188
In e na ional Jou nal o Ad ance and Applied Resea ch
www.ijaa .co.in
ISSN – 2347-7075
Impac Fac o – 8.141
Pee Re iewed
Bi-Mon hly
Vol. 6 No. 38
Sep embe - Oc obe - 2025
Impac o A i icial In elligence on Heal hca e: A Re iew o Cu en
Applica ions and Fu u e Possibili ies
D . Mukesh Tiwa i1 & D . Khalid A. Shaikh2
1Depa men o Mic obiology, D . D.Y.Pa il ACS College Aku di Pune.
2Depa men o S a is ics, D . D.Y.Pa il ACS College Aku di Pune
Co esponding Au ho –D . Mukesh Tiwa i
DOI - 10.5281/zenodo.17313161
Abs ac :
A i icial in elligence (AI) is apidly ans o ming heal hca e ac oss diagnosis, ea men
planning, pa ien moni o ing, d ug disco e y, adminis a ion, and public heal h. This e iew syn hesizes
ecen li e a u e on AI applica ions in heal hca e, highligh s conc e e examples whe e AI has in luenced
clinical p ac ice (medical imaging, diabe ic e inopa hy sc eening, p o ein-s uc u e p edic ion o d ug
disco e y), examines bene i s (accu acy, e iciency, scalabili y), and de ails pe sis en challenges (da a
bias, in e p e abili y, egula o y and e hical conce ns). We also su ey egula o y p og ess and
guidance, and ou line likely nea - e m and long- e m u u e possibili ies, including pe sonalized
medicine d i en by mul i-modal models, AI-assis ed clinical ials, and in eg a ion o la ge language
models (LLMs) in o clinical wo k lows. Finally, we p opose p io i ies o esea ch, go e nance, and
sa e deploymen o maximize bene i while minimizing ha ms. Key policy and scien i ic de elopmen s
indica e AI’s p omise bu unde line he need o obus egula ion, anspa ency, and emphasis on
equi y.
Keywo ds: A i icial In elligence, Machine Lea ning, Heal hca e, Medical Imaging, D ug
Disco e y, E hics, Regula ion, La ge Language Models, Alpha old.
In oduc ion:
A i icial in elligence (AI) — he
applica ion o machine lea ning (ML), deep
lea ning (DL), and ela ed compu a ional
echniques — has apidly shi ed om
esea ch p o o ypes o deployed sys ems in
clinical se ings. Applica ions ange om
diagnos ic imaging and clinical decision
suppo o d ug disco e y and adminis a i e
au oma ion. The COVID-19 pandemic
accele a ed in e es and deploymen o AI
ools o su eillance, diagnosis, and esou ce
planning. Howe e , he ield also aces
challenges: algo i hmic bias, ques ions abou
clinical gene alizabili y, da a go e nance,
explainabili y, and he need o egula o y
amewo ks o ensu e pa ien sa e y and
equi able bene i .
Heal hca e is unde going a p o ound
ans o ma ion d i en by he apid in eg a ion
o digi al echnologies, wi h A i icial
In elligence (AI) eme ging as one o he mos
in luen ial o ces o change. AI e e s o he
capabili y o compu e sys ems o simula e
human cogni i e unc ions such as lea ning,
easoning, and decision-making. In heal hca e,
hese echnologies a e no only suppo ing
clinicians in diagnos ic accu acy and
he apeu ic planning bu a e also eshaping
how pa ien s engage wi h heal h se ices, how
hospi als manage esou ces, and how esea ch
is conduc ed. Unlike con en ional digi al
ools, AI sys ems can p ocess as amoun s o
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
D . Mukesh Tiwa i &D . Khalid A. Shaikh
189
complex da a, ecognize sub le pa e ns, and
gene a e ac ionable insigh s wi h a le el o
e iciency and consis ency ha su passes
adi ional app oaches.
