A real-time artificial intelligence model accurately mapped several critical anatomical structures and dissection planes during total mesorectal excision, while processing external test images far faster than manual expert…

FA-NET-NT detected free intraperitoneal air on abdominal CT with high segmentation accuracy and maintained 95% patient-level sensitivity in both temporal and external validation cohorts. The model is intended…

Oncology AI is moving upstream to earlier risk signals and downstream into clinical workflows. The connective challenge is ensuring that models capture transferable cancer biology, improve decisions, and…

Foundation models achieved near-perfect accuracy on official German medical licensing questions, but aggregate scores concealed important weaknesses. Images disproportionately increased model errors, while the questions models found difficult…

Nearly three in ten adults with depression met the study definition of psychotherapy dropout, and most dropouts occurred immediately after the first session. Electronic health record data helped…

Routine kidney-biopsy narratives contain clinically relevant signals that standardized scores may not fully capture. In 3,078 patients with IgA nephropathy, a language-model pipeline identified two reproducible pathological subtypes…

FatigueVisual combines adaptive sweat handling, multimodal wearable sensing, and machine learning to monitor exercise-induced fatigue continuously. In a small intervention experiment, its real-time guidance reduced the fatigue index…

Internet-based cognitive behavioral therapy produced benefits even among patients whom a prediction model ranked least likely to respond. The model detected meaningful variation in relative benefit, but not…

PRECISE-HF replaces a single NT-proBNP cutoff with an individualized probability of heart failure. In internal validation, it ruled out nearly two-thirds of tested patients at 90% sensitivity while…

More than one-third of Medicaid beneficiaries receiving care coordination had at least one documented administrative burden. Together, these obstacles accounted for an estimated 18,822 hours of patient time.