Computer-Aided Diagnosis in Medicine

Computer-aided diagnosis has the longest history of any AI technology in medicine, with mammography CAD systems deployed since the late 1990s. Modern CADx has evolved far beyond those early systems, leveraging deep learning to achieve diagnostic accuracy that approaches or exceeds specialist physicians for specific conditions. CADx operates as a clinical decision support tool — analyzing images, signals, or data and providing diagnostic suggestions that physicians incorporate into their clinical judgment. The FDA classifies most CADx systems as Class II medical devices, with a well-established regulatory pathway.

Breast cancer screening CADx remains the most widely deployed application, with AI-enhanced systems replacing the first-generation rule-based CAD that achieved limited clinical acceptance. Modern deep learning mammography AI detects calcifications, masses, and architectural distortions with sensitivity exceeding 90%, while reducing false positives that plague traditional CAD. Studies in Europe have shown that AI mammography screening can replace one of the two radiologists in double-reading protocols without loss of sensitivity. IDx-DR (Digital Diagnostics) represents the pinnacle of CADx — the first FDA-authorized autonomous AI diagnostic that provides a diabetic retinopathy diagnosis from fundus photographs without physician review, now deployed in primary care settings and pharmacies to expand screening access.

CADx is expanding beyond imaging into multi-modal diagnosis. Cardiac CADx analyzes ECGs, echocardiograms, and cardiac CT data to detect arrhythmias, wall motion abnormalities, and coronary artery disease — with Cleerly providing AI-powered coronary artery analysis that quantifies plaque burden. Prenosis's AI sepsis diagnostic combines routine laboratory values into a predictive score, representing a non-imaging CADx approach. Dermatology CADx evaluates skin lesion photographs for melanoma risk. Pathology CADx from Paige AI grades prostate cancer tissue. The trend is toward multi-modal CADx that integrates imaging with clinical data, laboratory results, and genomics for more comprehensive diagnostic assessment — moving beyond single-test analysis toward holistic diagnostic reasoning.