The inc easing adop ion o AI in
heal hca e has been mo i a ed by mul iple
ac o s. The exponen ial g ow h o heal h da a
om elec onic medical eco ds, imaging
modali ies, genomics, wea able de ices, and
emo e moni o ing sys ems has c ea ed a
p essing need o ad anced analy ical
solu ions. A he same ime, heal hca e
sys ems wo ldwide a e acing signi ican
challenges, including ising cos s, a sho age
o skilled p o essionals, aging popula ions, and
he g owing bu den o ch onic diseases. AI
echnologies o e inno a i e solu ions o
add ess hese challenges by enhancing clinical
decision suppo , op imizing hospi al
ope a ions, imp o ing ea ly disease de ec ion,
and enabling pe sonalized medicine.
Cu en applica ions o AI in
heal hca e a e di e se and span almos e e y
aspec o he ca e con inuum. Machine
lea ning algo i hms a e inc easingly applied o
diagnos ic imaging, pa hology, and
de ma ology, whe e hey assis in de ec ing
abno mali ies wi h ema kable accu acy.
Na u al language p ocessing acili a es he
in e p e a ion o uns uc u ed medical no es,
s eamlining eco d-keeping and clinical
documen a ion. AI-enabled p edic i e models
a e p o ing aluable in iden i ying pa ien s a
isk o complica ions, eadmission, o disease
p og ession, hus allowing o p oac i e
in e en ions. Robo ics, powe ed by AI, a e
enhancing su gical p ecision and ehabili a ion
ou comes, while con e sa ional agen s and
cha bo s a e ex ending suppo in pa ien
engagemen , elemedicine, and men al heal h
ca e.
Despi e hese ad ancemen s, he
in eg a ion o AI in o heal hca e also b ings
o h se e al challenges and e hical
conside a ions. Issues such as da a p i acy,
algo i hmic bias, lack o anspa ency in
decision-making (o en e med he ―black box‖
p oblem), and egula o y unce ain y emain
signi ican obs acles o widesp ead adop ion.
Fu he mo e, he human aspec o ca e—
empa hy, us , and communica ion—canno
be eplaced by echnology and needs o be
p ese ed e en as AI sys ems become mo e
p e alen . Ensu ing equi able access o AI-
d i en heal hca e ac oss di e en egions and
popula ions is also a c i ical conce n,
pa icula ly in low- and middle-income
coun ies whe e heal hca e dispa i ies al eady
exis .
Looking ahead, he u u e possibili ies
o AI in heal hca e a e bo h exci ing and
complex. Wi h ad ancemen s in deep lea ning,
p ecision medicine, and in eg a ion wi h o he
eme ging echnologies such as blockchain and
he In e ne o Medical Things (IoMT), AI has
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
D . Mukesh Tiwa i &D . Khalid A. Shaikh
190
he po en ial o c ea e a mo e e icien ,
p edic i e, and pa ien -cen e ed heal hca e
ecosys em. Howe e , ealizing his ision
equi es ca e ul balancing o inno a ion wi h
e hical esponsibili y, mul idisciplina y
collabo a ion be ween clinicians,
echnologis s, and policymake s, and obus
amewo ks o go e nance and egula ion.
This e iew seeks o p o ide a
comp ehensi e examina ion o he cu en
applica ions o AI in heal hca e and o explo e
he u u e possibili ies ha could ede ine he
p ac ice o medicine. By analyzing he
bene i s, limi a ions, and e hical implica ions,
he pape aims o con ibu e o a deepe
unde s anding o how AI can be ha nessed o
imp o e heal h ou comes while main aining
he co e alues o pa ien ca e.
Me hods — li e a u e e iew app oach:
This e iew used a a ge ed na a i e
syn hesis o pee - e iewed jou nals, majo
ins i u ional epo s, and egula o y
announcemen s published h ough 2025.
