Pleomorphic dermal sarcoma closely resembled metastatic cutaneous squamous cell carcinoma across multiple genomic measures, while appearing distinctly different from other soft-tissue sarcomas. These findings may help reframe how…

More than one in five surveyed students reported turning to an AI chatbot for help with feelings or relationships. These students had substantially more emotional problems, even after…

APOE ε4 was associated with a conserved, dose-dependent inflammatory signature across peripheral and central tissues—even without an Alzheimer’s disease diagnosis. In patient-derived cortical organoids, immune activation appeared before…

The American College of Physicians places a clear ethical boundary around clinical AI: it should strengthen the patient–physician relationship, not displace the judgment, skills, and responsibility that sustain…

Neurology AI is shifting from episodic classification toward earlier, trajectory-aware intervention. Passive signals, longitudinal biomarkers, interpretable phenotypes, and scrutiny of evidence systems define the current direction.

Computer-aided feedback on colonoscopy technique increased adenoma detection after a positive faecal immunochemical test, despite already high detection rates. The multicentre Danish trial suggests that systematic monitoring can…

A CT-based deep learning model identified pathological high-risk features in early-stage invasive lung adenocarcinoma with substantially greater sensitivity than intraoperative frozen-section assessment. Its preoperative predictions could help surgeons…

High TNFAIP6 expression marked more aggressive low-grade glioma and poorer survival, while laboratory experiments indicated that TNFAIP6 directly promotes malignant behavior through JAK3-STAT6 signaling. Together, the findings position…

An integrated computational and experimental study identifies RAD51C as a central gene associated with aggressive bladder cancer. Silencing RAD51C impaired DNA repair and tumor-promoting behavior in cells and…

LLM-based digital twins can appear convincingly personalized while failing to reproduce the people they represent. Across 164 outcomes, their responses showed weak individual fidelity, compressed variability, and five…