سامانه همایش های دانشگاه آزاد اسلامی واحد نجف آباد

صفحه اصلی / اولین همایش بین المللی تحول خدمات سلامت: تلفیق مراقبت، درمان، بازتوانی و امیدبخشی

Artificial Intelligence and Machine Learning in Disease Prediction and Management: From Technical Performance to Clinical Value

نویسندگان :

Amirhossein Nikkhah¹ Malihe Jahromi²

1- موسسه آموزش عالی شهید اشرفی اصفهانی 2- دانشگاه آزاد اسلامی واحد نجف آباد

کلمات کلیدی :

artificial intelligence،machine learning،disease prediction،clinical decision support،prognostic modeling،medical imaging،digital health

چکیده :

Artificial intelligence (AI) and machine learning (ML) are now embedded in the vocabulary of modern clinical medicine, but their practical impact is not uniform across specialties, care settings, or outcome types. During the past decade, AI research has moved from algorithm development on retrospective datasets toward selective implementation in real clinical workflows. The most convincing progress has been observed in image-intensive fields, including ophthalmology, dermatology, pathology, and breast imaging, where deep learning systems can achieve high diagnostic discrimination and support clinician efficiency in specific tasks. At the same time, EHR-based models for risk prediction and early warning have expanded rapidly in cardiology, infectious disease, and inpatient medicine, often outperforming conventional risk scores in internal analyses. However, strong performance in development studies does not always translate into stable benefit after deployment. External validation studies repeatedly show the effects of dataset shift, calibration drift, documentation variability, and differences in patient case-mix across institutions. A second major issue is that much of the published literature emphasizes technical metrics such as AUROC while providing limited evidence on patient-centered endpoints, safety, workflow burden, or equity impact. Bias and representativeness challenges remain central, particularly when models trained on narrow populations are used in diverse health systems. Regulatory oversight for AI-enabled software has advanced, yet governance of adaptive models and post-deployment monitoring is still maturing. This review synthesizes current evidence and implementation experience across disease prediction and management use cases, with attention to what is clinically reliable, what remains uncertain, and what conditions are required for responsible scale-up. The central conclusion is that AI in healthcare has entered an accountability phase: success now depends less on achieving marginally higher benchmark performance and more on demonstrating reproducible, equitable, and operationally sustainable clinical value.