Da abases and sou ces included
PubMed/PMC, Na u e and Science amily
jou nals, Wo ld Heal h O ganiza ion (WHO)
guidance, and U.S. Food and D ug
Adminis a ion (FDA) documen s. Sea ches
combined e ms such as ―AI in heal hca e
e iew,‖ ―machine lea ning medical imaging,‖
―AI d ug disco e y,‖ ―WHO AI heal h
guidance,‖ and ―FDA AI medical de ices.‖
Recen high-impac examples (e.g.,
AlphaFold, DeepMind oph halmology
sys ems) and egula o y ac ions we e
p io i ized o illus a e majo ends and policy
esponses. This is a quali a i e e iew
in ended o syn hesize con empo a y e idence
and iden i y gaps and di ec ions o u u e
wo k
.
Cu en Applica ions:
1. Medical Imaging and Diagnos ics:
AI has a guably achie ed i s ea lies
clinical impac in medical imaging ( adiology,
pa hology, oph halmology). Con olu ional
neu al ne wo ks (CNNs) and ela ed
a chi ec u es ha e been used o de ec
ac u es, lung nodules, b eas lesions on
mammog ams, and e inal pa hology on
undus pho os and OCT scans. Real-wo ld
e alua ions show AI sys ems can each
sensi i i y and speci ici y compa able o
specialis s in na ow asks, and can educe
eading ime. No able examples include
DeepMind/Moo ields wo k on OCT
in e p e a ion (compa able accu acy o
oph halmologis s) and mul iple FDA-
au ho ized AI imaging ools o iage and
de ec ion. Howe e , eal-wo ld pe o mance
depends c i ically on da a dis ibu ion ma ch
and obus ex e nal alida ion.
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
D . Mukesh Tiwa i &D . Khalid A. Shaikh
191
2. Elec onic heal h eco ds (EHR) and
clinical decision suppo :
Na u al language p ocessing (NLP)
and p edic i e models applied o EHR da a a e
used o p edic pa ien de e io a ion (sepsis,
ICU ans e ), assis in medica ion dosing, and
au oma e documen a ion. LLMs and
specialized medical language models can
summa ize no es, d a discha ge summa ies,
and ex ac s uc u ed da a om ee ex .
While hese ools can educe clinician
adminis a i e bu den, conce ns abou
hallucina ions, da a p i acy, and eliabili y in
clinical easoning pe sis .
3 D ug disco e y and molecula biology:
AI ools accele a e se e al s ages o
d ug disco e y: a ge iden i ica ion, molecule
gene a ion and op imiza ion, and p o ein-
s uc u e p edic ion. AlphaFold (DeepMind)
has ans o med s uc u al biology by
p edic ing p o ein 3-D s uc u es a an
accu acy ha i als expe imen s o many
p o eins; i s impac on a ge cha ac e iza ion
and a ional d ug design is subs an ial. AI-
d i en gene a i e chemis y and i ual
sc eening pla o ms a e being used o p opose
candida e molecules, hough ansla ing AI
ou pu s in o clinically app o ed d ugs emains
challenging and esou ce-in ensi e.
Na u eScienceDi ec
4. Genomics and p ecision medicine:
AI models syn hesize genomic,
ansc ip omic, and clinical da a o isk
s a i ica ion and o sugges he apy op ions
(e.g., umo genomics d i ing a ge ed
he apies). Mul i-omic in eg a ion using
machine lea ning suppo s bioma ke
disco e y and pe sonalized he apeu ic
app oaches, hough ep oducible alida ion
and access o ep esen a i e da ase s a e
ongoing needs.
5. Robo ics, su ge y, and p ocedu al
assis ance:
Su gical obo s (e.g., da Vinci)
combined wi h AI o mo ion analysis,
augmen ed eali y o e lays, and au oma ed
su u ing esea ch a e e ol ing. These sys ems
a e enhancing minimally in asi e p ocedu es,
aining, and in aope a i e decision suppo ,
bu ully au onomous su gical obo s emain
expe imen al.
6. Telemedicine, emo e moni o ing, and
wea ables:
AI algo i hms p ocess con inuous da a
om wea ables (ECG, ac i i y acke s) o
a hy hmia de ec ion, all isk p edic ion, and
ch onic disease moni o ing. Teleheal h
pla o ms inc easingly inco po a e AI iage
and symp om checke s o suppo emo e
consul a ions, expanding access bu aising
ques ions abou accu acy and liabili y.
