Explainable Artificial Intelligence for Reducing the Global Cancer Burden

Explainable Artificial Intelligence for Reducing the Global Cancer Burden Authors: Fatemeh Hedayati Iran Polymer and Petrochemical Institute Reza Judi Chelan Yasuj University of Medical Sciences Mahfam Alijaniha Zanjan University of Medical Sciences Kamyab Komaee Koma Southeast University Parsa Irajian Iran University of Medical Sciences Negar Rajabi Azad University of Mashhad Maryam Khosravi Iran University of Science and Technology Kourosh Shahraki Zahedan University of Medical Sciences Behnoosh Omidvar Shiraz University Ahmad Jalali Lorestan University of Medical Sciences EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 1 Sepideh Sadat Babaei Islamic Azad UniversitySouth Tehran Branch Negin Rabiei Shiraz University of Medical Sciences Farbod Tabasi Kakhki Mashhad University of Medical Sciences Fatemeh Abediankenari Mazandaran University of Medical Sciences Omid Fakharzadeh Moghadam Mashhad University of Medical Sciences Behnam Hoorshad Iran University of Medical Sciences Sina Asadollahi Belarusian State Medical University Sara Razi Islamic Azad University of Tehran Medical Sciences Amir Arshia Beheshti Ardabil University of Medical Sciences Sina Montazeri University of North Texas Sadaf Taheri Loma Linda Dental University Nastaran Nemati University of Kentucky FATEMEH HEDAYATI 2 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 3 Book Details: Publisher: Kindle Publication Date: September 2025 Language: English Dimensions: 5 x 0.39 x 8 inches © Kindle and PreferPub 2025 ISBN-13: 979-8264785672 This peer-reviewed book is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specially the rights of translation, reprinting, result of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. FATEMEH HEDAYATI 4 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 5 Contents Chapter 1. Explainable AI in Reducing the Burden of Hematologic Malignancies 2. Explainable AI in Reducing the Burden of Neurological Cancers 3. Explainable AI in Reducing the Burden of Gastrointestinal Cancers 4. Explainable AI in Reducing the Burden of Breast and Lung Cancers 5. Explainable AI in Reducing the Burden of Urogenital Cancers 6. Explainable AI in Reducing the Burden of Skin and Soft Tissue Cancers 7. Explainable AI in Reducing the Burden of Head and Neck Cancers 8. Explainable AI in Reducing the Burden of Oral Cancers 9. Explainable AI in Reducing the Burden of Bone and Musculoskeletal Cancers 10. Explainable AI in Reducing the Burden of Rare and Other Cancers FATEMEH HEDAYATI 6 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 7 1. EXPLAINABLE AI IN REDUCING THE BURDEN OF HEMATOLOGIC MALIGNANCIES Background Hematologic malignancies, including leukemias, lymphomas, and myelomas, continue to represent a major cause of cancer-related morbidity and mortality worldwide. Their incidence rates are steadily increasing, and the diagnostic and therapeutic pathways remain complex and demanding. Despite significant progress in genomics, precision medicine, and targeted therapies, the clinical management of these malignancies remains challenging due to their biological heterogeneity and the intricate decision-making processes required for personalized care. Artificial Intelligence (AI), particularly machine learning (ML) and deep learning (DL) algorithms, has emerged as a promising approach to improve diagnostic accuracy, predict treatment responses, and stratify risk in hematologic cancers. Nevertheless, the so-called “black box” nature of many AI models has created obstacles to their adoption in clinical practice. Clinicians must be able to trust, interpret, and validate the outputs of such systems, particularly in high-stakes contexts 9 like oncology. This has generated growing interest in Explainable Artificial Intelligence (XAI), which seeks to make algorithmic decision-making more transparent, interpretable, and clinically actionable. Although numerous AI models have demonstrated impressive predictive power in hematologic malignancies, their lack of explainability significantly limits their practical use. Bridging the gap between algorithmic performance and interpretability remains a pressing challenge. In oncology, where opaque AI decisions could compromise patient safety and diminish trust among healthcare professionals and patients, the integration of XAI is not only a technical improvement but also a clinical and ethical necessity. Problem Statement A critical challenge in current AI applications for hematologic malignancies lies in their limited explainability, which impedes clinical adoption. While high-performing AI models exist, their opaque nature prevents meaningful incorporation into clinical decision-making workflows and limits their ability to reduce the clinical burden of blood cancers. Research Gap Despite the proliferation of AI models FATEMEH HEDAYATI 10 in hematologic oncology, few studies have systematically developed or implemented explainable frameworks that align with the cognitive and interpretive needs of hematologists. Most existing tools prioritize performance over transparency, resulting in a disconnect between model capability and clinical utility. Furthermore, domain-specific XAI approaches tailored to the pathophysiological and therapeutic complexity of hematologic malignancies remain scarce. Addressing this gap is essential to enable earlier diagnoses, optimize treatment strategies, and improve patient outcomes. The development of explainable AI approaches that are both accurate and interpretable has the potential to enhance clinician trust, support evidence-based decisionmaking, and facilitate the integration of AI into routine hematology practice. These approaches must be designed not only to maximize predictive performance but also to provide insights into how and why specific predictions are made, ensuring that AI contributes safely and effectively to patient care. Artificial Intelligence and Hematologic Malignancies Recent advances in artificial intelligence, particularly machine learning and deep learning, hold promise for transforming the management of hematologic malignancies. Machine learning EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 11 algorithms generate predictive models from training data and have been applied in pathology, radiology, genomics, and electronic health record analysis. Early applications primarily focus on automated classification of blood or bone marrow images, risk stratification, and genomic prediction. However, many modern deep learning models operate as “black boxes,” providing limited insight into the features driving their predictions. In high-stakes clinical environments, this opacity raises ethical, legal, and practical concerns, as clinicians cannot readily audit or trust the model’s reasoning, and patients and regulators demand transparency. Explainable artificial intelligence aims to address these challenges by providing humaninterpretable explanations of model outputs. XAI has the potential to reduce the burden of hematologic malignancies by enabling earlier diagnosis, enhancing prognostic assessment, guiding treatment selection, and improving resource allocation, while also addressing the ethical and practical limitations of traditional AI approaches. Why Explainability Matters in Hematology Ethical, Legal, and Clinical Considerations The opacity of many AI models creates a range FATEMEH HEDAYATI 12 of ethical and clinical challenges. Key aspects of explainability include identifying the addressee of explanations, whether clinician, patient, or regulator; ensuring explanations are relevant to decision-making; distinguishing between global and local explanations; balancing explainability with accuracy; addressing automation bias and epistemic authority; incorporating individual preferences and values; and considering the implications for patient autonomy and the clinician–patient relationship. AI systems must therefore be tailored to their specific context of use. For instance, a global explanation of model behavior may satisfy researchers but does little to assist a clinician in interpreting an individual patient’s result. The opacity of black-box models often arises from proprietary secrecy, technical complexity, and limited algorithmic literacy, making explainable AI a necessary strategy to facilitate trustworthy integration into clinical workflows. Critics note that XAI is not a universal solution. Adding post-hoc explanations to blackbox models may create a false sense of security. Rigorous internal and external validation of AI algorithms remains essential. In multiple myeloma, for example, clinicians must understand when and how to use AI tools and the degree of confidence they should place in their conclusions. Lack of transparency can erode EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 13 trust, and ethical implementation must ensure that AI does not disrupt the physician–patient relationship. Consequently, explainability must be complemented by robust validation, fairness assessment, and careful integration into practice. Potential Benefits of Explainable AI Despite these challenges, explainable AI offers substantial benefits in hematologic care. By clarifying the features that drive predictions, XAI can help identify novel biomarkers, validate biological hypotheses, and support personalized risk assessment. In multiple myeloma imaging, explainable models increase trustworthiness and allow expert evaluation; sharing code and data further enables external investigators to assess reliability. Explainable AI can reveal causal relationships and highlight new biomarkers. In chronic lymphocytic leukemia, an XAI algorithm analyzing multiparameter flow cytometry data identified seventeen cell populations predictive of inferior outcomes and improved risk stratification compared with the conventional CLL-IPI index. Such models not only enhance predictive accuracy but also direct clinicians toward specific cell populations or genes that might serve as therapeutic targets. XAI Applications in Hematologic Malignancies FATEMEH HEDAYATI 14 Explainable Morphologic Classification of Acute Myeloid Leukemia Deep learning has enabled automated classification of blood and bone marrow smears, but early models lacked transparency. The SingleCell-level Multiple Instance Learning Attention (SCEMILA) algorithm is an inherently explainable neural network designed for classifying acute myeloid leukemia subtypes from peripheral blood smears. SCEMILA employs a multiattention module that assigns attention scores to individual cells, making it possible to determine which cells contributed to the predicted label. Trained on eighty thousand single-cell images from 129 AML patients and sixty healthy controls, SCEMILA distinguished AML from healthy donors and achieved an F1-score of 0.86 for acute promyelocytic leukemia. Importantly, cells receiving high attention corresponded with those selected by human experts, demonstrating alignment between model reasoning and expert evaluation. By highlighting subtype-specific morphologic features without requiring cell-level labels, SCEMILA enables pathologists to verify the cells driving the classification and to detect potential errors. Such inherent explainability is critical for building trust and may accelerate adoption in clinical workflows. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 15 XAI for Prognostication in Chronic Lymphocytic Leukemia Prognostication in chronic lymphocytic leukemia traditionally relies on clinical staging and genetic markers, which may not fully capture disease heterogeneity. The ALPODS algorithm, a locally interpretable point-wise linear classifier, was applied to multiparameter flow cytometry data from 157 CLL patients. The model identified seventeen cell populations, including a CD4positive T-cell subset, associated with inferior outcomes and achieved an area-under-the-curve of 0.95, outperforming the CLL-IPI index, which had an AUC of 0.78. ALPODS provided interpretable explanations by highlighting the cell populations that contributed most to the risk score. When the identified CD4-positive Tcell population was integrated into the CLLIPI, predictive performance improved to an AUC of 0.83. This example illustrates how XAI can refine prognostic scores, uncover pathobiological insights such as immune cell involvement, and suggest potential therapeutic targets. Gene Selection and Time-toTherapy Prediction in Chronic Lymphocytic Leukemia Genomics provides another important arena for explainable artificial intelligence. Morabito and colleagues proposed the DeepSHAP Autoencoder FATEMEH HEDAYATI 16 Filter for Gene Selection (DSAF-GS), a deeplearning-based feature-selection method that combines an autoencoder with SHapley Additive exPlanations (SHAP) to identify genes influencing prognosis in chronic lymphocytic leukemia (CLL). Applied to a gene-expression dataset of 217 CLL cases encompassing approximately twenty thousand genes, DSAF-GS achieved 86.4 percent accuracy, with 85 percent sensitivity and 87.5 percent specificity. SHAP-based explanations revealed that predictions were strongly influenced by CEACAM19 and PIGP, moderately influenced by MKL1 and GNE, and less influenced by other genes. The ten most influential genes identified included FADD, IGF1R, GNE, and MKL1, which are involved in signal transduction, cell-cycle regulation, and apoptosis pathways. Incorporating these top genes into a multivariable model improved Harrell’s cindex and explained variation for time-to-firsttreatment compared with a basic prognostic model. Explainable Models in Multiple Myeloma Imaging Diagnosis and response assessment in multiple myeloma often rely on imaging modalities such as MRI, CT, or PET, along with morphologic criteria. A systematic review of AI in multiple myeloma imaging emphasized the need for explainable models to increase trustworthiness and allow experts to evaluate model decisions. RadiomicsEXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 17 Grad-CAM to highlight key image regions used for tumor identification. SECNN-MNet has demonstrated superior performance and enhanced explainability compared to previous CNN and ResNet-50 models. In several medical centers, an innovative model based on the GoogleNet architecture has been developed to extract complex features and data from brain MRI images of patients across four classes: glioma, meningioma, pituitary tumors, and non-tumor cases. Only the information derived from unique image features, including weights and gradients, is transmitted to a central server while preserving patient privacy from each medical center. The central server, utilizing federated learning, integrates data from all clients to design a global model, which is then distributed back to the clients. These clients further train the model on their local data, and the repetition of this process enhances the model’s accuracy. Alongside federated learning, this model employs XAI techniques such as Grad-CAM to display critical image regions as heatmaps and saliency maps to highlight the role of pixels in the decisionmaking process, enabling clinicians to gain a visual understanding of the model’s performance. To assess user satisfaction with the XAI system, quantitative metrics such as model and technique accuracy, interpretability, and response time, as well as qualitative tools such as the System Usability Score or NASA-TLX, are utilized, serving FATEMEH HEDAYATI 30 as an evaluation of the effectiveness of XAI. Overall, studies have demonstrated that explainable artificial intelligence plays a significant role in the accurate and reliable detection of brain tumors by providing a transparent and comprehensible explanation of how models interpret and predict unique tumor features in MRI images, aligning with clinical acceptance by medical specialists. The Complexity of Neurological Cancers Neurological cancers are distinct from other malignancies due to their location in a highly sensitive organ and the limited capacity for surgical resection without risking neurological deficits. Brain tumors often exhibit intratumoral heterogeneity, with diverse cell populations coexisting within the same lesion, making therapeutic targeting difficult. Furthermore, the blood-brain barrier (BBB) limits drug delivery, reducing the effectiveness of systemic therapies. Patients with neurological cancers frequently experience profound neurological symptoms, including seizures, personality changes, language deficits, and cognitive impairment, which diminish quality of life. Treatment itself can exacerbate these issues. Radiotherapy, while effective in slowing tumor progression, may cause long-term neurocognitive decline. Chemotherapeutics such as temozolomide may EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 31 improve survival but also carry significant side effects. Therefore, therapeutic strategies require careful balancing between efficacy and preservation of neurological function. The complexity extends to diagnosis and prognostication. Conventional imaging may not adequately distinguish between tumor recurrence and treatment-induced changes, such as pseudoprogression or radiation necrosis. Molecular biomarkers, including IDH mutations and MGMT promoter methylation, have proven valuable for prognosis and therapy selection, but their