Telemedicine and diagnostic artificial intelligence: evolution, prospects, and social challenges

DOI: h p://doi.o g/10.12795/A gumen os/2024.i27.07 A gumen os de Razón Técnica, n.º 27, 2024, pp. 215-230 TELEMEDICINE AND DIAGNOSTIC ARTIFICIAL INTELLIGENCE: EVOLUTION, PROSPECTS, AND SOCIAL CHALLENGES TELEMEDICINA E INTELIGENCIA ARTIFICIAL DIAGNÓSTICA: EVOLUCIÓN, PERSPECTIVAS Y DESAFÍOS SOCIALES REGINA PENNER Sou h U al S a e Uni e si y, Russia penne . [email p o ec ed] h ps://o cid.o g/0000-0002-3277-7274 RECIBIDO: 30/09/2024 ACEPTADO: 15/11/2024 Abs ac : This a icle examines he e olu ion o elemedicine as one o he key ields in mode n medicine, u ilizing in o ma ion and communica ion echnologies o p o ide medical se ices emo ely. The s udy’s goal is o analyze he his o ical de elopmen o elemedicine and i s impac on medical p ac ice, as well as assess he p ospec s o using Diagnos ic A i icial In elligence (Diagnos ic AI) in elemedicine. The esea ch me hodology includes discou se analysis, based on he wo ks o Michel Foucaul and Gilles Deleuze, allowing he examina ion o medical ins i u ions as mechanisms o con olling and no malizing pa ien beha io . The s udy concludes ha elemedicine signi ican ly imp o es access o heal hca e, educes cos s, and enhances ea men e iciency. Diagnos ic AI ep esen s a new phase in elemedicine, capable o accele a ing and imp o ing diagnos ic p ocesses, hough i s implemen a ion is accompanied by signi ican e hical and social challenges. The no el y o his a icle lies in c i ically examining he ole o elemedicine and diagnos ic echnologies in con empo a y socie y, wi h a ocus on hei disciplina y unc ions and he in luence on he ela ionships be ween pa ien s and heal hca e p o essionals. The conclusion emphasizes he impo ance o u he s udy and egula ion o Diagnos ic AI in medical p ac ice o minimize isks and ensu e equi able access o heal hca e. Keywo ds: elemedicine; a i icial in elligence; diagnos ic a i icial in elligence; Google; AMIE; OpenAI; e olu ion; p ospec s; social challenges. REGINA PENNER A gumen os de Razón Técnica, n.º 27, 2024, pp. 215-230 216 Resumen: Es e a ículo examina la e olución de la elemedicina como uno de los campos cla e en la medicina mode na, u ilizando ecnologías de la in o mación y la comunicación pa a p opo ciona se icios médicos de o ma emo a. El obje i o del es udio es analiza el desa ollo his ó ico de la elemedicina y su impac o en la p ác ica médica, así como e alua las pe spec i as de uso de la In eligencia A i icial Diagnós ica (IA Diagnós ica) en la elemedicina. La me odología de in es igación incluye el análisis del discu so, basado en los abajos de Michel Foucaul y Gilles Deleuze, lo que pe mi e examina las ins i uciones médicas como mecanismos pa a con ola y no maliza el compo amien o del pacien e. El es udio concluye que la elemedicina mejo a signi ica i amen e el acceso a la a ención médica, educe los cos os y mejo a la e iciencia del a amien o. La IA diagnós ica ep esen a una nue a ase en la elemedicina, capaz de acele a y mejo a los p ocesos de diagnós ico, aunque su implemen ación es á acompañada de impo an es desa íos é icos y sociales. La no edad de es e a ículo adica en examina c í icamen e el papel de la elemedicina y las ecnologías de diagnós ico en la sociedad con empo ánea, con un en oque en sus unciones disciplina ias y la in luencia en las elaciones en e pacien es y p o esionales de la salud. La conclusión en a iza la impo ancia de segui es udiando y egulando la IA diagnós ica en la p ác ica médica pa a minimiza los iesgos y ga an iza un acceso equi a i o a la a ención médica. Palab as cla e: elemedicina; in eligencia a i icial; in eligencia a i icial diagnós ica; Google, AMIE; OpenAI; e olución; pe spec i as; e os sociales. In oduc ion. The Social Ins i u ion o Medicine The phenomenon o medicine and he discou se su ounding i a e no s a ic o comple e p ojec s; hey e ol e o e ime, ac oss a