7. Adminis a i e and ope a ional asks:
AI s eamlines billing, coding,
appoin men scheduling, and esou ce
op imiza ion. These applica ions can educe
cos s and ee clinician ime bu also aise
wo k o ce and ai ness conce ns.
Rep esen a i e case s udies / Examples:
Diabe ic e inopa hy sc eening:
Mul iple deep-lea ning sys ems ha e
demons a ed specialis -le el de ec ion o
e e able diabe ic e inopa hy on e inal
images and ha e been ialed in sc eening
p og ams o inc ease co e age whe e
specialis access is limi ed. PMC
Oph halmic OCT in e p e a ion
(DeepMind & Moo ields): Sys ems
de ec ing a wide ange o e inal
pa hologies om OCT images ma ched
clinician pe o mance in s udies,
illus a ing how imaging AI can iage
and p io i ize pa ien s. STAT
AlphaFold o s uc u al biology:
AlphaFold’s accu a e p o ein s uc u e
p edic ions ha e accele a ed molecula
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
D . Mukesh Tiwa i &D . Khalid A. Shaikh
192
unde s anding and in o med s uc u e-
based d ug design wo k lows. Na u e
Bene i s and oppo uni ies:
Imp o ed diagnos ic accu acy and
speed: AI can de ec sub le pa e ns and
p ocess la ge image olumes as e han
humans in na ow asks.
Scalabili y: Once alida ed, AI ools can
be deployed a scale — pa icula ly
aluable in low- esou ce se ings (e.g., AI
sc eening p og ams).
Pe sonaliza ion: ML models ha
in eg a e mul i-modal da a (imaging +
genomics + EHR) suppo indi idualized
isk s a i ica ion and ea men planning.
D ug disco e y accele a ion: AI educes
ea ly-s age candida e space explo a ion
ime and can sugges no el chemis ies
and a ge s using p o ein s uc u e
p edic ions.
Ope a ional e iciencies: Au oma ion o
documen a ion and adminis a i e
wo k lows can ee clinician ime o
pa ien ca e.
Challenges, isks, and e hical
conside a ions:
Da a quali y, ep esen a i eness, and bias:
Models ained on biased o
un ep esen a i e da ase s isk poo
gene alizabili y and can pe pe ua e heal h
dispa i ies. Sou cing di e se, high-quali y
labeled da ase s and anspa en epo ing o
da ase composi ion a e essen ial. PMC
Explainabili y and us :
Many high-pe o ming models (deep
neu al ne wo ks, LLMs) lack anspa en
easoning pa hways. Clinicians and pa ien s
may dis us sys ems ha canno explain
decisions; explainable AI (XAI) esea ch aims
o b idge his gap bu is no a uni e sal
solu ion.
Sa e y, alida ion, and egula o y o e sigh :
Clinical sa e y equi es igo ous
p ospec i e e alua ion and con inuous
moni o ing pos -deploymen because model
pe o mance can deg ade o e ime as clinical
p ac ice and da a dis ibu ions change.
Regula o s wo ldwide a e issuing guidance o
AI/ML medical de ices; he FDA has issued
and e ined guidance o li ecycle managemen
and ma ke ing o AI-enabled de ices. Robus
p ema ke e idence and pos ma ke
su eillance a e inc easingly emphasized. U.S.
Food and D ug Adminis a ion Exponen
P i acy and da a go e nance:
Use o pa ien da a in aining la ge
models aises consen , de-iden i ica ion, and
da a-sha ing conce ns. Na ional and
in e na ional policy amewo ks mus balance
inno a ion wi h p i acy p o ec ion.
Liabili y and p o essional oles:
Responsibili y o AI-assis ed
decisions — clinician, de elope , o ins i u ion
— needs clea legal amewo ks. Changes o
clinical oles and aining a e necessa y i AI
akes o e ou ine asks.
Hallucina ion and misin o ma ion ( o
LLMs):
La ge language models can p oduce
plausible bu inco ec ou pu s
(―hallucina ions‖), which is haza dous in
clinical con ex s unless p ope ly cons ained
and alida ed.