interpretation is challenging when integrated with clinical and imaging data. Clinicians must navigate these complexities rapidly while ensuring decisions are evidencebased. The Role of Artificial Intelligence in Neurological Oncology Artificial intelligence has demonstrated substantial potential in neurological oncology by processing large, multimodal datasets that surpass human interpretive capacity. Deep learning models trained on imaging data can achieve high accuracy in classifying tumor types, segmenting lesions, and predicting treatment outcomes. Radiogenomics, an emerging field combining imaging features with genomic data, leverages AI to non-invasively predict molecular FATEMEH HEDAYATI 32 subtypes of tumors, enabling precision medicine approaches without requiring invasive biopsies. In surgical planning, AI-based tools provide 3D reconstructions and predictive maps of tumor boundaries, improving surgical precision and reducing postoperative deficits. During radiotherapy, AI algorithms assist in dose optimization by accurately contouring tumor margins and sparing healthy tissue. In clinical decision-making, predictive models estimate overall survival and recurrence risks, guiding patient-specific treatment strategies. Despite these achievements, the majority of AI models operate as opaque systems, offering limited insight into how decisions are made. This opacity poses risks in medicine, where incorrect predictions can have life-threatening consequences. Clinicians are hesitant to trust systems they cannot interpret, particularly when model outputs conflict with established clinical knowledge. Here lies the importance of explainability, which ensures that AI is not only accurate but also transparent and clinically actionable. Foundations of Explainable AI Explainable AI encompasses a set of techniques designed to make AI models interpretable and understandable. Interpretability refers to the degree to which a human can comprehend how EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 33 a model processes input to produce output. Transparency ensures that the rationale behind a model’s decision is visible and justifiable. Techniques in XAI range from inherently interpretable models, such as decision trees and generalized linear models, to post hoc explanations for complex models, such as deep neural networks. Key methods in XAI include feature importance analysis, which highlights the variables most influential in a decision; saliency maps, which identify regions of images driving predictions; local interpretable model-agnostic explanations (LIME), which approximate model behavior for individual predictions; and SHAP (Shapley additive explanations), which attribute contributions of each feature to an output. These methods allow clinicians to assess whether AI models base decisions on clinically relevant factors or spurious correlations. In neurological oncology, saliency maps may reveal whether an AI model relies on tumor margins, edema patterns, or unrelated image artifacts. SHAP values may show the influence of genomic markers such as IDH mutation status on survival predictions. By providing such clarity, XAI helps clinicians integrate AI insights into their decision-making process while maintaining accountability and trust. FATEMEH HEDAYATI 34 Applications of Explainable AI in Imaging of Neurological Cancers Medical imaging remains the cornerstone of neurological cancer management, from diagnosis to treatment monitoring. Explainable AI enhances imaging-based AI tools by enabling clinicians to verify that predictions align with known radiological features. For tumor detection, saliency maps allow radiologists to visualize which areas of the brain MRI influenced the classification of a glioma versus metastasis. This transparency reduces diagnostic uncertainty and ensures that AI models do not misinterpret noise as pathology. In tumor segmentation, explainability confirms that algorithms correctly delineate tumor boundaries and edema, which are crucial for surgical and radiotherapy planning. Another promising application is differentiating pseudoprogression from true tumor recurrence. Conventional imaging often fails to make this distinction, leading to inappropriate treatment changes. XAI-driven models highlight imaging features, such as contrast-enhancement patterns and diffusion abnormalities, that underpin predictions. Clinicians can assess these features for plausibility, increasing confidence in adopting AI recommendations. Furthermore, radiogenomics benefits from XAI EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 35 by linking imaging features to genetic alterations. For example, an explainable model predicting IDH mutation from MRI scans can indicate which tumor regions reflect mutational status, supporting biologically meaningful interpretations. This enhances the utility of noninvasive diagnostics and facilitates personalized therapeutic strategies. Genomic and Molecular Applications of Explainable AI Molecular profiling has revolutionized the classification and treatment of neurological cancers. However, genomic datasets are highdimensional and complex, requiring advanced computational approaches. Explainable AI offers a pathway to extract actionable insights while maintaining transparency. Predictive models built on genomic data can identify molecular signatures associated with prognosis or therapeutic response. By applying SHAP values or feature attribution methods, clinicians can understand which mutations, gene expression patterns, or epigenetic markers drive survival predictions. This prevents over-reliance on uninterpretable black-box models and supports biologically coherent hypotheses. For example, in glioblastoma, an explainable model may reveal that MGMT promoter methylation contributes significantly to FATEMEH HEDAYATI 36 predictions of temozolomide response, aligning with established clinical evidence. Conversely, if the model attributes importance to genes with no known association, clinicians can question its validity and avoid erroneous conclusions. Such transparency fosters trust and facilitates clinical translation of AI-driven genomic insights. XAI also enables integration of multimodal data, combining imaging, genomics, and clinical parameters. By making the contributions of each data type explicit, XAI models allow clinicians to understand how different factors interact in shaping predictions. This holistic understanding is particularly valuable in neurological cancers, where decisions must consider both biological and functional outcomes. Enhancing Clinical DecisionMaking with Explainable AI Clinical decision-making in neurological oncology involves high-stakes choices regarding surgery, radiotherapy, and systemic treatment. Explainable AI empowers clinicians by providing not only predictions but also the reasoning behind them. When predicting surgical outcomes, an explainable model can show that tumor location near eloquent brain areas drives a higher risk of postoperative deficits. Such insights help surgeons plan procedures with greater awareness of functional risks. In radiotherapy planning, EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 37 explainable dose-optimization models highlight which tissue-sparing priorities influenced recommended dosing, supporting informed tradeoffs. In survival prediction, models that indicate which clinical and biological factors contribute most strongly to prognosis allow clinicians to communicate predictions transparently to patients and families. This fosters shared decisionmaking, where patients understand the rationale behind recommended therapies and are more likely to adhere to treatment plans. Explainable AI also supports multidisciplinary tumor boards by providing interpretable evidence that integrates diverse data sources. Oncologists, radiologists, neurosurgeons, and pathologists can collectively evaluate model explanations, ensuring that recommendations align with clinical expertise. This reduces the risk of overreliance on any single clinician or algorithm and enhances the robustness of care. Reducing Healthcare Burden Through Explainable AI The burden of neurological cancers extends beyond individual patients to healthcare systems and societies. High costs of care, frequent hospitalizations, and the need for long-term rehabilitation strain resources. Explainable AI can mitigate this burden by improving efficiency, FATEMEH HEDAYATI 38 reducing diagnostic errors, and optimizing treatment strategies. Early diagnosis facilitated by XAI-supported imaging models reduces delays in initiating treatment, improving survival outcomes and lowering costs of advanced disease management. Accurate differentiation between recurrence and pseudoprogression prevents unnecessary therapies, minimizing side effects and financial waste. Personalized treatment recommendations informed by transparent AI models ensure that resources are directed toward interventions most likely to benefit individual patients. XAI also reduces medicolegal risks associated with AI adoption. By providing interpretable justifications for predictions, XAI safeguards against liability concerns and enhances regulatory compliance. Healthcare providers and payers are more likely to adopt AI tools when their outputs are transparent and defensible. Furthermore, explainability fosters patient trust, which is essential for adherence to treatment and engagement in care. Patients who understand the reasoning behind AI-driven recommendations are more likely to accept interventions, reducing dropout rates and improving long-term outcomes. Collectively, these benefits reduce the overall burden of neurological cancers at both patient and societal levels. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 39 Ethical and Regulatory Dimensions of Explainable AI in Neurological Cancers The ethical integration of AI into healthcare requires transparency, accountability, and fairness. Black-box models that make inscrutable decisions raise concerns about bias, safety, and trustworthiness. Explainable AI addresses these issues by making decision-making processes visible and justifiable. In neurological cancers, where decisions profoundly affect survival and quality of life, ethical responsibility is paramount. XAI ensures that models do not inadvertently rely on confounding factors such as demographic biases or imaging artifacts. By exposing feature importance, XAI allows developers and clinicians to identify and correct biases, promoting equitable care. Regulatory agencies increasingly emphasize explainability as a prerequisite for AI approval in healthcare. Transparent models are easier to validate and audit, accelerating their clinical translation. By aligning with regulatory requirements, XAI accelerates safe deployment of AI tools in neurological oncology. Future Directions of Explainable AI in Neurological Cancers FATEMEH HEDAYATI 40 The future of XAI in neurological cancers lies in developing more sophisticated, userfriendly tools that integrate seamlessly into clinical workflows. Advances in visual explanation methods, interactive dashboards, and natural language generation will make model reasoning more accessible to non-technical clinicians. Multimodal explainability will gain prominence, enabling simultaneous interpretation of imaging, genomic, pathology, and clinical data. This holistic perspective will reflect the multifactorial nature of neurological cancers and support precision oncology. Collaborative platforms integrating XAI into tumor boards, electronic health records, and decision-support systems will enhance multidisciplinary care. As more real-world evidence accumulates, XAI models will continue to evolve, improving accuracy and trustworthiness. Ultimately, explainable AI has the potential to transform neurological oncology by reducing diagnostic delays, optimizing therapies, and empowering patients and clinicians alike. By addressing the transparency gap, XAI paves the way for responsible and effective use of AI in combating one of the most challenging categories of human malignancies. Conclusion EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 41 Neurological cancers remain among the most devastating diseases, with profound impacts on patients, families, and healthcare systems. While AI offers unprecedented opportunities to enhance diagnosis, treatment, and prognosis, its adoption has been limited by concerns over opacity and trust. Explainable AI addresses this challenge by making model outputs transparent, interpretable, and clinically meaningful. By illuminating the reasoning behind predictions, XAI empowers clinicians to make informed decisions, enhances patient trust, reduces diagnostic and therapeutic errors, and fosters ethical integration of AI into practice. Its applications in imaging, genomics, and clinical decision-making demonstrate tangible benefits in reducing the burden of neurological cancers. As research advances, XAI will play a central role in bridging the gap between cutting-edge computational tools and humancentered clinical care. Through transparency and accountability, explainable AI ensures that technological innovation translates into realworld improvements in outcomes, ultimately alleviating the immense burden of neurological cancers. FATEMEH HEDAYATI 42 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 43 3. EXPLAINABLE AI IN REDUCING THE BURDEN OF GASTROINTESTINAL CANCERS Background Gastrointestinal cancers are among the most prevalent and deadly forms of malignancy worldwide, encompassing a wide range of tumors affecting the esophagus, stomach, pancreas, liver, gallbladder, and colon. These cancers account for a significant proportion of global cancer mortality, with colorectal, gastric, liver, and pancreatic cancers ranking among the top causes of cancerrelated deaths. Their impact extends beyond individual patients, burdening healthcare systems with high diagnostic costs, complex treatments, and long-term palliative care needs. The challenges of gastrointestinal cancers lie in their insidious onset, late diagnosis, and biological heterogeneity. Early stages often present with vague symptoms such as abdominal discomfort or dyspepsia, which can be mistaken for benign conditions. As a result, many patients are diagnosed only at advanced stages, where curative treatment options are limited. Endoscopic screening and imaging modalities such as computed tomography, magnetic resonance imaging, and positron emission tomography play 44 pivotal roles in detection and staging, yet interpretation is time-consuming and subject to inter-observer variability. Molecular profiling has revealed that gastrointestinal cancers are driven by diverse genetic and epigenetic alterations, with certain biomarkers such as KRAS mutations in colorectal cancer or HER2 overexpression in gastric cancer influencing prognosis and therapy. However, integrating genomic insights into clinical practice remains challenging due to the overwhelming complexity and volume of data generated by sequencing technologies. Artificial intelligence, particularly machine learning and deep learning, has emerged as a transformative force in oncology. AI systems can detect subtle imaging features, analyze vast genomic datasets, and generate predictive models for survival, recurrence, and treatment response. In gastrointestinal cancers, AI has been applied to tasks such as detecting polyps in colonoscopy, differentiating benign from malignant liver lesions, and predicting outcomes of chemotherapy or immunotherapy. Despite its promise, AI adoption has been limited by the “black box” nature of many models. Clinicians and patients are hesitant to trust predictions when the underlying reasoning is hidden. Explainable AI (XAI) addresses this limitation by creating models whose decisionEXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 45 making process is transparent and interpretable. XAI allows healthcare providers to understand why a model recommended a certain diagnosis, highlighted a suspicious lesion, or predicted a specific treatment outcome. This interpretability builds trust, supports accountability, and ensures that AI complements rather than replaces human expertise. In the context of gastrointestinal cancers, explainable AI can reduce the burden by enabling earlier diagnosis, improving therapeutic decisions, and guiding personalized care while maintaining transparency. Its integration into endoscopy, imaging, genomics, and clinical workflows provides a pathway toward more effective and ethical cancer management. The Burden of Gastrointestinal Cancers The global burden of gastrointestinal cancers is immense. Colorectal cancer remains one of the most commonly diagnosed malignancies worldwide and a leading cause of mortality, particularly in high-income countries. Gastric cancer, though declining in incidence in certain regions, continues to be prevalent in East Asia and carries a poor prognosis. Hepatocellular carcinoma associated with chronic viral hepatitis and metabolic disorders imposes a heavy toll in both Asia and Africa. Pancreatic cancer, despite FATEMEH HEDAYATI 46 being less common, is notorious for its lethality due to late detection and aggressive biology. Patients face not only the risk of death but also significant morbidity. Symptoms such as gastrointestinal bleeding, pain, malnutrition, and