ious cul u es and socie ies. Fo ins ance, one majo e ela ion o Wes e n Eu opean socie y a he u n o he 19 h and 20 h cen u ies was he ecogni ion o emale sexuali y (e.g., in he wo ks o Sigmund F eud, such as “S udies on Hys e ia” o “Th ee Essays on he Theo y o Sexuali y”). Michel Foucaul , a guably he mos enowned “a chaeologis o knowledge,” p oposed he concep o he clinical gaze in his book “The Bi h o he Clinic” (Foucaul , 2003). Th ough his lens, we obse e a shi om iewing diseases TELEMEDICINE AND DIAGNOSTIC ARTIFICIAL INTELLIGENCE A gumen os de Razón Técnica, nº 27, 2024, pp. 215-230 217 as abs ac en i ies o s udying illness as some hing isible and diagnosable wi hin he human body. This ansi ion, on one hand, ans o ms he phenomenon o medicine i sel and i s social p esence, while on he o he hand, i se s he di ec ion o he discu si e p ac ices and social ela ions ha o m a ound i . Following Foucaul ’s insigh s, we can ace se e al s ages in he his o ical ans o ma ion o medical discou ses: 1) Classical Medicine (17 h – 18 h cen u ies): Cha ac e ized by a pu ely medical iew o disease, whe e diseases we e ea ed as indi idual en i ies, independen o he pa ien . Physicians, inspi ed by na u al sciences, aimed o classi y and sys ema ize diseases, much like bo anis s classi y plan s. Symp oms we e conside ed ex e nal mani es a ions o in e nal diseases, and he physician’s ask was o in e p e hese symp oms accu a ely. 2) Clinical Medicine (La e 18 h – Ea ly 19 h cen u ies): Du ing his pe iod, he clinic as a disciplina y ins i u ion was bo n. The ocus shi ed om abs ac diseases o he speci ic body o he pa ien , whe e diseases could be localized, connec ed o ce ain o gans and issues, and diagnosed, e en h ough dissec ion. This new pa adigm o bodily obse a ion con as ed wi h he pa ien ’s subjec i e expe ience. 3) Biopoli ics (19 h – 20 h cen u ies): Medicine became a ool o powe , aimed a con olling indi idual bodies and popula ions. A his s age, s a e heal hca e ins i u ions we e es ablished, se ing no only o ea pa ien s bu also as mechanisms o con olling public heal h (Foucaul , 2008). Fo example, psychia y o ms he unde s anding op ics h ough which he no ms o beha io and he deg ees o de ia ion om hem a e illumina ed; pa ien s o psychia ic clinics in such op ics a e ans o med in o objec s o disciplina y con ol, whe e he ask o medicine is no so much in ea men as in he no maliza ion o de ian beha io . In sum, medicine is in eg a ed in o he s a e machine, o ien ed owa d REGINA PENNER A gumen os de Razón Técnica, n.º 27, 2024, pp. 215-230 218 main aining and e en imp o ing he heal h o he na ion, bu mainly in s a is ical indica o s. La e , poin ing o he ansi ion om disciplina y sys ems o socie ies o con ol, G. Deleuze de eloped he Foucauldian concep , no ing he medicaliza ion o e e yday li e, when e e yday p ac ices (nu i ion, physical ac i i y, sleep) become objec s o medical a en ion, mo ing om he p i a e o he public sphe e, mo e accu a ely subjec o measu emen and con ol; and he echnologiza ion o medicine, as a esul o which medicine is ans o med in o a complex ne wo k, whe e heal h managemen is ca ied ou h ough in o ma ion pla o ms ha collec and analyze da a abou pa ien s in eal ime (Deleuze, 1992). This echnological u n is he implemen a ion o elemedicine, which is now a legi ima e a ea o medicine ha uses con empo a y in o ma ion and communica ion echnologies o p o ide medical se ices and pa ien consul a ions a a dis ance. Telemedicine has main ained i s ele ance o e he pas decade; he mos s iking illus a ion o his is s ill he COVID-19 pandemic, when access o adi ional heal h se ices became di icul a ound he wo ld. The so-called elemedicine “body” includes indi ec consul a ions, pa ien heal h moni o ing, eme gency cases and educa ional p og ams as in eg al elemen s. Fo example, consul a ions allow pa ien s o ecei e answe s o hei que ies om doc o s online, wi hou he need o di ec physical