A flexible framework for sepsis prediction: Standardizing data management and imputation in time series using MIMIC-III

Abstract

Sepsis is a life-threatening immune response to infections, leading to organ dysfunction. Despite technological advances, the application of AI in sepsis prediction faces challenges, particularly due to the lack of standardized approaches for data preprocessing and imputation. This work introduces a new framework aimed at simplifying data management, ensuring AI models trained on time series data are both reliable and comprehensive. The framework facilitates the construction, preprocessing, and imputation of the Mimic-III database from PhysioNet, providing a standardized benchmark for future AI research in early sepsis prediction.