cachexia severely impair quality of life. Treatments including surgery, chemotherapy, radiotherapy, and immunotherapy may prolong survival but often carry debilitating side effects. The economic burden is equally substantial, involving direct medical costs, indirect costs from lost productivity, and the emotional toll on families. The complexity of gastrointestinal cancers stems from their heterogeneity at multiple levels. Morphological differences complicate pathological classification, while diverse genomic and epigenomic alterations influence treatment responses. Tumor microenvironments vary widely, with immune cell infiltration and stromal interactions shaping prognosis. These factors necessitate advanced tools capable of integrating and interpreting multidimensional data, which is where explainable AI becomes crucial. The Role of Artificial Intelligence in Gastrointestinal Oncology Artificial intelligence has shown enormous promise in gastrointestinal oncology by augmenting diagnostic accuracy, improving risk EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 47 stratification, and personalizing therapies. Deep learning models trained on endoscopic videos detect polyps with remarkable sensitivity, reducing miss rates compared with traditional human observation. AI-enhanced imaging techniques differentiate between malignant and benign lesions with higher precision, assisting radiologists in liver or pancreatic cancer diagnosis. In pathology, AI algorithms analyze digital slides to identify histological patterns predictive of tumor grade or aggressiveness. In genomics, machine learning processes large-scale sequencing data to uncover mutational signatures associated with prognosis or therapeutic response. Predictive models combining clinical and molecular variables provide individualized survival estimates, enabling more informed treatment planning. Despite these advances, most AI systems operate as opaque models that provide outputs without explanations. While they may achieve high accuracy, clinicians remain reluctant to rely on predictions they cannot interpret, particularly in high-stakes decisions involving cancer care. Without transparency, AI risks being perceived as untrustworthy or even unsafe. Foundations of Explainable AI Explainable AI seeks to bridge the gap between accuracy and transparency. Interpretability refers FATEMEH HEDAYATI 48 to how easily a human can understand the internal mechanics of an AI system, while explainability encompasses both the interpretability and the ability to communicate reasoning to end-users. There are two main strategies in XAI. The first is building inherently interpretable models, such as decision trees, generalized linear models, and rule-based systems, where reasoning is transparent by design. The second involves post hoc explanations applied to complex models like deep neural networks. These include techniques such as feature importance ranking, local interpretable model-agnostic explanations (LIME), SHAP (Shapley additive explanations), saliency maps, and counterfactual explanations. For example, in an AI system detecting gastric cancer from endoscopic images, a saliency map can highlight the exact lesion area influencing the decision. In survival prediction models, SHAP values can show which clinical features or gene mutations contributed most to an individual patient’s prognosis. This transparency ensures that clinicians can verify the plausibility of AI outputs and detect potential biases. Applications of Explainable AI in Endoscopy Endoscopy is central to detecting and managing gastrointestinal cancers. Colonoscopy remains the EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 49 and social determinants. For example, Poissonbased approaches often assume homogeneous risk levels across populations, which can lead to biased predictions when environmental or individual variations exist. Although advanced probabilistic models such as Poisson-Gamma can address issues like overdispersion, they may still lack sufficient flexibility to represent complex multivariate dependencies. Furthermore, spatial and socioeconomic heterogeneity exerts strong influence on cancer outcomes, yet regionspecific and nonlinear interactions are often underrepresented in traditional models. The interplay between behavioral, environmental, and social factors requires models capable of adapting to contextual variations and hidden correlations, which conventional statistical frameworks frequently fail to accommodate. Machine Learning and Public Health Integration In contrast, machine learning (ML) techniques can effectively model these relationships, providing more tailored regional predictions and supporting timely public health responses. These technologies play an essential role in developing comprehensive strategies for disease prevention and management. Such approaches are especially effective in integrating diverse data sources, strengthening risk stratification, and informing well-rounded disease prevention programs. By FATEMEH HEDAYATI 62 doing so, ML not only advances predictive accuracy but also broadens the scope of public health integration. Importance of Explainability (XAI) in Clinical Practice To ensure that AI-driven insights are interpretable and actionable, the integration of XAI into medical applications is indispensable. AI has shown remarkable promise across healthcare, including in diagnostics, imaging, drug discovery, and personalized treatment planning. However, its adoption is frequently hindered by opaque decision-making processes. XAI enhances transparency and fosters trust by helping clinicians and patients understand, monitor, and validate AI-generated recommendations. This approach supports fairer healthcare delivery by detecting and mitigating biases, strengthening equity, and empowering individuals. The case of breast cancer serves as a clear example where explainability is especially important, underscoring the broader role of XAI across medical domains. Conclusion This review emphasizes the transformative role of XAI in improving the diagnostic and prognostic management of breast and lung cancers. While AI technologies have made significant advancements in medical imaging and mortality prediction, their EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 63 application in clinical practice is often constrained by their opaque black-box characteristics. XAI bridges this gap by making AI models more interpretable and trustworthy, thereby enabling clinicians to make informed decisions based on algorithmic outputs. In high-stakes fields such as oncology, this level of clarity is essential not only for patient safety but also for building confidence among clinicians, patients, and healthcare institutions. Moreover, XAI contributes to healthcare equity by uncovering biases and adapting models to diverse populations and settings. In this way, XAI is not merely a technical enhancement but a fundamental step toward ethical, transparent, and effective cancer care. FATEMEH HEDAYATI 64 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 65 5. EXPLAINABLE AI IN REDUCING THE BURDEN OF UROGENITAL CANCERS Background Urogenital cancers represent a diverse and challenging group of malignancies that affect the kidneys, bladder, prostate, testes, and other structures of the urinary and reproductive systems. These cancers account for a significant proportion of the global cancer burden, with prostate and bladder cancers ranking among the most frequently diagnosed malignancies in men, while kidney cancer contributes substantially to cancer-related mortality worldwide. Testicular cancer, though relatively rare, is the most common malignancy in young adult males, and its treatment outcomes depend heavily on timely detection and individualized therapy. The burden of urogenital cancers is compounded by late detection, complex tumor biology, and the long-term impact of treatment on quality of life. For instance, prostate cancer may present as an indolent disease in some men, while in others it progresses aggressively with metastatic potential. Differentiating between these forms is critical for guiding therapy but remains difficult with current clinical tools. Similarly, bladder 66 cancer is characterized by high recurrence rates, necessitating frequent surveillance cystoscopies that are invasive and costly. Kidney cancers, particularly renal cell carcinoma, often present incidentally but can progress silently until advanced stages. The heterogeneity of these cancers across genetic, molecular, and clinical dimensions complicates treatment decisions. While precision medicine approaches, such as the use of targeted therapies and immunotherapies, have improved outcomes in certain subgroups, their effectiveness varies among patients. Clinicians face the ongoing challenge of integrating imaging, pathology, genomic, and clinical data into cohesive treatment strategies. Artificial intelligence has emerged as a transformative tool in oncology by enabling automated image analysis, predictive modeling, and integration of large-scale datasets. In urogenital cancers, AI has shown promise in tasks such as prostate cancer detection on magnetic resonance imaging, bladder tumor segmentation, and prediction of treatment response based on histopathological and genomic features. Yet, the adoption of AI has been hindered by the opacity of many models. Clinicians often hesitate to rely on systems whose decision-making processes remain hidden, especially in high-stakes clinical contexts where transparency is essential. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 67 Explainable artificial intelligence addresses this challenge by creating models that not only perform well but also provide interpretable and transparent reasoning. This interpretability fosters trust among clinicians and patients, ensures accountability, and facilitates regulatory approval. In the context of urogenital cancers, explainable AI can reduce the disease burden by improving diagnostic accuracy, guiding treatment selection, and enhancing patient engagement, all while ensuring that decision-making remains clinically meaningful and ethically sound. The Burden of Urogenital Cancers Urogenital cancers impose a profound global health burden, both in terms of incidence and mortality, as well as their economic and psychosocial impact. Prostate cancer is the most frequently diagnosed cancer among men in many regions, with incidence rates rising due to aging populations and increased use of prostatespecific antigen testing. Despite improvements in survival, prostate cancer continues to cause significant morbidity, particularly when advanced disease leads to bone metastases, urinary dysfunction, and sexual impairment. Bladder cancer is a costly malignancy to manage, primarily due to its high recurrence rates and the need for lifelong surveillance. Patients often undergo repeated cystoscopies, FATEMEH HEDAYATI 68 biopsies, and treatments such as intravesical therapy, which add to the financial and physical burden. Kidney cancers contribute substantially to mortality, with renal cell carcinoma being the most common subtype. These tumors are often detected incidentally during imaging for unrelated conditions, yet advanced cases carry a poor prognosis despite recent therapeutic advances. Testicular cancer, while relatively rare, has a unique psychosocial burden due to its impact on younger men at the peak of their productive years. Treatment, although often curative, can affect fertility, hormonal balance, and psychological well-being. The clinical complexity of these cancers often results in challenging treatment decisions. For example, distinguishing between indolent and aggressive prostate cancer is crucial to avoid overtreatment or undertreatment. Similarly, selecting systemic therapies for advanced kidney or bladder cancers requires weighing potential benefits against toxicity and cost. These challenges necessitate tools that can integrate vast amounts of clinical, imaging, and genomic information to support evidence-based decisionmaking, a need that explainable AI is wellpositioned to meet. The Role of Artificial Intelligence EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 69 in Urogenital Oncology Artificial intelligence has demonstrated considerable potential in enhancing the management of urogenital cancers. Deep learning models applied to multiparametric MRI have improved prostate cancer detection, enabling identification of clinically significant tumors that might be missed by radiologists. AI algorithms trained on histopathological slides can classify prostate or bladder tumor grade with high accuracy, assisting pathologists in reducing interobserver variability. Radiomics, the extraction of quantitative features from medical imaging, has been applied to kidney and bladder cancers to predict tumor stage, grade, and treatment response. When combined with machine learning, radiomics can reveal subtle imaging biomarkers invisible to the human eye. In genomics, AI algorithms have been used to identify prognostic signatures from large sequencing datasets, guiding precision therapies. Clinical decision support systems powered by AI can integrate multimodal data to provide individualized risk predictions. For instance, models predicting biochemical recurrence after prostatectomy or response to immunotherapy in renal cell carcinoma can help tailor treatment strategies. Despite these advances, the black-box nature of many AI models limits clinical adoption. Without clear explanations, FATEMEH HEDAYATI 70 clinicians remain cautious about incorporating AI recommendations into practice, particularly when model predictions conflict with established guidelines or clinical judgment. Foundations of Explainable AI Explainable AI aims to make the decisionmaking processes of AI systems transparent and interpretable. Interpretability refers to the ability of humans to understand how inputs are transformed into outputs, while explainability encompasses the capacity to communicate this reasoning effectively to users. There are two broad approaches to explainability. The first involves inherently interpretable models, such as logistic regression, decision trees, or rule-based systems, where the reasoning is straightforward. The second involves post hoc explanation techniques applied to complex models like deep neural networks. Methods such as feature importance rankings, local interpretable model-agnostic explanations, Shapley additive explanations, and saliency maps allow users to understand which features most influenced a given prediction. In urogenital cancers, explainability ensures that AI predictions are aligned with clinical knowledge. For example, a model predicting aggressive prostate cancer should highlight relevant features such as PSA level, MRI lesion characteristics, and EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 71 Gleason score rather than spurious correlations. Similarly, an AI system for bladder tumor recurrence should base its predictions on known risk factors such as tumor grade and multiplicity, not irrelevant data artifacts. By making reasoning transparent, XAI builds confidence and facilitates integration into clinical decision-making. Applications of Explainable AI in Imaging Imaging is central to the management of urogenital cancers, from detection to treatment monitoring. Multiparametric MRI is widely used for prostate cancer diagnosis, yet interpretation requires expertise and is subject to variability. AI models have demonstrated high accuracy in detecting clinically significant tumors, and XAI enhances these models by showing radiologists which regions of the prostate influenced predictions. Saliency maps can highlight suspicious lesions, allowing radiologists to validate AI reasoning and avoid false positives caused by artifacts. In bladder cancer, imaging plays a crucial role in staging and monitoring. AI-based segmentation models can delineate bladder tumors on CT or MRI, and explainability ensures that the contours are clinically appropriate. For kidney cancer, radiomics-based models predicting histological subtype or response to therapy can use feature FATEMEH HEDAYATI 72 attribution methods to show which imaging characteristics drive predictions, aligning results with radiological expertise. These explainable imaging tools reduce inter-observer variability, enhance diagnostic confidence, and support more accurate treatment planning, ultimately reducing the burden of disease. Pathology and Explainable AI Histopathology remains the gold standard for diagnosing urogenital cancers. However, pathologists face growing workloads as cancer incidence rises. AI algorithms analyzing digital pathology slides can identify tumor regions, grade prostate cancer, or classify bladder cancer subtypes. Explainable AI ensures that these classifications are based on relevant morphological features. For example, in prostate cancer grading, an XAI model can highlight glandular structures and cellular patterns that drove its classification of Gleason grade. This allows pathologists to verify model outputs, reducing the risk of misclassification. In bladder cancer, explainable models can show which histological patterns indicate muscle invasion, a critical determinant of treatment strategy. Such transparency enhances confidence in AI-assisted pathology and facilitates adoption in clinical practice. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 73 Genomic and Molecular Applications of Explainable AI Genomic profiling has revealed numerous biomarkers in urogenital cancers. In prostate cancer, mutations in DNA repair genes such as BRCA1 and BRCA2 predict sensitivity to PARP inhibitors. In kidney cancer, mutations in VHL and PBRM1 influence disease biology and therapeutic response. In bladder cancer, alterations in FGFR3 and TP53 have prognostic and therapeutic implications. AI models can integrate these complex datasets to predict prognosis and therapy response. Explainable AI ensures that such predictions are grounded in meaningful biology. For instance, an XAI model predicting immunotherapy response in renal cell carcinoma might reveal that gene expression profiles of immune checkpoint pathways were key drivers, aligning with clinical expectations. In bladder cancer, XAI could highlight the contribution of FGFR3 mutations in predicting response to targeted therapy, supporting clinical decision-making. By clarifying how genomic features influence predictions, XAI facilitates the translation of complex molecular insights into actionable clinical strategies. Clinical Decision-Making FATEMEH HEDAYATI 74 with Explainable AI Treatment of urogenital cancers requires careful balancing between efficacy, toxicity, and quality of life. Explainable AI enhances clinical decisionmaking by providing transparent predictions that clinicians can interpret and trust. In prostate cancer, models predicting disease progression after active surveillance can indicate which variables—such as PSA velocity, MRI findings, and biopsy results—drive predictions. This helps clinicians and patients weigh the risks of surveillance versus definitive treatment. In bladder cancer, XAI models predicting recurrence risk can show whether tumor grade, stage, or patient comorbidities contributed most, enabling personalized surveillance protocols. In advanced kidney cancer, explainable models predicting response to immunotherapy can clarify the role of clinical and molecular features, helping oncologists choose between immune checkpoint inhibitors, targeted therapies, or combination regimens. Such interpretability fosters shared decision-making, where patients can understand the rationale behind recommendations and actively participate in their care. Reducing Healthcare Burden through Explainable AI The healthcare burden of urogenital cancers includes high costs of treatment, EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 75 repeated surveillance procedures, and long-term management of side effects. Explainable AI can reduce this burden by improving diagnostic efficiency, minimizing unnecessary interventions, and guiding optimal therapy selection. For example, in prostate cancer, XAI-assisted MRI interpretation reduces unnecessary biopsies by clarifying which lesions warrant further investigation. In bladder cancer, transparent recurrence prediction models help tailor surveillance intervals, reducing the frequency of unnecessary cystoscopies while ensuring timely detection of recurrences. In kidney cancer, explainable prediction of therapy response prevents patients from undergoing costly and toxic treatments unlikely to benefit them. By reducing overtreatment, avoiding ineffective therapies, and enhancing resource allocation, XAI contributes to more sustainable healthcare delivery while improving patient outcomes. Ethical and Regulatory Considerations Ethical deployment of AI in urogenital oncology requires transparency, accountability, and fairness. Black-box models risk perpetuating biases based on demographic or socioeconomic factors. Explainable AI mitigates these risks by making model reasoning visible, enabling developers and clinicians to identify and correct FATEMEH HEDAYATI 76 biases. From a regulatory perspective, agencies increasingly emphasize explainability as a requirement for clinical approval of AI systems. Transparent models are easier to validate, audit, and monitor, accelerating their integration into clinical workflows. Explainability also supports informed consent by enabling clinicians to explain AI-derived recommendations to patients, empowering them to make informed choices. Future Directions of Explainable AI in Urogenital Cancers The future of XAI in urogenital cancers lies in multimodal integration, where imaging, pathology, genomics, and clinical data are combined into comprehensive predictive models. Explainability will be crucial to ensure that the contributions of each modality are clear and clinically meaningful. Advances in visualization techniques and userfriendly interfaces will make XAI outputs more accessible to clinicians. Interactive dashboards could allow oncologists to explore how different variables affect predictions, fostering deeper understanding and trust. Explainable AI will also play a role in precision oncology clinical trials, guiding patient selection and monitoring treatment response. By clarifying model reasoning, XAI enhances trial transparency EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 77 and accelerates drug development. Integration into electronic health records and clinical decision support systems will ensure that XAI becomes part of routine practice, improving efficiency and equity in cancer care. Conclusion Urogenital cancers impose a significant global burden due to their prevalence, heterogeneity, and long-term management challenges. Artificial intelligence offers transformative potential in enhancing diagnosis, prognosis, and treatment, yet adoption is limited by the opacity of many models. Explainable AI addresses this challenge by providing transparency, interpretability, and accountability. By clarifying how predictions are made, XAI fosters trust, supports clinical decisionmaking, and enhances patient engagement. Its applications span imaging, pathology, genomics, and treatment planning, offering tangible benefits in reducing misdiagnosis, optimizing therapies, and alleviating healthcare costs. As explainable AI continues to evolve, its integration into urogenital oncology holds the promise of transforming cancer care, reducing disease burden, and advancing toward a future where technology and human expertise work together to achieve better outcomes for patients worldwide. FATEMEH HEDAYATI 78 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 79 6. EXPLAINABLE AI IN REDUCING THE BURDEN OF SKIN AND SOFT TISSUE CANCERS Background Skin and soft tissue cancers represent a significant global health concern, encompassing a wide spectrum of malignant conditions that include melanoma, non-melanoma skin cancers such as basal cell carcinoma and squamous cell carcinoma, and rare but aggressive malignancies like sarcomas. Collectively, these cancers impose a considerable clinical, societal, and economic burden. Skin cancers alone are among the most common malignancies worldwide, with their incidence steadily increasing due to risk factors such as ultraviolet radiation exposure, aging populations, and lifestyle patterns. Meanwhile, soft tissue sarcomas, though less common, are particularly challenging due to their heterogeneity, late presentation, and high recurrence rates. Traditional approaches to diagnosing, treating, and monitoring these cancers rely on dermatologic examinations, imaging modalities, biopsy, histopathological evaluation, and a combination of surgery, radiotherapy, and systemic therapies. While these methods have 80 advanced substantially in the past decades, they still face considerable limitations. Clinical diagnosis often depends on subjective evaluation by physicians, leading to variability in accuracy. Histological assessment, though considered the gold standard, requires invasive procedures and can be time-consuming. Moreover, treatment decisions frequently rest on complex and multidimensional clinical data, where prognostic uncertainty complicates personalized care. The rise of artificial intelligence (AI) has brought new opportunities to address these challenges. Machine learning and deep learning algorithms have already shown impressive performance in cancer detection, classification, and outcome prediction. In dermatology, AI-based image analysis tools have achieved dermatologistlevel accuracy in recognizing malignant lesions from dermoscopic and photographic images. Similarly, AI systems analyzing radiological scans, pathology slides, and genomic data are increasingly aiding the early detection and treatment of soft tissue malignancies. However, one of the main criticisms of conventional AI models is their lack of interpretability. These systems often function as “black boxes,” generating predictions without providing clear explanations for their reasoning. This opacity raises concerns in clinical contexts where transparency, trust, accountability, and EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 81 7. EXPLAINABLE AI IN REDUCING THE BURDEN OF HEAD AND NECK CANCERS Background The burden of head and neck cancers (HNC) continues to rise globally, demanding innovative approaches to improve diagnosis, treatment, and prognosis. Among the various technological advances, Artificial Intelligence (AI) and its subfield of explainable AI (XAI) have received significant attention. These technologies present substantial potential in addressing the complexities of HNC, where clinical decisions often depend on subjective interpretation of medical images and patient data. AI-driven solutions, particularly when combined with radiomics, enable the extraction of meaningful insights from medical images and patient records, improving diagnostic precision and facilitating personalized treatment strategies. This chapter examines the role of explainable AI in reducing the burden of HNC, especially through improving the interpretability and clinical utility of predictive models used in HNC management. Early Detection of Oral Cancer through Explainable AI 94 Oral cancer constitutes a significant portion of head and neck cancers, with high mortality rates largely due to delayed diagnosis and the complexity of clinical examination. Early detection is essential for improving survival rates, making AI-driven diagnostic tools highly valuable for early intervention. One such tool, the Lightweight Explainable Network (LWENet), combines convolutional neural networks (CNN) with label-guided attention (LGA) for accurate and interpretable oral cancer detection. By utilizing depth-wise separable convolutions, LWENet reduces computational overhead while ensuring efficiency for clinical use. The addition of axial multi-head self-attention (AMSA) based Vision Transformer (ViT) encoders further strengthens the model’s ability to focus on significant features such as tissue texture and boundaries, which are essential for cancer detection. The application of Grad-CAM for visualizing the model's decisionmaking process enhances interpretability, allowing clinicians to understand how predictions are generated. In a study performed on the MOD and OCI datasets, LWENet achieved outstanding performance, with precision scores of 96.97 percent and 99.48 percent for oral cancer detection. This combination of high accuracy and interpretability makes LWENet a promising tool for early oral cancer detection, providing clinicians with a valuable second opinion to support timely treatment decisions. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 95 AI-Based Radiomics in HNC Radiomics, an emerging field applying quantitative analysis to medical imaging, has been integrated with AI to advance the management of head and neck cancers. Radiomics can extract a wide array of features from medical images, such as CT scans or MRIs, which characterize tumor phenotypes more objectively than traditional approaches. When paired with machine learning algorithms, these radiomic features help construct models capable of predicting cancer outcomes, including diagnosis, prognosis, and treatment response. Despite their potential, AI-based radiomics models face challenges related to model generalizability, data imbalance, and interpretability. However, the integration of explainable AI techniques, such as Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP), introduces transparency and allows clinicians to understand the factors influencing predictions. This improves trust in AI systems for decisionmaking in HNC management. Machine Learning Explainability for Survival Outcomes in HNSCC In head and neck squamous cell carcinoma (HNSCC), predicting survival outcomes is essential for guiding treatment strategies and patient management. Machine learning (ML) models FATEMEH HEDAYATI 96 have demonstrated promise in predicting overall survival (OS) by integrating clinicopathological, treatment-related, and sociodemographic data. However, the use of ML models in clinical practice is restricted by their inherent lack of interpretability. Recent research has applied XAI techniques, including SHAP and LIME, to increase the transparency of ML models and to provide clinicians with critical insights into the most influential features for survival predictions. By clarifying the contributions of variables such as cancer stage, HPV status, and p16 protein levels, these models enable healthcare providers to make more informed and individualized treatment decisions for HNSCC patients. The addition of explainable AI tools ensures that predictions are not only accurate but also clinically actionable. Radiomics and Deep Learning for Esophageal Cancer Grading Although esophageal cancer is not directly classified as HNC, its treatment challenges share many similarities, particularly in the use of AI for tumor grading and prognosis prediction. A new framework for esophageal cancer grading employs both radiomics and deep learning techniques to classify cancer stages with greater accuracy. By combining CT imaging features with machine learning models such as XGBoost and Random Forest, the study developed a robust system for improving diagnostic accuracy and EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 97 interpretability. For HNC, such frameworks could be adapted to enhance tumor staging, predict recurrence, and detect potential metastasis. Furthermore, the framework highlights the importance of model interpretability, ensuring that healthcare providers can trust the reasoning process when making treatment decisions. Interpretable Models for Laryngeal Cancer Diagnosis Laryngeal cancer, another major type of HNC, presents distinct diagnostic challenges, especially when depending on histopathological images. A recent study introduced a deep learning model using transformers for the grading of laryngeal squamous cell carcinoma (LSCC), which incorporates learned-parameter-free attention (LA) to minimize background noise in image data. This technique strengthens the model’s capacity to focus on critical areas contributing to diagnosis. By including explainable AI techniques such as attention mechanisms, the model allows clinicians to identify which image features are most influential in the decision-making process. The ability to interpret how the model derives its conclusions is essential for clinical adoption, as it promotes trust and supports improved decisionmaking. Predicting Survival in Laryngeal Squamous Cell Carcinoma FATEMEH HEDAYATI 98 For laryngeal squamous cell carcinoma (LSCC), a practical online prediction platform has been developed to estimate patient survival over five years. This model, created with machine learning algorithms such as SVM and XGBoost, uses clinical and demographic data to predict patient outcomes. The incorporation of SHAP for model interpretation improves transparency, showing which clinical factors, including cancer stage and patient demographics, most strongly influence survival predictions. This is particularly significant in HNC, where personalized treatment planning is essential to improving patient outcomes. By providing clear and interpretable predictions, such platforms can guide clinical decisions and ensure interventions are tailored to individual patient needs. Radiomics for Nasopharyngeal Carcinoma Prognostication Nasopharyngeal carcinoma (NPC) is another subtype of head and neck cancer that creates diagnostic challenges, mainly due to latestage detection and the absence of distinct early symptoms. Radiomics, combined with AI, addresses this by extracting detailed tumor characteristics from CT scans that can predict locoregional recurrence (LRR) and overall survival. By applying a model that integrates radiomic features with clinical data, predictions for LRR and survival outcomes have been substantially EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 99 improved. This strategy not only strengthens prognostication but also supports personalized treatment approaches. The inclusion of AI enhances the accuracy and interpretability of these models, ensuring clinicians can make informed decisions by combining data-driven insights with clinical expertise. Predicting Lymph Node Metastasis in Thyroid Cancer Although thyroid cancer is not formally categorized within HNC, predicting metastasis in thyroid cancer is closely related to predicting lymph node involvement in head and neck cancers. A recent study applied delta radiomics derived from enhanced