Regula ion, go e nance, and e hics —
ecen de elopmen s: In e na ional bodies
and egula o s ha e begun issuing guidance o
go e n heal h AI. The WHO has published
e hical and go e nance ecommenda ions o
AI in heal h emphasizing equi y, anspa ency,
and human o e sigh . Na ional egula o s,
including he FDA, ha e de eloped guidance
on AI/ML in medical de ices, add essing
p ema ke e idence, cybe secu i y, and
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
D . Mukesh Tiwa i &D . Khalid A. Shaikh
193
p ede e mined change con ol plans o
adap i e sys ems. These de elopmen s e lec
ecogni ion ha AI equi es li ecycle
o e sigh : om da a p o enance and model
e alua ion o pos -ma ke moni o ing and
handling o model upda es. Wo ld Heal h
O ganiza ion U.S. Food and D ug
Adminis a ion
Fu u e Possibili ies (nea - e m and long-
e m):
Nea - e m (1–5 yea s):
B oad in eg a ion o AI iage and
decision suppo in o EHR wo k lows
— au oma ed ale s o de e io a ion,
medica ion in e ac ions, and
documen a ion assis ance.
LLM-powe ed documen a ion and
in o ma ion e ie al — educing
clinician adminis a i e bu den while
equi ing gua d ails o p e en
hallucina ions.
AI-assis ed imaging iage deployed in
sc eening p og ams o expand access in
unde se ed egions.
Imp o ed model moni o ing and
―model-as-medical-de ice‖ p ac ices —
con inuous alida ion pipelines, da a d i
de ec ion, and sa e model upda es unde
egula o y amewo ks.
8.2 Mid/long- e m (5–15+ yea s)
Pe sonalized, mul i-modal AI clinicians:
in eg a ed sys ems ha combine genomics,
con inuous physiologic moni o ing,
imaging, and social de e minan s o
deli e ailo ed ca e plans.
AI-accele a ed d ug pipelines:
compu a ional pla o ms ha signi ican ly
sho en lead disco e y and op imize
molecules—po en ially educing cos s and
imelines o ce ain d ug classes, bu s ill
equi ing clinical ials. AlphaFold and
successo models will unde pin many
design wo k lows. PMC
Assis i e au onomy in ce ain
p ocedu es: semi-au onomous obo s o
epe i i e su gical asks unde human
supe ision.
Popula ion-le el public heal h
su eillance: AI sys ems o ea ly
ou b eak de ec ion, esou ce alloca ion
o ecas ing, and epidemic con ol.
Resea ch & policy p io i ies:
To enable sa e and equi able bene i s
om AI in heal hca e, p io i ies include:
1. High-quali y, di e se da ase s wi h clea
p o enance and s anda dized epo ing.
2. P ospec i e, mul i-cen e clinical ials
and eal-wo ld e idence o AI ools, no
only e ospec i e alida ion.
3. In e ope abili y s anda ds and
anspa en e alua ion me ics o
compa ing algo i hms ac oss se ings.
4. Robus egula o y amewo ks ha
add ess adap i e sys ems, pos -ma ke
su eillance, and anspa ency
equi emen s.
5. E hics and ai ness audi ing wi h
s akeholde pa icipa ion, especially om
communi ies a isk o being
disad an aged by biased models.
6. Clinician aining and ole edesign o
wo k wi h AI, including how o in e p e
ou pu s and manage ailu e modes. Wo ld
Heal h O ganiza ionExponen
Limi a ions o cu en e idence:
Many AI s udies emain single-cen e ,
e ospec i e, o use con enience da ase s ha
o e s a e eal-wo ld pe o mance.
He e ogeneous epo ing s anda ds make
c oss-s udy compa isons di icul . The apid
pace o AI inno a ion also means ha e iews
can become quickly da ed; con inuous
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
D . Mukesh Tiwa i &D . Khalid A. Shaikh
194
syn hesis and li ing sys ema ic e iews a e
needed o keep guidance cu en . PMC
Conclusion:
AI p esen s ans o ma i e po en ial
o heal hca e: imp o ed diagnos ics, no el
d ug disco e y wo k lows, ope a ional
e iciencies, and expanded access o ca e.