con ac . The impo ance o such a p ocedu e inc eases o people li ing in emo e o ha d- o- each a eas. Bu consul a ions also help educe he ac ual load on medical ins i u ions (sup a-indi idual u ili y) and educe he wai ing ime o an appoin men (indi idual u ili y). Dissemina ion o sma de ices ac oss indi idual households is ano he elemen o elemedicine. Using mobile de ices and senso s, pa ien s can emo ely moni o key heal h indica o s (blood p essu e, blood suga , hea a e). This da a is au oma ically ansmi ed o TELEMEDICINE AND DIAGNOSTIC ARTIFICIAL INTELLIGENCE A gumen os de Razón Técnica, nº 27, 2024, pp. 215-230 219 doc o s, which allows hem o p omp ly espond o de ia ions om he no m. In eme gency si ua ions, elemedicine echnologies allow he ic im o p omp ly con ac a doc o who can p o ide i s aid, gi e ecommenda ions o s abilizing he condi ion o decide on he need o hospi aliza ion. Mo eo e , oday elemedicine is inc easingly used o ain medical wo ke s. Online cou ses, webina s and o he educa ional esou ces help imp o e he quali ica ions o medical pe sonnel and in o m he popula ion. Telemedicine is a spec acula illus a ion o e ol ing disciplina y ins i u ions; i signi ican ly imp o es access o medical ca e o he gene al popula ion, educes heal h ca e cos s, and inc eases he e ec i eness o ea men . A he same ime, he need o ensu e he secu i y o pe sonal da a, p o ec pa ien p i acy, and s anda dize elemedicine se ices is g owing exponen ially. In addi ion, o e he pas i e yea s, non-human ac o s ha e inc easingly been included in he doc o -pa ien ela ionship. As ecen ly as he end o he 20 h cen u y, he unc ions o a machine in a medical con ex we e limi ed o analyzing he da a i ecei ed. Today, a machine claims o be a “specialis ” capable o p o iding p o essional se ices in consul ing a pa ien , diagnosing him, and suppo ing his ea men . Diagnos ic AI: A New Miles one in Telemedicine Opens New Challenges A i icial in elligence implemen ing diagnos ic goals in medical discou se (so-called Diagnos ic AI) is one o such agen s. Diagnos ic AI is one o he mos p omising a eas in medicine and in o ma ion echnology. I s applica ion co e s a ange o asks: om in e p e ing images and analyzing gene ic da a o p edic ing ea men esul s and suppo ing medical decisions. REGINA PENNER A gumen os de Razón Técnica, n.º 27, 2024, pp. 215-230 220 A i icial in elligence in medicine began wi h he i s expe sys ems in he 1960s. Howe e , a echnological leap has only been made in he las decade in he con ex o machine lea ning, in he a eas o deep lea ning and big da a p ocessing. Mode n AI-based diagnos ic sys ems a e able o analyze huge amoun s o da a, which signi ican ly exceeds human capabili ies, and p o ide doc o s wi h ecommenda ions based on his da a. One o he key b eak h oughs has been he de elopmen o AI sys ems o analyzing medical images. Fo example, a s udy by analys s om India showed ha such sys ems can iden i y pa hological changes in images wi h an accu acy compa able o o e en supe io o he capabili ies o expe ienced adiologis s ( he a icle speci ies ha sys ems ained on MRI o CT da a a e able o ecognize cance in he ea ly s ages, which signi ican ly inc eases he chances o success ul ea men ) (Ma hu e al., 2022). The use o AI in diagnos ics has a numbe o ob ious ad an ages. Fi s ly, i is he high speed and accu acy o da a p ocessing, which is especially impo an in he con ex o a sho age o quali ied medical pe sonnel and an inc easing bu den on heal hca e (Pin o-Coelho, 2023). Secondly, AI can help minimize he human ac o , which o en leads o diagnos ic e o s (Russell & No ig, 2021). Howe e , he implemen a ion o Diagnos ic AI is associa ed wi h a numbe o challenges and limi a ions. One o he main p oblems is he us o doc o s and pa ien s in AI sys ems. Ex ensi e clinical ials and ce i ica ion o such sys ems a e needed o ensu e hei eliabili y and sa e y. This o malizes he e hical p oblems in he use