CT scans to predict peripheral lymph node metastasis (LNM) in thyroid cancer patients. By merging clinical and radiomics data, machine learning algorithms achieved high predictive accuracy. SHAP values were employed to improve model interpretability, making clear the contribution of each feature to the prediction. This interpretability is vital for clinical decision-making, as it provides clinicians with a transparent understanding of how tumor characteristics influence metastasis predictions. Such approaches can readily be adapted to predict lymph node involvement in HNC, thereby enhancing patient management. Addressing Radiotherapy FATEMEH HEDAYATI 100 Toxicity Using AI AI has also been implemented to predict the adverse effects of radiotherapy in HNC patients, particularly with respect to long-term side effects such as xerostomia and parotid gland shrinkage. A fuzzy logic-based machine learning model was developed to forecast these outcomes using radiomics data from CT scans. By combining diverse types of data including clinical, dosimetric, and radiomic information, AI models can predict the likelihood of these toxicities, enabling early interventions. The interpretability of the model is improved through fuzzy logic, which provides transparent rule-based classifiers. This transparency supports clinicians in understanding the rationale behind treatment adjustments, ensuring that decisions are grounded in comprehensive analysis of the available data. Conclusion The integration of explainable AI into the management of head and neck cancers offers great potential to reduce the burden of these diseases by improving diagnostic accuracy, treatment planning, and prognostication. By introducing transparency and interpretability, AI models empower clinicians to make better-informed decisions that are tailored to the specific needs of each patient. The future of HNC care depends on the ongoing development of these AI-based EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 101 tools, ensuring they remain not only accurate but also understandable and practical in clinical applications. FATEMEH HEDAYATI 102 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 103 8. EXPLAINABLE AI IN REDUCING THE BURDEN OF ORAL CANCERS Background Founded by John McCarthy in 1956, it is now evident that Artificial Intelligence (AI) has transformed our lives by simulating human thinking through its capacity to process data and recognize patterns. In addition to its role in other health industries such as medical and pharmacological fields, AI is now being widely applied in different areas of dentistry, with applications ranging from caries detection and prediction to diagnosing various diseases such as oral cancers. More than 377,000 people globally each year are affected by oral cancer, with survival rates dropping significantly as the disease advances. Frequently developing as a progression of oral potentially malignant disorders (OPMD), early detection plays a critical role in saving patients. However, it remains challenging due to the subtle and often asymptomatic nature of this disease, which leads to delayed diagnoses and consequently poorer outcomes. AI is effectively being used in the oncology field and shows significant promise in detecting OPMD and oral cancer. Recent research has 104 employed different algorithms of AI such as deep learning (DL), machine learning (ML), and image recognition to detect and classify oral cancer, to predict how malignant it may become, to estimate its progression and nodal metastasis, and ultimately to detect its recurrence rate. JingWen Li and colleagues evaluated the diagnostic accuracy of AI-assisted clinical imaging in detecting OPMD and oral cancer, with a specific focus on comparing performance across different imaging modalities. The findings of seventeen studies involving various AI algorithms and imaging tools demonstrated that AI-assisted detection shows high diagnostic performance overall, with clinical photography yielding the highest diagnostic odds ratio (DOR = 77.772) and sensitivity (93.9 percent). The study concluded that AI, particularly when integrated with accessible tools such as clinical photography, holds substantial promise for enhancing the early diagnosis of OPMD and oral cancer, especially in resource-limited settings. Machine learning-based tools such as Straticyte™ utilize protein biomarkers to predict the risk of oral epithelial dysplasia progressing to oral squamous cell carcinoma, achieving high sensitivity and specificity. Advanced multiplex immunohistochemistry (mIHC) combined with ML allows precise spatial profiling of immune cells in the tumor microenvironment, thereby EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 105 aiding prognostication and therapy planning. In digital pathology, deep learning models analyze histopathological images to identify prognostic and predictive biomarkers directly from tissue morphology, bypassing conventional molecular assessments. These models are capable of predicting treatment responses and survival outcomes. Additionally, AI-assisted epigenomic profiling is emerging, with machine learning algorithms being used to detect and interpret DNA methylation and histone modifications relevant to oral squamous cell carcinoma pathogenesis. Collectively, these AI-driven approaches support more accurate, objective, and personalized oral cancer management. Despite concerns about these AI models functioning as so-called black boxes, ongoing advancements are aiming to enhance their transparency in order to ensure safety, accuracy, and clinical relevance prior to widespread clinical adoption. The burden of oral cancers and the need for explainability Oral cancers impose a dual burden on patients and health systems. Patients often experience physical disfigurement, impaired speech, difficulty in swallowing, and social stigma. Health systems are strained by the high cost of treatments, rehabilitation, and the long-term care required FATEMEH HEDAYATI 106 for survivors. The late-stage presentation of most patients is a major driver of poor survival outcomes. Early-stage oral cancers are associated with significantly better survival rates and less invasive treatment, underscoring the importance of timely diagnosis and intervention. Traditional diagnostic approaches rely heavily on the expertise of clinicians, pathologists, and radiologists. However, variability in expertise, resource limitations, and the subjective nature of clinical evaluations contribute to diagnostic delays and inaccuracies. AI offers solutions to these challenges by providing consistent and data-driven insights. Yet, without explainability, clinicians may be hesitant to rely on AI systems, particularly in life-altering decisions such as cancer diagnosis and treatment planning. Explainable AI addresses this gap by bridging the divide between the computational accuracy of AI models and the interpretability required in healthcare. For oral cancers, explainability is crucial in contexts such as identifying premalignant lesions, predicting treatment response, and distinguishing between tumor subtypes. Clinicians must understand why an AI model predicts malignancy in a lesion or why it recommends a particular treatment pathway, to ensure that the outputs are clinically valid and aligned with patient needs. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 107 XAI in early detection and screening of oral cancers Early detection remains one of the most effective strategies for reducing the burden of oral cancers. Screening programs, especially in high-risk populations, can identify precancerous or early malignant lesions before they progress to advanced disease. Visual oral examination, toluidine blue staining, brush cytology, and imaging-based methods are commonly used, but they are limited by subjectivity and resource availability. AI models have demonstrated impressive performance in detecting early lesions from clinical photographs, histopathological slides, and imaging modalities. XAI enhances these models by offering interpretability. For instance, in imagebased screening, XAI can highlight the specific regions of an oral lesion that contributed to the model’s prediction of malignancy. This visual explanation not only increases clinician trust but also provides an educational tool for less experienced healthcare workers. In resource-limited settings where specialist expertise is scarce, XAI-enabled AI tools can empower primary care providers to conduct effective oral cancer screenings. A community health worker equipped with a smartphone-based AI system could capture images of suspicious FATEMEH HEDAYATI 108 lesions and receive an explainable assessment that indicates the likelihood of malignancy and highlights the concerning features. This approach reduces reliance on specialist availability while maintaining transparency in decision-making. XAI in pathology and molecular profiling Histopathological evaluation remains the gold standard for diagnosing oral cancers. Pathologists examine tissue biopsies to assess cell morphology, tissue architecture, and other features that distinguish malignant from benign lesions. However, interpretation can vary between pathologists, and subtle features may be overlooked. AI models have been trained to analyze histopathological images with high accuracy, identifying features that correlate with malignancy or prognosis. Explainable AI strengthens this process by showing pathologists which features within the slide influenced the model’s prediction. For example, an XAI system might highlight irregular nuclear morphology, disrupted basement membranes, or abnormal mitotic figures as the basis for classifying a tissue as malignant. This explanation helps pathologists validate the AI’s decision and integrate it into their diagnostic workflow. Beyond morphology, molecular profiling plays EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 109 a critical role in understanding oral cancers. Genomic alterations, such as mutations in TP53 or amplification of EGFR, can influence prognosis and therapeutic options. AI models can analyze large genomic datasets to identify patterns associated with tumor behavior and treatment response. With explainability, these models can indicate which genetic alterations or pathways were most influential in predicting outcomes, providing valuable insights for precision oncology. XAI in imaging and treatment planning Imaging modalities, including MRI, CT, and PET scans, are essential for staging oral cancers, assessing tumor extent, and planning treatment. AI systems trained on imaging data can automate tumor segmentation, predict invasion of adjacent structures, and evaluate treatment response. Yet clinicians often hesitate to trust these automated outputs without clear explanations. XAI techniques such as heatmaps and saliency maps can show clinicians which regions of an image influenced the AI’s prediction. For instance, if an AI model predicts perineural invasion or lymph node involvement, the XAI explanation can highlight the anatomical regions that led to this conclusion. This transparency allows radiologists and oncologists to cross-validate the AI’s insights FATEMEH HEDAYATI 110 and make more confident treatment decisions. Treatment planning for oral cancers is highly individualized, often involving surgery, radiation, chemotherapy, or a combination of these. XAIenabled predictive models can support treatment decisions by explaining why a particular approach is recommended for a patient. For example, an XAI system might predict that a patient is likely to respond well to chemoradiotherapy based on tumor size, molecular markers, and imaging features, while also highlighting the specific variables that influenced the prediction. This level of transparency not only builds clinician trust but also facilitates shared decision-making with patients. XAI in prognostication and follow-up care Prognostication is a central aspect of managing oral cancers. Clinicians must estimate the likely course of the disease, including risks of recurrence and survival probabilities, to guide treatment intensity and follow-up schedules. AI models can integrate diverse datasets to generate highly personalized prognostic predictions. However, without explainability, these predictions risk being dismissed as opaque and untrustworthy. With XAI, clinicians gain insight into the variables driving prognostic predictions. For instance, an AI system may predict a high risk of recurrence in a EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 111 patient and explain that this is due to perineural invasion, lymph node involvement, and specific molecular markers. By making these factors explicit, XAI supports evidence-based decisionmaking and allows clinicians to align prognostic predictions with established clinical knowledge. Follow-up care is another domain where XAI can be transformative. Monitoring patients for recurrence or secondary malignancies often involves imaging, clinical examination, and biomarker testing. XAI-enabled AI tools can detect subtle early signs of recurrence and explain the features responsible, allowing for earlier intervention and improved outcomes. XAI and patient engagement Beyond clinicians, patients themselves benefit from explainability in AI-driven care. Cancer diagnosis and treatment can be overwhelming for patients, who may struggle to understand the rationale behind complex medical decisions. XAI can support patient engagement by presenting transparent explanations of AI-driven recommendations in understandable terms. For example, when an AI system recommends surgery over radiotherapy, XAI can break down the reasoning, highlighting tumor characteristics, expected treatment outcomes, and side effect profiles. By demystifying the decision-making process, patients can make more informed choices FATEMEH HEDAYATI 112 and feel more empowered in their care journey. Increased patient understanding and trust can also improve adherence to treatment plans, thereby enhancing overall outcomes. XAI in oral cancer research Research on oral cancers involves analyzing vast and complex datasets, including genomic data, imaging records, histopathological slides, and clinical outcomes. AI has the potential to accelerate discoveries by identifying novel patterns and associations. However, the lack of transparency in AI outputs can limit their utility in research. XAI provides the interpretability needed for scientific discovery. For instance, when an AI model identifies a new molecular signature associated with poor prognosis, XAI can highlight the genes or pathways involved, providing researchers with a clear direction for further investigation. Similarly, in drug discovery, XAI can explain why a particular compound is predicted to be effective against oral cancer, helping researchers prioritize candidates for laboratory validation. By enhancing transparency and interpretability, XAI ensures that AI-driven research is not only accurate but also scientifically actionable. This can accelerate the development of new diagnostic tools, biomarkers, and therapeutic strategies, EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 113 For example, if an AI system predicts poor prognosis for a patient with Ewing sarcoma, it can specify that the prediction was driven by large tumor size, presence of metastases, and specific genetic alterations. This allows clinicians to validate the AI’s output and consider additional interventions for high-risk patients. Follow-up care is another critical area where XAI can make an impact. Sarcoma survivors require long-term monitoring for recurrence, metastasis, and late treatment effects. AI tools can analyze imaging and clinical data to detect subtle signs of recurrence earlier than traditional methods. Explainability ensures that when a suspicious finding is flagged, clinicians can see which features drove the alert, improving trust and facilitating timely intervention. Patient engagement through explainable AI Patients with bone and musculoskeletal cancers often face difficult decisions regarding surgery, chemotherapy, and long-term rehabilitation. These decisions involve weighing oncologic outcomes against functional and qualityof-life considerations. XAI can enhance patient engagement by making AI-driven recommendations understandable to patients and their families. For example, if an AI system recommends limbFATEMEH HEDAYATI 126 sparing surgery over amputation, XAI can present the rationale in terms of tumor location, surgical feasibility, and predicted outcomes. By providing clear and interpretable explanations, XAI fosters shared decision-making and helps patients feel more empowered in their care. This transparency can improve adherence to treatment plans and enhance overall satisfaction with the care process. XAI in research and drug discovery Research on bone and musculoskeletal cancers is hindered by the rarity of these diseases and the complexity of their biology. AI can accelerate discoveries by analyzing large datasets to identify novel biomarkers, therapeutic targets, and drug candidates. However, opaque AI models limit scientific interpretation and reproducibility. XAI addresses this issue by making AI-driven discoveries interpretable. For instance, when an AI model identifies a gene expression signature associated