Source
IJCAI
DATE OF PUBLICATION
08/16/2025
Authors
Dmitriy Umerenkov Aleksandr Nesterov Vladimir Shaposhnikov Elena Sokolova Ruslan Abramov Nikolay Romanenko Vladimir Kokh Marina Kirina Anton Abrosimov Dmitry Dylov Ivan Oseledets
Share

AI Diagnostic Assistant (AIDA): A Predictive Model for Diagnoses from Health Records in Clinical Decision Support Systems

Abstract

Clinical Decision Support Systems (CDSS) playan increasingly important role in medical diagnostics.We present AI Diagnostic Assistant (AIDA),a real-time predictive model designed to assist doctorsin interpreting patient conditions while workingin a CDSS. AIDA analyzes electronic healthrecords (EHR), including medical history, laboratoryresults, and complaints, to suggest potentialdiagnoses from 95 common conditions before doctormakes final decision. The model acts as a verificationand backup tool, ensuring that no criticaldetails are overlooked. Trained on 1.5 million patientrecords and validated on a dataset curated bya panel of experts, AIDA proves trustworthy as adiagnosis-making assistant (87.7% accuracy comparedto 91.7% accuracy among doctors).Integrated into a megapolis-wide CDSS, AIDA hasassisted doctors in making over 3 million realworlddiagnoses to date.



Join AIRI