o AI in eal medical p ac ices, ounded, on he one hand, by he con iden iali y o pa ien da a, and on he o he hand, by he ques ion o he esponsibili y o AI o he decisions made (in gene al, abou he legi imacy o applying he c i e ia o esponsibili y o so wa e) (Flo idi, 2023). TELEMEDICINE AND DIAGNOSTIC ARTIFICIAL INTELLIGENCE A gumen os de Razón Técnica, nº 27, 2024, pp. 215-230 221 In any case, he e hical issues su ounding he use o diagnos ic AI ouch on se e al a eas a once. One o he key issues is ensu ing ai access o new echnologies. The e is a isk ha such sys ems will be used p edominan ly in de eloped coun ies, while de eloping coun ies may be le wi hou access o hese achie emen s ( he e y o mula ion o he p oblem epea s he classic plo o class s uggle, ansc ibed by McKenzie Wa k on he “hacke ” ma e ial (Wa k, 2004). An equally undamen al ques ion is who is esponsible o he decisions made by AI. I he sys em makes a mis ake, who will be held accoun able: he de elope s, he doc o s using he sys em, o he in elligen sys em i sel (Jobin, e al., 2019)? AMIE: P elimina y esul s o es ing he Google’s neu al ne wo k Diagnos ic AI is a e olu iona y ool ha has he po en ial o signi ican ly imp o e he quali y o medical ca e and make diagnos ics mo e accessible and accu a e. Howe e , success ully implemen ing AI in medical p ac ice equi es add essing a numbe o echnical, e hical, and social issues. A icula e Medical In elligence Explo e (AMIE), Google’s medical neu al ne wo k, demons a es his pa adox o echnological de elopmen and e hical u bulence. The de elopmen and es ing o AMIE, an inno a i e sys em based on la ge language models (LLM) designed o conduc medical diagnos ic dialogues, is discussed in he pape “Towa ds Con e sa ional Diagnos ic AI” by esea che s om Google Resea ch and Google DeepMind (Tu, e al., 2024). The au ho s de ine he goal o he s udy as he c ea ion o a i icial in elligence capable o imp o ing he a ailabili y and quali y o medical ca e by in oducing ad anced echnologies in o he diagnos ic p ocess and REGINA PENNER A gumen os de Razón Técnica, n.º 27, 2024, pp. 215-230 222 gene a ing AI ecommenda ions o he ea men o a speci ic pa ien (Tu, e al., 2024, p. 3). They cla i y ha dialogue be ween a doc o and a pa ien is he basis o medical p ac ice. Mo eo e , up o 80% o diagnoses a e made on he basis o anamnesis ob ained du ing communica ion wi h he pa ien (see, o example, Pe e son, 1992). Howe e , acco ding o expe s, he con empo a y heal hca e sys em a ound he wo ld aces many challenges, including a sho age o quali ied doc o s, especially in emo e and disad an aged a eas, as well as he high cos o medical se ices (see, o example, Roshan & Rao, 2000). In esponse o he iden i ied p oblems, he AMIE neu al ne wo k was de eloped, which uses a mode n a chi ec u e on LLM base. A he inpu AMIE ecei es ex in o ma ion in he o m o a ques ion o complain o he pa ien ; based on he ini ial in o ma ion and answe s o cla i ying ques ions a he ou pu he neu al ne wo k issues om wo diagnoses co esponding o he indica ed symp oms, cla i ies he mos p obable o hem and o e s he pa ien speci ic p esc ip ions. AMIE was ained using a ich da ase . These included eal medical dialogues o c ea e a da abase ha closely esembles eal- li e doc o -pa ien in e ac ions; mul iple-choice medical ques ions o imp o e AMIE’s abili y o gene alize and sol e clinical p oblems o a ying complexi y; and expe medical summa ies o p o ide a sys em wi h eliable and alida ed in o ma ion o diagnosis and decision-making (Tu, e al., 2024, p. 4–5). One o he key aspec s o AMIE aining was he sel -lea ning me hod, which is widely used in mode n neu al ne wo ks (see, o example, OpenAI, e al., 2024 o Google, 2024). In a simula ed en i onmen , AMIE in e ac s wi h o he AI agen s, simula ing in e ac ions wi h pa ien s and ecei ing cons an eedback. This allowed he sys em no only o eco d e o s, bu also o imp o e i s skills wi h each new