with chemotherapy resistance, XAI can specify which genes and pathways contributed to the prediction. This provides researchers with actionable hypotheses for laboratory validation. In drug discovery, XAI can explain why a candidate compound is predicted to inhibit a sarcomarelated pathway, helping researchers prioritize compounds for preclinical testing. By enhancing transparency and interpretability, XAI ensures that AI-driven research contributes meaningfully EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 127 to advancing sarcoma care. Challenges in implementing XAI for musculoskeletal cancers Despite its potential, several challenges hinder the adoption of XAI in bone and musculoskeletal cancers. One challenge is the complexity of generating explanations that are both accurate and understandable. Highly technical explanations may satisfy data scientists but fail to be useful for clinicians. Conversely, overly simplistic explanations may lack sufficient detail to support clinical decision-making. Another challenge is data availability. The rarity of bone and musculoskeletal cancers means that datasets are often small and fragmented across institutions. Training robust AI models requires large and diverse datasets, and without them, XAI tools may not generalize well to different patient populations. Efforts to build international consortia and share data are essential to overcome this barrier. Integrating XAI into clinical workflows is also a practical challenge. Clinicians already face significant time pressures, and XAI tools must provide explanations that are concise, relevant, and seamlessly integrated into existing systems. Ethical and regulatory considerations are particularly important in oncology. XAI must be developed and validated with strict attention to FATEMEH HEDAYATI 128 patient privacy, data security, and accountability. Moreover, the legal implications of AI-driven decisions remain an area of active debate. Future perspectives The future of XAI in reducing the burden of bone and musculoskeletal cancers lies in developing hybrid models that combine AI-driven insights with human expertise. Such models would leverage the strengths of both approaches while ensuring interpretability and trust. Personalized explanations tailored to different users will also play an important role. For example, pathologists may require detailed feature-level explanations, while patients may benefit from simplified, patient-friendly explanations. Another promising direction is the integration of multimodal data. Bone and musculoskeletal cancers generate diverse types of data, from imaging and pathology to genomics and clinical records. AI models capable of integrating these data streams, coupled with explainability, could provide unprecedented insights into tumor biology and treatment response. Collaborative efforts across institutions, disciplines, and countries will be essential to realize the potential of XAI. By pooling resources and expertise, the medical and scientific community can develop robust, validated, and clinically relevant XAI tools that address the EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 129 unique challenges of sarcomas. Conclusion Explainable AI offers a transformative approach to reducing the burden of bone and musculoskeletal cancers. By making AI-driven insights transparent and interpretable, XAI enhances early diagnosis, improves accuracy in histopathology and molecular profiling, optimizes treatment planning, supports prognostication, and engages patients in their care. In research, XAI accelerates discoveries by clarifying the mechanisms behind AI predictions, fostering scientific understanding and innovation. Although challenges remain in terms of data availability, workflow integration, and regulatory frameworks, the potential of XAI to improve outcomes for patients with these rare and devastating cancers is immense. By bridging the gap between computational power and human interpretability, XAI has the potential to transform the management of bone and musculoskeletal cancers into a more precise, equitable, and effective field of oncology. FATEMEH HEDAYATI 130 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 131 10. EXPLAINABLE AI IN REDUCING THE BURDEN OF RARE AND OTHER CANCERS Explainable AI in Reducing the Burden of Pediatric Cancers Background Pediatric cancers, although relatively rare compared to adult malignancies, represent a significant health challenge due to their aggressive nature, complex management, and profound impact on affected children and their families. Childhood cancers include a wide spectrum of hematologic malignancies, such as acute lymphoblastic leukemia and acute myeloid leukemia, as well as solid tumors like neuroblastoma, Wilms tumor, rhabdomyosarcoma, and medulloblastoma. Unlike adult cancers, pediatric malignancies often arise from developmental and genetic abnormalities rather than prolonged exposure to environmental risk factors, highlighting the distinct biological and molecular underpinnings that drive these diseases. Despite advances in pediatric oncology, which have led to remarkable improvements in survival rates for many cancer types, significant challenges 132 remain. Early diagnosis is difficult because symptoms are often nonspecific, including fatigue, pallor, pain, or swelling, and may mimic common childhood illnesses. Delayed diagnosis can lead to advanced-stage disease at presentation, limiting therapeutic options and worsening prognosis. Furthermore, the intensive treatments required, including chemotherapy, radiation, and surgery, carry substantial short-term and longterm toxicities that affect growth, cognitive development, fertility, and overall quality of life. These factors contribute to the enduring burden of pediatric cancers on children, families, and healthcare systems. Artificial intelligence has emerged as a transformative tool in pediatric oncology, offering potential solutions for early detection, risk stratification, personalized treatment, and longterm survivorship care. However, the adoption of AI in clinical practice has been limited by the black-box nature of many algorithms, which produce predictions without clear explanations. In the context of pediatric care, where treatment decisions have life-altering consequences, transparency, interpretability, and trust are essential. Explainable AI, or XAI, addresses these challenges by providing interpretable and actionable insights from complex AI models. By revealing the factors contributing to predictions, XAI enables clinicians, families, and researchers EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 133 to understand, validate, and apply AI-driven recommendations in a safe and effective manner. The application of XAI in pediatric cancers has the potential to reduce the burden of disease by improving early detection, optimizing treatment decisions, predicting outcomes, supporting survivorship care, and accelerating research into the unique biology of childhood malignancies. This discussion explores the multifaceted role of XAI in pediatric oncology, highlighting current applications, challenges, and future directions. The burden of pediatric cancers Although childhood cancers account for a small percentage of overall cancer cases, they carry disproportionately high physical, emotional, and economic consequences. Acute lymphoblastic leukemia is the most common pediatric malignancy, followed by brain tumors, neuroblastoma, and Wilms tumor. Survival rates vary by cancer type, stage at diagnosis, and access to specialized care. For example, survival for children with acute lymphoblastic leukemia exceeds 85 percent in high-income countries, whereas survival for high-risk neuroblastoma or diffuse intrinsic pontine glioma remains below 50 percent, despite intensive therapy. The rarity and heterogeneity of pediatric cancers pose unique challenges. Pediatric oncologists often encounter few cases of rare tumor subtypes, FATEMEH HEDAYATI 134 making standardized treatment protocols less effective. Additionally, treatment-related toxicity is a significant concern, as children are more vulnerable to long-term adverse effects, including cognitive impairment, cardiotoxicity, secondary malignancies, growth disturbances, and endocrine disorders. These long-term consequences underscore the importance of precise, individualized treatment strategies that maximize efficacy while minimizing harm. Early diagnosis is another critical area of need. Pediatric cancers often present with subtle or nonspecific symptoms that overlap with common childhood conditions. For instance, fatigue, pallor, or bone pain may be mistakenly attributed to infections or growth-related changes. Delayed diagnosis can result in advanced disease, reducing the likelihood of curative outcomes. Implementing tools that support early recognition of high-risk cases is therefore a priority for reducing the burden of pediatric cancers. Explainable AI for early detection and diagnosis Early and accurate detection of pediatric cancers is crucial for improving outcomes. Traditional diagnostic approaches rely on clinical evaluation, laboratory tests, imaging, and histopathology. However, these methods can be limited by the rarity of the conditions, inter-observer variability, EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 135 and the subtlety of early-stage disease. AI models have shown promise in detecting pediatric cancers from medical images, electronic health records, genomic profiles, and other multimodal data sources. For example, AI algorithms trained on imaging data, such as MRI or CT scans, can identify tumors, characterize tissue abnormalities, and detect features indicative of malignancy that may be missed by the human eye. In laboratory data analysis, AI can identify patterns in blood counts, biomarkers, and genetic mutations that signal the presence of leukemia or other hematologic malignancies. Explainable AI enhances these applications by revealing which features or patterns contributed to a given prediction. In imaging, XAI can generate heatmaps that highlight tumor regions or areas of abnormal tissue structure, enabling radiologists and clinicians to verify the AI’s findings. In laboratory-based predictions, XAI can indicate specific laboratory values, gene expression levels, or mutation profiles that influenced the risk assessment. By providing interpretable outputs, XAI allows clinicians to integrate AI recommendations into their diagnostic workflow, fostering trust and supporting timely referral to specialized care centers. XAI in histopathology and FATEMEH HEDAYATI 136 molecular profiling Histopathological evaluation remains the gold standard for confirming pediatric cancer diagnoses. Pathologists examine tissue morphology, cell differentiation, mitotic activity, and other features to distinguish malignant from benign lesions and to classify tumor subtypes. However, interpretation can be challenging, particularly in rare pediatric tumors, due to overlapping features and variability in expertise. AI models trained on digitized histopathology slides can assist by classifying tumor types, grading malignancy, and identifying prognostic features. Explainable AI strengthens this process by making predictions interpretable. For instance, an XAI model may classify a neuroblastoma biopsy and highlight areas of high mitotic activity, necrosis, or specific cellular patterns that influenced the decision. This not only helps pathologists validate the AI’s output but also provides a teaching tool for less experienced practitioners. Molecular profiling is increasingly central to pediatric oncology. Pediatric cancers often harbor unique genetic alterations, such as ALK mutations in neuroblastoma, NTRK fusions in infantile fibrosarcoma, or specific chromosomal translocations in leukemias. These alterations have diagnostic, prognostic, and therapeutic implications. AI can integrate EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 137 genomic, transcriptomic, and epigenomic data to predict disease behavior, treatment response, or likelihood of relapse. Explainable AI ensures that the model identifies which mutations, gene expression patterns, or pathways contributed to its prediction, allowing clinicians and researchers to interpret results in a biologically meaningful way. XAI in imaging and treatment planning Imaging is critical for staging pediatric cancers, assessing tumor response, and guiding treatment planning. MRI, CT, PET scans, and ultrasound provide detailed information about tumor size, location, and involvement of adjacent structures. AI models can automate tumor segmentation, detect metastases, and predict treatment response with high accuracy. However, clinicians need to understand the basis of AI predictions to rely on them for decision-making. Explainable AI addresses this by highlighting specific regions of images that informed predictions. For example, an XAI model may indicate areas of a brain tumor that are likely to be highly proliferative or infiltrative, guiding neurosurgeons in planning resection. In radiation oncology, XAI can explain dose planning predictions, identifying regions of potential toxicity and allowing for safer and more precise FATEMEH HEDAYATI 138 targeting. By providing interpretable outputs, XAI helps clinicians optimize treatment plans while minimizing harm to developing tissues and organs. Prognostication and followup care with XAI Prognostic predictions are essential for tailoring treatment intensity, scheduling follow-ups, and counseling families. AI models can integrate multiple data sources, including tumor histology, molecular profiles, imaging features, and clinical variables, to predict outcomes such as survival, relapse risk, or likelihood of therapy response. Explainable AI ensures that prognostic predictions are transparent and actionable. For example, if an AI model predicts high relapse risk for a child with leukemia, XAI can clarify that this is due to specific cytogenetic abnormalities, minimal residual disease levels, or early treatment response. Such interpretability allows clinicians to justify treatment intensification, plan closer monitoring, or consider experimental therapies. Follow-up care can also benefit from XAI by detecting subtle changes in imaging, laboratory results, or clinical symptoms that may indicate recurrence, providing early alerts and guiding timely intervention. Patient and family EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 139 engagement through XAI Pediatric cancer care involves not only clinicians but also families, who play a central role in decision-making. AI recommendations can be complex and difficult to interpret, particularly for non-specialists. Explainable AI can improve patient and family engagement by providing clear and understandable explanations of risk assessments, treatment options, and expected outcomes. For example, when an AI model recommends a particular chemotherapy regimen or surgical approach, XAI can break down the rationale in terms of tumor characteristics, predicted response, and potential side effects. This transparency fosters shared decision-making, improves adherence to treatment plans, and enhances confidence in care. Families who understand the reasoning behind clinical decisions are better equipped to support their child through treatment and recovery. XAI in pediatric cancer research and drug development Research in pediatric oncology is complicated by the rarity of many cancers, limiting sample sizes and slowing the discovery of novel therapies. AI can accelerate research by identifying patterns across diverse datasets, including genomics, imaging, and clinical outcomes. However, opaque FATEMEH HEDAYATI 140 AI models can produce results that are difficult to interpret, limiting their translational potential. Explainable AI enhances research by clarifying the mechanisms behind predictions. For instance, an XAI model might identify a gene expression signature associated with high-risk neuroblastoma and highlight the key genes driving the association. In drug development, XAI can explain why a candidate compound is predicted to inhibit tumor growth or overcome resistance, guiding prioritization for preclinical and clinical testing. By making AI-driven research interpretable, XAI accelerates the development of new diagnostics, targeted therapies, and precision treatment strategies in pediatric oncology. Challenges in implementing XAI in pediatric cancers Several challenges must be addressed to fully realize the potential of XAI in pediatric oncology. One key challenge is data availability. Pediatric cancers are rare, and high-quality, annotated datasets are limited. Training robust AI models requires multicenter collaborations, data sharing, and standardization of imaging, genomic, and clinical data. Another challenge is developing explanations that are both accurate and clinically meaningful. Highly technical explanations may be difficult for clinicians or families to interpret, while EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 141 oversimplified explanations may omit critical information. Integrating XAI into clinical workflows is also essential, ensuring that interpretability tools are user-friendly, actionable, and seamlessly incorporated into decisionmaking processes. Ethical and