RenalCLIP is a kidney cancer-focused vision-language foundation model designed to assess renal masses from CT without invasive sampling. In this multicenter retrospective study, it outperformed conventional CNNs and…

Fundus2Video converts a routine color fundus photograph into a dynamic fluorescein angiography video, offering an AI-synthesized approximation of vascular phase information without dye injection.

Recent Radiology & Imaging coverage signals a maturation: the field is moving beyond narrow task-specific tools toward foundation models, systematic harmonization, and clinical workflow integration—with emerging evidence that…

Medicine is shifting from diagnosis to prediction. Cancer detected five years early. Sleep apnea from infrared video. Prostate cancer detection that beats experts. This week's papers show AI…

A large multi-cohort fMRI study identifies reproducible brain signatures for mentalizing about oneself, others, and mentalizing more generally.

Ophthalmology is shifting rapidly: AI is democratizing screening through smartphones and telemedicine, the eye is being reframed as a systemic health sensor, and prediction is moving from population…

Deep learning applied to routine histopathology slides accurately predicts meningioma molecular classification and recurrence risk without expensive DNA methylation profiling.

Two major vascular diseases—atherosclerosis and abdominal aortic aneurysm—share risk factors but are driven by fundamentally different lipid metabolic pathways, suggesting the need for disease-specific treatments.

A new analysis of over 27,000 abstracts reveals that AI-generated text has infiltrated ophthalmology journals at scale, with over 25% of research articles showing AI signatures and 90%…

Researchers developed MBD-Check, an interpretable machine learning tool that accurately predicts mild bleeding disorders before surgery with 90% sensitivity, outperforming existing clinical screening methods.