dialogue, g adually mo ing owa ds he expe le el. To es he e ec i eness and accu acy o he AMIE sys em, a TELEMEDICINE AND DIAGNOSTIC ARTIFICIAL INTELLIGENCE A gumen os de Razón Técnica, nº 27, 2024, pp. 215-230 223 blinded s udy was conduc ed in which he sys em was compa ed wi h p ima y ca e physicians based on in e ac ions wi h “simula ed” pa ien s. The esul s showed ha AMIE demons a ed highe diagnos ic accu acy compa ed o doc o s di ec ly. In pa icula , he sys em ou pe o med physicians on 28 o 32 pe o mance indica o s assessed by expe s and on 24 o 26 indica o s assessed by simula o s playing he ole o pa ien s (Tu, e al., 2024, p. 12–15). These esul s con i m ha AMIE is capable o no only suppo ing high-quali y medical dialogues, bu also making a signi ican con ibu ion o imp o ing he quali y o diagnos ics, especially in cases whe e he physician may be limi ed in ime o esou ces. The au ho s o he s udy emphasize he ad an ages o using AMIE, including: 1) A ailabili y: AMIE can ope a e a ound he clock, p o iding suppo o pa ien s ega dless o ime and loca ion; 2. 2) Da a consis ency and objec i i y: he sys em minimizes human e o s associa ed wi h a igue, s ess, o pe sonal biases o medical pe sonnel; 3) Po en ial o lea ning and sel -imp o emen o he neu al ne wo k i sel : AMIE is able o adap o new da a and medical disco e ies, which makes i a p oduc i e ool in he con ex o cons an ly e ol ing medicine (Tu, e al., 2024, p. 17–21). A he same ime, he esea che s no ed se e al limi a ions iden i ied du ing he ope a ion o he neu al ne wo k. The key one was he ex in e ace on which AMIE was launched; messages abou he pa ien ’s condi ion in ex o ma we e a ed by doc o s as unusual and incon enien ; adi ional me hods o in e ac ion wi h pa ien s, di ec con e sa ion, emain amilia o hem. Despi e he success ul esul s in he simula ion, he au ho s poin ou he need o u he es ing o he sys em in eal condi ions o assess i s p ac ical applicabili y and sa e y. And in his case, as no ed ea lie , REGINA PENNER A gumen os de Razón Técnica, n.º 27, 2024, pp. 215-230 230 Google. (2024). PaLM 2. Technical Repo . h ps://ai.google/s a ic/documen s/palm2 ech epo .pd Jobin, A. e al. (2019). The Global Landscape o AI E hics Guidelines. Na u e Machine In elligence, 1(9), 389–399. h ps://doi.o g/10.1038/s42256-019-0088-2 Ma hu , P., Mish a, S., Awas hi, R., Cywinski, J., e al. (2022). A i icial In elligence in Heal hca e: 2021 Yea in Re iew. h ps://doi.o g/10.13140/RG.2.2.25350.24645/1 OpenAI, e al. (4 Ma . 2024) GPT-4 Technical Repo . h p://a xi .o g/abs/2303.08774 Pe e son, M. C., Holb ook, J. H., Von Hales, D., Smi h, N. & S ake , L. (1992). Con ibu ions o he his o y, physical examina ion, and labo a o y in es iga ion in making medical diagnoses. Wes e n Jou nal o Medicine, 156, 163. Pin o-Coelho, L. (2023). How A i icial In elligence Is Shaping Medical Imaging Technology: A Su ey o Inno a ions and Applica ions. Bioenginee ing, 10(12), 1435. h ps://doi.o g/10.3390/bioenginee ing10121435 PRIME S udy P og ess Upda e (2024). Second Pa icipan . h ps://neu alink.com/blog/p ime-s udy-p og ess-upda e-second- pa icipan / Roshan, M., & Rao, A. (2000). A s udy on ela i e con ibu ions o he his o y, physical examina ion and in es iga ions in making medical diagnosis. The Jou nal o he Associa ion o Physicians o India, 48, 771–775. Russell, S., & No ig, P. (2021). A i icial In elligence: A Mode n App oach. Pea son. S auman, E., & Goodie , B. (2011). The Doc o (s) in House: An Analysis o he E olu ion o he Tele ision Doc o -He o. Jou nal o Medical Humani ies, 32(1), 31–46. h ps://doi.o g/10.1007/s10912- 010-9124-2. TELEMEDICINE AND DIAGNOSTIC ARTIFICIAL INTELLIGENCE A gumen os de Razón Técnica, nº 27, 2024, pp. 215-230 231 Tu, T., Palepu, A., Schaeke mann, M., Saab, K., e al. (10 Jan. 2024). Towa ds Con e sa ional Diagnos ic AI. h p://a xi .o g/abs/2401.05654 Wa k, Mc. (2004). A Hacke Mani es o. Ha a d Uni e si y P ess.