regulatory considerations are particularly important in pediatric oncology. AI systems must protect patient privacy, comply with legal frameworks, and ensure accountability. The use of AI in making treatment decisions for children requires careful oversight, rigorous validation, and adherence to evidence-based practice standards. Future directions The future of XAI in pediatric oncology includes the development of hybrid models that combine AI-driven insights with human expertise. These models can leverage computational power while ensuring interpretability and trust. Personalized explanations tailored to different stakeholders, such as oncologists, radiologists, pathologists, or families, will further enhance adoption and utility. Integration of multimodal data is another promising direction. Pediatric cancers generate diverse data types, including imaging, histopathology, genomics, and clinical information. AI models capable of synthesizing FATEMEH HEDAYATI 142 these data streams, coupled with explainable outputs, can provide holistic insights into tumor biology, treatment response, and long-term outcomes. Collaborative efforts across institutions, disciplines, and countries will be essential for building robust, validated XAI systems. By pooling resources and expertise, researchers and clinicians can accelerate discovery, optimize care, and improve outcomes for children with cancer. Conclusion Explainable AI has the potential to transform pediatric oncology by enhancing early detection, supporting accurate diagnosis, optimizing treatment planning, improving prognostication, and engaging families in shared decision-making. By providing transparent and interpretable insights, XAI addresses the key barrier of trust that has limited AI adoption in pediatric care. Although challenges remain, including data scarcity, workflow integration, and ethical considerations, the opportunities presented by XAI are profound. By bridging the gap between computational complexity and human interpretability, XAI can reduce the burden of pediatric cancers, improve survival and quality of life, and accelerate research into the unique biology of childhood malignancies. Through collaboration, innovation, and careful EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 143 implementation, XAI can become an integral component of precision pediatric oncology, offering hope for children and families affected by cancer. FATEMEH HEDAYATI 144 EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 145 W. Quantitative analysis of studies that use artificial intelligence on thyroid cancer: a 20-year bibliometric analysis. Front Oncol. 2025;15:1525650. 72. Ghane G, Karimi R, Chekeni AM, Darvishi M, Imani R, Vafaeinezhad FZ. Pain Management in Cancer Patients With Artificial Intelligence: Narrative Review. Scientifica (Cairo). 2025;2025:6888213. 73. Goh S, Goh RSJ, Chong B, Ng QX, Koh GCH, Ngiam KY, et al. Challenges in Implementing Artificial Intelligence in Breast Cancer Screening Programs: Systematic Review and Framework for Safe Adoption. J Med Internet Res. 2025;27:e62941. 74. Gonçalves N, Chaves J, Marques-Sá I, DinisRibeiro M, Libânio D. Early diagnosis of gastric cancer: Endoscopy and artificial intelligence. Best Pract Res Clin Gastroenterol. 2025;75:101979. 75. Grunhut J, Newland JJ, Brown RF. Implications of Artificial Intelligence for Colorectal Cancer in Young Populations. J Surg Oncol. 2025;131(7):1368-72. 76. HaghighiKian SM, Shirinzadeh-Dastgiri A, Vakili-Ojarood M, Naseri A, Barahman M, Saberi A, et al. A Holistic Approach to Implementing Artificial Intelligence in Lung Cancer. Indian J Surg Oncol. 2025;16(1):257-78. 77. Hamamoto R, Komatsu M, Yamada M, FATEMEH HEDAYATI 158 Kobayashi K, Takahashi M, Miyake M, et al. Current status and future direction of cancer research using artificial intelligence for clinical application. Cancer Sci. 2025;116(2):297-307. 78. Haque F, Simon BD, Özyörük KB, Harmon SA, Türkbey B. Generative Artificial Intelligence in Prostate Cancer Imaging. Balkan Med J. 2025;42(4):286-300. 79. Hardacre C, Hibbs T, Fok M, Wiles R, Bashar N, Ahmed S, et al. Predicting Surgical Difficulty in Rectal Cancer Surgery: A Systematic Review of Artificial Intelligence Models Applied to PreOperative MRI. Cancers (Basel). 2025;17 )5(. 80. Hashem H, Sultan I. Revolutionizing precision oncology: the role of artificial intelligence in personalized pediatric cancer care. Front Med (Lausanne). 2025;12:1555893. 81. Hashim HT, Alhatemi AQM, Daraghma M, Ali HT, Khan MA, Sulaiman FA, et al. Artificial intelligence versus radiologists in detecting earlystage breast cancer from mammograms: a metaanalysis of paradigm shifts. Pol J Radiol. 2025;90:e1-e8. 82. Haue AD, Hjaltelin JX, Holm PC, Placido D, Brunak SR. Artificial intelligence-aided data mining of medical records for cancer detection and screening. Lancet Oncol. 2024;25(12):e694e703. 83. Hays P. Artificial intelligence in EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 159 cytopathological applications for cancer: a review of accuracy and analytic validity. Eur J Med Res. 2024;29(1):553. 84. Hengky A, Lionardi SK, Kusumajaya C. Can artificial intelligence aid the urologists in detecting bladder cancer? Indian J Urol. 2024;40(4):221-8. 85. Hilbers D, Nekain N, Bates A, Nunez JJ. Patient Attitudes Toward Artificial Intelligence in Cancer Care: Scoping Review. JMIR Cancer. 2025;11:e74010. 86. Hofman P, Ourailidis I, Romanovsky E, Ilié M, Budczies J, Stenzinger A. Artificial intelligence for diagnosis and predictive biomarkers in NonSmall cell lung cancer Patients: New promises but also new hurdles for the pathologist. Lung Cancer. 2025;200:108110. 87. Howard HR, Hasanova M, Tiwari A, Ghose A, Winayak R, Nahar T, et al. The landscape of conventional and artificial intelligence-based clinical prediction models in non-small-cell lung cancer: from development to real-world validation. ESMO Open. 2025;10(9):105557. 88. Huang D, Li Z, Jiang T, Yang C, Li N. Artificial intelligence in lung cancer: current applications, future perspectives, and challenges. Front Oncol. 2024;14:1486310. 89. Huang L, Wu X, You J, Jin Z, He W, Sun J, et al. Artificial Intelligence Can Predict Personalized FATEMEH HEDAYATI 160 Immunotherapy Outcomes in Cancer. Cancer Immunol Res. 2025;13(7):964-77. 90. Huang R, Jin X, Liu Q, Bai X, Karako K, Tang W, et al. Artificial intelligence in colorectal cancer liver metastases: From classification to precision medicine. Biosci Trends. 2025;19(2):150-64. 91. Hussain MS, Ramalingam PS, Chellasamy G, Yun K, Bisht AS, Gupta G. Harnessing Artificial Intelligence for Precision Diagnosis and Treatment of Triple Negative Breast Cancer. Clin Breast Cancer. 2025;25(5):406-21. 92. Imani S, Li X, Chen K, Maghsoudloo M, Jabbarzadeh Kaboli P, Hashemi M, et al. Computational biology and artificial intelligence in mRNA vaccine design for cancer immunotherapy. Front Cell Infect Microbiol. 2024;14:1501010. 93. Ingman WV, Britt KL, Stone J, Nguyen TL, Hopper JL, Thompson EW. Artificial intelligence improves mammography-based breast cancer risk prediction. Trends Cancer. 2025;11(3):188-91. 94. Javanmard Z, Zarean Shahraki S, Safari K, Omidi A, Raoufi S, Rajabi M, et al. Artificial intelligence in breast cancer survival prediction: a comprehensive systematic review and metaanalysis. Front Oncol. 2024;14:1420328. 95. Jeong S, Choi HI, Yang KI, Kim JS, Ryu JW, Park HJ. Artificial Intelligence in the Diagnosis of Tongue Cancer: A Systematic Review with MetaEXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 161 Analysis. Biomedicines. 2025;13 )8(. 96. Ji J, Duan F, Liao Q, Wang H, Liu S, Liu Y, et al. Artificial Intelligence-Based Pathology to Assist Prediction of Neoadjuvant Therapy Responses for Breast Cancer. Cancer Med. 2025;14(15):e71132. 97. Jiang CQ, Li XJ, Zhou ZY, Xin Q, Yu L. Imaging based artificial intelligence for predicting lymph node metastasis in cervical cancer patients: a systematic review and meta-analysis. Front Oncol. 2025;15:1532698. 98. K RU, Elango AP, Subramanian R, Ks S. The Evolving Landscape of Ovarian Cancer: Innovations in Biotechnology and Artificial IntelligenceBased Screening and Treatment. Curr Treat Options Oncol. 2025;26(7):622-37. 99. Kadir A, Asaduzzaman M, Kundu J, Rahman MA, Rabbany MG, Shemanto MU, et al. Integrating Genetic Insights and Artificial Intelligence for Enhanced Oral and Maxillofacial Cancer Care. Methods Mol Biol. 2025;2952:107-24. 100. Kaidar-Person O, Pfob A, Valentini V, Aznar M, Dekker A, Meattini I, et al. Artificial intelligence in breast cancer radiotherapy: Insights from the Toolbox Consortium Delphi study. Breast. 2025;83:104537. 101. Karimzadhagh S, Ghodous S, Robati RM, Abbaspour E, Goldust M, Zaresharifi N, et al. Performance of Artificial Intelligence in Skin Cancer Detection: An Umbrella Review of FATEMEH HEDAYATI 162 Systematic Reviews and Meta-Analyses. Int J Dermatol. 2025. 102. Karuppan Perumal MK, Rajan Renuka R, Kumar Subbiah S, Manickam Natarajan P. Artificial intelligence-driven clinical decision support systems for early detection and precision therapy in oral cancer: a mini review. Front Oral Health. 2025;6:1592428. 103. Kemna R, Zeeuw JM, Ziesemer KA, Ali M, Bereska JI, Marquering H, et al. From Development to Implementation: A Systematic Review on the Current Maturity Status of Artificial Intelligence Models for Patients with Colorectal Cancer Liver Metastases. Oncology. 2025:1-10. 104. Khosravi P, Fuchs TJ, Ho DJ. Artificial Intelligence-Driven Cancer Diagnostics: Enhancing Radiology and Pathology through Reproducibility, Explainability, and Multimodality. Cancer Res. 2025;85(13):2356-67. 105. Kim SJ, Clark V, Hancock JT, Rawassizadeh R, Liu H, Taylor EA, et al. Leveraging artificial intelligence-mediated communication for cancer prevention and control and drug addiction: A systematic review. Transl Behav Med. 2025;15 )1(. 106. Kotoulas SC, Spyratos D, Porpodis K, Domvri K, Boutou A, Kaimakamis E, et al. A Thorough Review of the Clinical Applications of Artificial Intelligence in Lung Cancer. Cancers (Basel). 2025;17 )5(. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 163 107. Kuppanda PM, Janda M, Soyer HP, Caffery LJ. What Are Patients' Perceptions and Attitudes Regarding the Use of Artificial Intelligence in Skin Cancer Screening and Diagnosis? Narrative Review. J Invest Dermatol. 2025;145(8):1858-65. 108. Lei C, Sun W, Wang K, Weng R, Kan X, Li R. Artificial intelligence-assisted diagnosis of early gastric cancer: present practice and future prospects. Ann Med. 2025;57(1):2461679. 109. Leonard S, Patel MA, Zhou Z, Le H, Mondal P, Adams SJ. Comparing Artificial Intelligence and Traditional Regression Models in Lung Cancer Risk Prediction Using A Systematic Review and MetaAnalysis. J Am Coll Radiol. 2025;22(6):675-90. 110. Li G, Shi Q, Wu Q, Sui X. Target identification of natural products in cancer with chemical proteomics and artificial intelligence approaches. Cancer Biol Med. 2025;22(6):549-97. 111. Li K, Wu S, Zhang Y, Zhu B, Qi Z, Hou S, et al. The Usability and Experience of Artificial Intelligence-Based Conversational Agents in Health Education for Cancer Patients: A Scoping Review. J Clin Nurs. 2025. 112. Li R, Li J, Wang Y, Liu X, Xu W, Sun R, et al. The artificial intelligence revolution in gastric cancer management: clinical applications. Cancer Cell Int. 2025;25(1):111. 113. Li W, Hu R, Zhang Q, Yu Z, Deng L, Zhu X, et al. Artificial intelligence in prostate cancer. Chin FATEMEH HEDAYATI 164 Med J (Engl). 2025;138(15):1769-82. 114. Liao W, Xu X. Progress in the application research of cervical cancer screening developed by artificial intelligence in large populations. Discov Oncol. 2025;16(1):1282. 115. Lin X, Zhang Z, Zhou T, Li J, Jin Q, Li Y, et al. The Role of Computed Tomography and Artificial Intelligence in Evaluating the Comorbidities of Chronic Obstructive Pulmonary Disease: A OneStop CT Scanning for Lung Cancer Screening. Int J Chron Obstruct Pulmon Dis. 2025;20:1395-406. 116. Lin Y, Cheng M, Wu C, Huang Y, Zhu T, Li J, et al. MRI-based artificial intelligence models for post-neoadjuvant surgery personalization in breast cancer: a narrative review of evidence from Western Pacific. Lancet Reg Health West Pac. 2025;57:101254. 117. Liu JY, Sun RH, Li C. [Advances in the application of artificial intelligence technology in the surgical diagnosis and treatment of thyroid cancer]. Zhonghua Er Bi Yan Hou Tou Jing Wai Ke Za Zhi. 2025;60(1):80-6. 118. Liu L, Liu J, Su Q, Chu Y, Xia H, Xu R. Performance of artificial intelligence for diagnosing cervical intraepithelial neoplasia and cervical cancer: a systematic review and metaanalysis. EClinicalMedicine. 2025;80:102992. 119. Liu L, Pei Q, Qadir J, Chen Y, Li J, Luo Y, et al. Application of artificial intelligenceEXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 165 based stemness index in cancer. Front Oncol. 2025;15:1608712. 120. Lococo F, Ghaly G, Flamini S, Campanella A, Chiappetta M, Bria E, et al. Artificial intelligence applications in personalizing lung cancer management: state of the art and future perspectives. J Thorac Dis. 2024;16(10):7096-110. 121. Long X, Sun K, Lai S, Liu Y, Su J, Chen W, et al. Artificial intelligence and anti-cancer drugs' response. Acta Pharm Sin B. 2025;15(7):3355-71. 122. Lukac S, Putz F, De Micheli G, Corti C, Janni W, Tolaney SM, et al. Artificial intelligence as treatment support in breast cancer: current perspectives. Breast. 2025;83:104564. 123. Ma X, Zhang Q, He L, Liu X, Xiao Y, Hu J, et al. Artificial intelligence application in the diagnosis and treatment of bladder cancer: advance, challenges, and opportunities. Front Oncol. 2024;14:1487676. 124. Macheka S, Ng PY, Ginsburg O, Hope A, Sullivan R, Aggarwal A. Prospective evaluation of artificial intelligence (AI) applications for use in cancer pathways following diagnosis: a systematic review. BMJ Oncol. 2024;3(1):e000255. 125. Maggi M, Chierigo F, Fallara G, Jannello LMI, Tozzi M, Pellegrino F, et al. Shaping the Future of Personalized Therapy in Bladder Cancer Using Artificial Intelligence. Eur Urol Focus. 2025. FATEMEH HEDAYATI 166 126. Martella S, Cusumano G, Senevirathne TH, Stylianakis D, Palmas E, Denaro N, et al. Evolutionary Overview and Future Perspectives: ESR1 Mutations, Liquid Biopsy, and Artificial Intelligence for a New Era of Personalized Medicine in ER+ Breast Cancer. Mol Diagn Ther. 2025. 127. Mastella E, Calderoni F, Manco L, Ferioli M, Medoro S, Turra A, et al. A systematic review of the role of artificial intelligence in automating computed tomography-based adaptive radiotherapy for head and neck cancer. Phys Imaging Radiat Oncol. 2025;33:100731. 128. Mastroleo F, Marvaso G, Jereczek-Fossa BA. Artificial intelligence in muscle-invasive bladder cancer: opportunities, challenges, and clinical impact. Curr Opin Urol. 2025;35(5):543-8. 129. Megat Ramli PN, Aizuddin AN, Ahmad N, Abdul Hamid Z, Ismail KI. A Systematic Review: The Role of Artificial Intelligence in Lung Cancer Screening in Detecting Lung Nodules on Chest XRays. Diagnostics (Basel). 2025;15 )3(. 130. Mehri-Kakavand G, Mdletshe S, Wang A. A Comprehensive Review on the Application of Artificial Intelligence for Predicting Postsurgical Recurrence Risk in Early-Stage Non-Small Cell Lung Cancer Using Computed Tomography, Positron Emission Tomography, and Clinical Data. J Med Radiat Sci. 2025. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 167 131. Mela E, Tsapralis D, Papaconstantinou D, Sakarellos P, Vergadis C, Klontzas ME, et al. Current Role of Artificial Intelligence in the Management of Esophageal Cancer. J Clin Med. 2025;14 )6(. 132. Migliorelli A, Manuelli M, Ciorba A, Stomeo F, Pelucchi S, Bianchini C. Role of Artificial Intelligence in Human Papillomavirus Status Prediction for Oropharyngeal Cancer: A Scoping Review. Cancers (Basel). 2024;16 )23(. 133. Mirfendereski P, Li GY, Pearson AT, Kerr AR. Artificial intelligence and the diagnosis of oral cavity cancer and oral potentially malignant disorders from clinical photographs: a narrative review. Front Oral Health. 2025;6:1569567. 134. Moglia V, Johnson O, Cook G, de Kamps M, Smith L. Artificial intelligence methods applied to longitudinal data from electronic health records for prediction of cancer: a scoping review. BMC Med Res Methodol. 2025;25(1):24. 135. Montazer F, Mehdikhani B, Noroozi N, Nezameslami R, Nezameslami A, Shahbazi A, et al. GPU-Accelerated Artificial Intelligence Applications in Cancer Diagnosis, Imaging, and Treatment Planning. Asian Pac J Cancer Prev. 2025;26(8):2725-39. 136. Moradi Kashkooli F, Bhandari A, Gu B, Kolios MC, Kohandel M, Zhan W. Multiphysics modelling enhanced by imaging and artificial intelligence for personalised cancer FATEMEH HEDAYATI 168 nanomedicine: Foundations for clinical digital twins. J Control Release. 2025;386:114138. 137. Murray K, Oldfield L, Stefanova I, Gentiluomo M, Aretini P, O'Sullivan R, et al. Biomarkers, omics and artificial intelligence for early detection of pancreatic cancer. Semin Cancer Biol. 2025;111:76-88. 138. Mushcab H, Al Ramis M, AlRujaib A, Eskandarani R, Sunbul T, AlOtaibi A, et al. Application of Artificial Intelligence in CardioOncology Imaging for Cancer Therapy-Related Cardiovascular Toxicity: Systematic Review. JMIR Cancer. 2025;11:e63964. 139. Naemi A, Tashk A, Sorayaie Azar A, Samimi T, Tavassoli G, Bagherzadeh Mohasefi A, et al. Applications of Artificial Intelligence for Metastatic Gastrointestinal Cancer: A Systematic Literature Review. Cancers (Basel). 