Howe e , ealizing hese bene i s equi ably
and sa ely demands igo ous alida ion,
anspa en go e nance, and global
coo dina ion on e hics and egula ion.
Ad ances such as AlphaFold demons a e AI’s
powe o sol e p e iously in ac able scien i ic
p oblems, while egula o y mo es om bodies
like he WHO and FDA show policymake s
a e adap ing o he echnology’s eali ies. The
u u e landscape will likely ea u e hyb id
sys ems whe e human clinicians and AI co-
ope a e — combining human judgmen ,
alues, and o e sigh wi h machine scalabili y
and pa e n ecogni ion. Fo policymake s,
clinicians, and esea che s, he challenge is o
accele a e inno a ion while embedding
sa egua ds ha p o ec pa ien s and ensu e
equi able dis ibu ion o bene i s. Na u e
Wo ld Heal h O ganiza ion
Acknowledgmen : The au ho s g a e ully
acknowledge inancial suppo om he D . D.
Y. Pa il ACS College, Aku di, Pune,
Maha ash a.
Re e ences:
1. Alhejaily AMG. A i icial in elligence in
heal hca e (Re iew). PMC 2024. PMC
2. Al Kuwai i A, e al. A Re iew o he Role
o A i icial In elligence in Heal hca e.
PMC 2023. PMC
3. Wo ld Heal h O ganiza ion. Ha nessing
a i icial in elligence o heal h / WHO
guidance on AI in heal h. (WHO digi al
heal h & inno a ion pages). Wo ld Heal h
O ganiza ion+1
4. U.S. Food and D ug Adminis a ion.
A i icial In elligence and Machine
Lea ning in So wa e as a Medical De ice
(FDA pages; d a / inal guidance). 2024–
2025 upda es. U.S. Food and D ug
Adminis a ionExponen
5. Jumpe J, E ans R, P i zel A, e al. Highly
accu a e p o ein s uc u e p edic ion wi h
AlphaFold. Na u e 2021. Na u e
6. Desai D. Re iew o AlphaFold 3:
T ans o ma i e Ad ances in D ug Design
and The apeu ics. PMC 2024/2025. PMC
7. Kong M, e al. A i icial In elligence
Applica ions in Diabe ic Re inopa hy.
PMC 2024. PMC
8. Khali a M, e al. AI in diagnos ic imaging:
Re olu ionising accu acy and ... 2024.
ScienceDi ec /Na u e e iew.
ScienceDi ec
9. Goodman KW, e al. Global heal h and big
da a: The WHO's a i icial ... (E hics &
go e nance e iews). PMC 2023. PMC
10. Chus ecki M. Bene i s and Risks o AI in
Heal h Ca e: Na a i e Re iew. In e ac J
Med Res. 2024;13(1):e53616. IJMR
11. Khan MM, e al. Towa ds Secu e and
T us ed AI in Heal hca e: A Sys ema ic
Re iew. (2010–2023). Compu Me hods
P og ams Biomed. 2024; e-pub.
ScienceDi ec
12. Bo ha NN, e al. Pe cei ed Th ea s Posed
by he Use o AI Tools in Heal hca e on
Pa ien s’ Righ s and Sa e y: A Scoping
Re iew (2010–2023). A ch Public Heal h.
2024; e-pub. BioMed Cen al
13. Mo one G, e al. A i icial In elligence in
Clinical Medicine: O e iew o Sys ema ic
Re iews. F on Digi Heal h. 2025; e-pub.
F on ie s
14. Awas hi R. A i icial In elligence in
Heal hca e: 2024 Yea in Re iew.
medRxi . 2025; e-pub. MedRxi
15. De Micco F, e al. A i icial In elligence
in Heal hca e: T ans o ming Pa ien
Clinical Risk Managemen . F on Med.
2025; e-pub.