2025;17 )3(. 140. Ng XJK, Mohd Khairuddin AS, Liu HC, Loh TC, Tan JL, Khor SM, et al. Artificial intelligenceassisted point-of-care devices for lung cancer. Clin Chim Acta. 2025;570:120191. 141. Nguyen MH, Le MHN, Bui AT, Le NQK. Artificial intelligence in predicting EGFR mutations from whole slide images in lung Cancer: A systematic review and Meta-Analysis. Lung Cancer. 2025;204:108577. 142. Ni HM, Kouzy R, Sabbagh A, Rooney MK, Feng J, Castillo SP, et al. The state of the art EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 169 in artificial intelligence and digital pathology in prostate cancer. Nat Rev Urol. 2025. 143. Noei Teymoordash S, Zendehdel H, Norouzi AR, Kashian M. Diagnostic accuracy of artificial intelligence algorithms to predict remove all macroscopic disease and survival rate after complete surgical cytoreduction in patients with ovarian cancer: a systematic review and metaanalysis. BMC Surg. 2025;25(1):27. 144. Norouzkhani N, Mobaraki H, Varmazyar S, Zaboli H, Mohamadi Z, Nikeghbali G, et al. Artificial intelligence networks for assessing the prognosis of gastrointestinal cancer to immunotherapy based on genetic mutation features: a systematic review and meta-analysis. BMC Gastroenterol. 2025;25(1):310. 145. Novellino T, Masciocchi C, Tudor AM, Casà C, Chiloiro G, Romano A, et al. Artificial Intelligence and Rectal Cancer: Beyond Images. Cancers (Basel). 2025;17 )13(. 146. O'Connor O, McVeigh TP. Increasing use of artificial intelligence in genomic medicine for cancer carethe promise and potential pitfalls. BJC Rep. 2025;3(1):20. 147. Oldan JD, Anugu A, Islam MZ, Amindarolzarbi A, Werner RA, Pomper MG, et al. Theranostics 2.0: Target-driven, artificial intelligence-enabled cancer therapy across tumor types. Diagn Interv Imaging. 2025. FATEMEH HEDAYATI 170 148. Paiboonborirak C, Abu-Rustum NR, Wilailak S. Artificial intelligence in the diagnosis and management of gynecologic cancer. Int J Gynaecol Obstet. 2025;171 Suppl 1:199-209. 149. Pallumeera M, Giang JC, Singh R, Pracha NS, Makary MS. Evolving and Novel Applications of Artificial Intelligence in Cancer Imaging. Cancers (Basel). 2025;17 )9(. 150. Patil MR, Bihari A. Role of artificial intelligence in cancer detection using protein p53: A Review. Mol Biol Rep. 2024;52(1):46. 151. Picchio V, Pontecorvi V, Dhori X, Bordin A, Floris E, Cozzolino C, et al. The emerging role of artificial intelligence applied to exosome analysis: from cancer biology to other biomedical fields. Life Sci. 2025;375:123752. 152. Pourakbar N, Motamedi A, Pashapour M, Sharifi ME, Sharabiani SS, Fazlollahi A, et al. Effectiveness of Artificial Intelligence Models in Predicting Lung Cancer Recurrence: A Gene Biomarker-Driven Review. Cancers (Basel). 2025;17 )11(. 153. Purohit L, Kiamos A, Ali S, AlvarezPinzon AM, Raez L. Incidental Pulmonary Nodule (IPN) Programs Working Together with Lung Cancer Screening and Artificial Intelligence to Increase Lung Cancer Detection. Cancers (Basel). 2025;17 )7(. 154. Ramchandani R, Guo E, Biglou SG, Sabbah EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 171 SG, Mostowy M, Mahiny D, et al. Representation and Bias in Artificial Intelligence Models for Thyroid Cancer: A Systematic Review. Thyroid. 2025. 155. Ramos R, Moura CS, Costa M, Lamas NJ, Castro LPE, Correia R, et al. Heterogeneity of Lung Cancer: The Histopathological Diversity and Tumour Classification in the Artificial Intelligence Era. Pathobiology. 2025;92(4):239-50. 156. Riaz IB, Khan MA, Osterman TJ. Artificial intelligence across the cancer care continuum. Cancer. 2025;131(16):e70050. 157. Rizkala T, Menini M, Massimi D, Repici A. Role of Artificial Intelligence for Colon Polyp Detection and Diagnosis and Colon Cancer. Gastrointest Endosc Clin N Am. 2025;35(2):389-400. 158. Roche JJ, Seyedshahi F, Rakovic K, Thu AW, Le Quesne J, Blyth KG. Current and future applications of artificial intelligence in lung cancer and mesothelioma. Thorax. 2025. 159. Romeo M, Dallio M, Napolitano C, Basile C, Di Nardo F, Vaia P, et al. Clinical Applications of Artificial Intelligence (AI) in Human Cancer: Is It Time to Update the Diagnostic and Predictive Models in Managing Hepatocellular Carcinoma (HCC)? Diagnostics (Basel). 2025;15 )3(. 160. Safarian A, Mirshahvalad SA, Nasrollahi H, Jung T, Pirich C, Arabi H, et al. Impact of [(18)F]FDG FATEMEH HEDAYATI 172 PET/CT Radiomics and Artificial Intelligence in Clinical Decision Making in Lung Cancer: Its Current Role. Semin Nucl Med. 2025;55(2):156-66. 161. Sahoo K, Lingasamy P, Khatun M, Sudhakaran SL, Salumets A, Sundararajan V, et al. Artificial Intelligence in cancer epigenomics: a review on advances in pan-cancer detection and precision medicine. Epigenetics Chromatin. 2025;18(1):35. 162. Sahoo RK, Sahoo KC, Dash GC, Kumar G, Baliarsingh SK, Panda B, et al. Diagnostic performance of artificial intelligence in detecting oral potentially malignant disorders and oral cancer using medical diagnostic imaging: a systematic review and meta-analysis. Front Oral Health. 2024;5:1494867. 163. Samathoti P, Kumarachari RK, Bukke SPN, Rajasekhar ESK, Jaiswal AA, Eftekhari Z. The role of nanomedicine and artificial intelligence in cancer health care: individual applications and emerging integrations-a narrative review. Discov Oncol. 2025;16(1):697. 164. Sampieri C, Peretti G. Democratizing cancer detection: artificial intelligence-enhanced endoscopy could address global disparities in head and neck cancer outcomes. Eur Arch Otorhinolaryngol. 2025;282(5):2739-43. 165. Sarvepalli S, Vadarevu S. Role of artificial intelligence in cancer drug discovery and EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 173 development. Cancer Lett. 2025;627:217821. 166. Scalia IG, Pathangey G, Abdelnabi M, Ibrahim OH, Abdelfattah FE, Pietri MP, et al. Applications of Artificial Intelligence for the Prediction and Diagnosis of Cancer TherapyRelated Cardiac Dysfunction in Oncology Patients. Cancers (Basel). 2025;17 )4(. 167. Šebestová G, Klinger T, Švajdler M, Jr., Daum O, Jirásek T. Utilization of Artificial Intelligence Algorithms for the Diagnosis of Breast, Lung, and Prostate Cancer. Cesk Patol. 2025;61(2):70-90. 168. Seth L, Ladbury C, Amini A. Artificial Intelligence and Machine Learning Approaches in Designing Immunotherapy in Cancer. Cancer Treat Res. 2025;129:17-32. 169. Sguanci M, Palomares SM, Cangelosi G, Petrelli F, Sandri E, Ferrara G, et al. Artificial Intelligence in the Management of Malnutrition in Cancer Patients: A Systematic Review. Adv Nutr. 2025;16(7):100438. 170. Shahzad K, Abu-Zanona M, Elzaghmouri BM, AbdelRahman SM, Fadol Osman AA, Al-Khateeb A, et al. USING ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING APPROACHES TO ENHANCE CANCER THERAPY AND DRUG DISCOVERY: A NARRATIVE REVIEW. J Ayub Med Coll Abbottabad. 2024;36(1):183-9. 171. Shen K, Tan M, Liu Y, Xu X, Yang S. FATEMEH HEDAYATI 174 From data to Diagnosis: How artificial intelligence is revolutionizing preoperative assessment of thyroid nodules and cancer. Eur J Surg Oncol. 2025;51(9):110191. 172. Shi J, Chen J, He G, Peng Q. Artificial intelligence in high-dose-rate brachytherapy treatment planning for cervical cancer: a review. Front Oncol. 2025;15:1507592. 173. Shirzad M, Salahvarzi A, Razzaq S, JavidNaderi MJ, Rahdar A, Fathi-Karkan S, et al. Revolutionizing prostate cancer therapy: Artificial intelligence - Based nanocarriers for precision diagnosis and treatment. Crit Rev Oncol Hematol. 2025;208:104653. 174. Silveira JA, da Silva AR, de Lima MZT. Harnessing artificial intelligence for predicting breast cancer recurrence: a systematic review of clinical and imaging data. Discov Oncol. 2025;16(1):135. 175. Silvestre-Barbosa Y, Castro VT, Di Carvalho Melo L, Reis PED, Leite AF, Ferreira EB, et al. Worldwide research trends on artificial intelligence in head and neck cancer: a bibliometric analysis. Oral Surg Oral Med Oral Pathol Oral Radiol. 2025;140(1):64-78. 176. Smiley A, Reategui-Rivera CM, VillarrealZegarra D, Escobar-Agreda S, Finkelstein J. Exploring Artificial Intelligence Biases in Predictive Models for Cancer Diagnosis. Cancers EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 175 (Basel). 2025;17 )3(. 177. Song B, Liang R. Integrating artificial intelligence with smartphone-based imaging for cancer detection in vivo. Biosens Bioelectron. 2025;271:116982. 178. Stevenson E, Esengur OT, Zhang H, Simon BD, Harmon SA, Turkbey B. An overview of utilizing artificial intelligence in localized prostate cancer imaging. Expert Rev Med Devices. 2025;22(4):293-310. 179. Sun K, Wang Y, Qu R, Yang Q, Luo R, Jiang Z, et al. Comprehensive application of artificial intelligence in colorectal cancer: A review. iScience. 2025;28(7):112980. 180. Tabataba Vakili S, Haywood D, Kirk D, Abdou AM, Gopalakrishnan R, Sadeghi S, et al. Application of Artificial Intelligence in Symptom Monitoring in Adult Cancer Survivorship: A Systematic Review. JCO Clin Cancer Inform. 2024;8:e2400119. 181. Taha HA, Zeilani RS, Haddad RH, Abdalrahim MS. Artificial intelligence and machine learning techniques for predicting neuropathic pain in patients with cancer: A systematic review. Digit Health. 2025;11:20552076251358315. 182. Tang C, Xu Z, Duan H, Zhang S. Advancements in artificial intelligence for ultrasound diagnosis of ovarian cancer: FATEMEH HEDAYATI 176 a comprehensive review. Front Oncol. 2025;15:1581157. 183. Tang X, Zhou H, Liu Y, Gao S, Zhou Y. Diagnostic performance of the ultrasound -based artificial intelligence diagnostic system in predicting cervical lymph node metastasis in patients with thyroid cancer: A systematic review and meta-analysis. Sci Prog. 2025;108(2):368504251346906. 184. Tayyil Purayil AL, Joseph RM, Raj A, Kooriyattil A, Jabeen N, Beevi SF, et al. Role of Artificial Intelligence in MRI-Based Rectal Cancer Staging: A Systematic Review. Cureus. 2024;16(12):e76185. 185. Thalambedu N, Balla M, Sivasubramanian BP, Sadaram P, Malla KP, Vasipalli KP, et al. Integrating artificial intelligence with circulating tumor DNA for non-small cell lung cancer: opportunities, challenges, and future directions. Front Med (Lausanne). 2025;12:1612376. 186. Tiwari A, Ghose A, Hasanova M, Faria SS, Mohapatra S, Adeleke S, et al. The current landscape of artificial intelligence in computational histopathology for cancer diagnosis. Discov Oncol. 2025;16(1):438. 187. Tun HM, Rahman HA, Naing L, Malik OA. Artificial intelligence utilization in cancer screening program across ASEAN: a scoping review. BMC Cancer. 2025;25(1):703. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 177 R, Sequist LV, Fintelmann FJ. Artificial Intelligence and Machine Learning in Lung Cancer Screening. Thorac Surg Clin. 2023;33(4):401-9. 262. Huang B, Huang H, Zhang S, Zhang D, Shi Q, Liu J, et al. Artificial intelligence in pancreatic cancer. Theranostics. 2022;12(16):6931-54. 263. You Y, Lai X, Pan Y, Zheng H, Vera J, Liu S, et al. Artificial intelligence in cancer target identification and drug discovery. Signal Transduct Target Ther. 2022;7(1):156. 264. Xu Y, Su GH, Ma D, Xiao Y, Shao ZM, Jiang YZ. Technological advances in cancer immunity: from immunogenomics to single-cell analysis and artificial intelligence. Signal Transduct Target Ther. 2021;6(1):312. 265. Kang J, Chowdhry AK, Pugh SL, Park JH. Integrating Artificial Intelligence and Machine Learning Into Cancer Clinical Trials. Semin Radiat Oncol. 2023;33(4):386-94. 266. Zhang J, Wu J, Zhou XS, Shi F, Shen D. Recent advancements in artificial intelligence for breast cancer: Image augmentation, segmentation, diagnosis, and prognosis approaches. Semin Cancer Biol. 2023;96:11-25. 267. Huang S, Yang J, Shen N, Xu Q, Zhao Q. Artificial intelligence in lung cancer diagnosis and prognosis: Current application and future perspective. Semin Cancer Biol. 2023;89:30-7. FATEMEH HEDAYATI 190 268. He X, Liu X, Zuo F, Shi H, Jing J. Artificial intelligence-based multi-omics analysis fuels cancer precision medicine. Semin Cancer Biol. 2023;88:187-200. 269. Chen M, Copley SJ, Viola P, Lu H, Aboagye EO. Radiomics and artificial intelligence for precision medicine in lung cancer treatment. Semin Cancer Biol. 2023;93:97-113. 270. Yang F, Darsey JA, Ghosh A, Li HY, Yang MQ, Wang S. Artificial Intelligence and Cancer Drug Development. Recent Pat Anticancer Drug Discov. 2022;17(1):2-8. 271. Zhou Y, Xu X, Song L, Wang C, Guo J, Yi Z, et al. The application of artificial intelligence and radiomics in lung cancer. Precis Clin Med. 2020;3(3):214-27. 272. Zhang Y, Ma W, Huang Z, Liu K, Feng Z, Zhang L, et al. Research and application of omics and artificial intelligence in cancer. Phys Med Biol. 2024;69 )21(. 273. M SA, Al-Musawi SG, Al-Alwany AA, Uinarni H, Rasulova I, Rodrigues P, et al. Artificial intelligence in cancer diagnosis: Opportunities and challenges. Pathol Res Pract. 2024;253:154996. 274. Kenner B, Chari ST, Kelsen D, Klimstra DS, Pandol SJ, Rosenthal M, et al. Artificial Intelligence and Early Detection of Pancreatic Cancer: 2020 Summative Review. Pancreas. 2021;50(3):251-79. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 191 275. Wang Y, Lin W, Zhuang X, Wang X, He Y, Li L, et al. Advances in artificial intelligence for the diagnosis and treatment of ovarian cancer (Review). Oncol Rep. 2024;51 )3(. 276. Calderaro J, Žigutytė L, Truhn D, Jaffe A, Kather JN. Artificial intelligence in liver cancer - new tools for research and patient management. Nat Rev Gastroenterol Hepatol. 2024;21(8):585-99. 277. Bera K, Braman N, Gupta A, Velcheti V, Madabhushi A. Predicting cancer outcomes with radiomics and artificial intelligence in radiology. Nat Rev Clin Oncol. 2022;19(2):132-46. 278. Perez-Lopez R, Ghaffari Laleh N, Mahmood F, Kather JN. A guide to artificial intelligence for cancer researchers. Nat Rev Cancer. 2024;24(6):427-41. 279. Shmatko A, Ghaffari Laleh N, Gerstung M, Kather JN. Artificial intelligence in histopathology: enhancing cancer research and clinical oncology. Nat Cancer. 2022;3(9):1026-38. 280. Salama V, Godinich B, Geng Y, HumbertVidan L, Maule L, Wahid KA, et al. Artificial Intelligence and Machine Learning in Cancer Related Pain: A Systematic Review. medRxiv. 2023. 281. Abdel Razek AAK, Khaled R, Helmy E, Naglah A, AbdelKhalek A, El-Baz A. Artificial Intelligence and Deep Learning of Head and Neck Cancer. Magn Reson Imaging Clin N Am. FATEMEH HEDAYATI 192 2022;30(1):81-94. 282. Sebastian AM, Peter D. Artificial Intelligence in Cancer Research: Trends, Challenges and Future Directions. Life (Basel). 2022;12 )12(. 283. Chassagnon G, De Margerie-Mellon C, Vakalopoulou M, Marini R, Hoang-Thi TN, Revel MP, et al. Artificial intelligence in lung cancer: current applications and perspectives. Jpn J Radiol. 2023;41(3):235-44. 284. Cesario A, D'Oria M, Calvani R, Picca A, Pietragalla A, Lorusso D, et al. The Role of Artificial Intelligence in Managing Multimorbidity and Cancer. J Pers Med. 2021;11 )4(. 285. Bassani S, Santonicco N, Eccher A, Scarpa A, Vianini M, Brunelli M, et al. Artificial intelligence in head and neck cancer diagnosis. J Pathol Inform. 2022;13:100153. 286. Salama V, Godinich B, Geng Y, HumbertVidan L, Maule L, Wahid KA, et al. Artificial Intelligence and Machine Learning in Cancer Pain: A Systematic Review. J Pain Symptom Manage. 2024;68(6):e462-e90. 287. Jan Z, El Assadi F, Abd-Alrazaq A, Jithesh PV. Artificial Intelligence for the Prediction and Early Diagnosis of Pancreatic Cancer: Scoping Review. J Med Internet Res. 2023;25:e44248. 288. Brancaccio G, Balato A, Malvehy J, Puig S, EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 193 Argenziano G, Kittler H. Artificial Intelligence in Skin Cancer Diagnosis: A Reality Check. J Invest Dermatol. 2024;144(3):492-9. 289. Zhang C, Xu J, Tang R, Yang J, Wang W, Yu X, et al. Novel research and future prospects of artificial intelligence in cancer diagnosis and treatment. J Hematol Oncol. 2023;16(1):114. 290. Gao Q, Yang L, Lu M, Jin R, Ye H, Ma T. The artificial intelligence and machine learning in lung cancer immunotherapy. J Hematol Oncol. 2023;16(1):55. 291. Kinoshita T, Komatsu M. Artificial Intelligence in Surgery and Its Potential for Gastric Cancer. J Gastric Cancer. 2023;23(3):400-9. 292. Kim KW, Huh J, Urooj B, Lee J, Lee J, Lee IS, et al. Artificial Intelligence in Gastric Cancer Imaging With Emphasis on Diagnostic Imaging and Body Morphometry. J Gastric Cancer. 2023;23(3):388-99. 293. Ilhan B, Lin K, Guneri P, WilderSmith P. Improving Oral Cancer Outcomes with Imaging and Artificial Intelligence. J Dent Res. 2020;99(3):241-8. 294. Liu X, Shi J, Li Z, Huang Y, Zhang Z, Zhang C. The Present and Future of Artificial Intelligence in Urological Cancer. J Clin Med. 2023;12 )15(. 295. Yan S, Li J, Wu W. Artificial intelligence in breast cancer: application and future perspectives. FATEMEH HEDAYATI 194 J Cancer Res Clin Oncol. 2023;149(17):16179-90. 296. Katta MR, Kalluru PKR, Bavishi DA, Hameed M, Valisekka SS. Artificial intelligence in pancreatic cancer: diagnosis, limitations, and the future prospects-a narrative review. J Cancer Res Clin Oncol. 2023;149(9):6743-51. 297. Ahn JS, Shin S, Yang SA, Park EK, Kim KH, Cho SI, et al. Artificial Intelligence in Breast Cancer Diagnosis and Personalized Medicine. J Breast Cancer. 2023;26(5):405-35. EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR REDUCIN... 195 Proof