Treatment reasoning in medicine is inherently iterative: it requires identifying missing information, gathering evidence from multiple sources, evaluating contraindications and comorbidities, and revising conclusions as new data emerges. Most current AI systems struggle with this because they lack a structured way to seek and integrate evidence before forming a recommendation.

Researchers introduce ATHENA-R1, an AI agent trained to perform treatment reasoning across all FDA-approved drugs since 1939 by interacting with a universe of 212 real biomedical tools (covering indications, interactions, adverse events, disease phenotypes, and more). The agent learns to identify knowledge gaps, call the appropriate tools, incorporate retrieved evidence, and iterate until it reaches a grounded decision.

Sudden cardiac death remains difficult to predict. The current clinical standard — reduced left ventricular ejection fraction (LVEF) — misses most at-risk patients and leads to many unnecessary defibrillator implants.

Researchers used deep learning on a large Swedish dataset of routine ECGs linked to death records and discovered a previously unknown ECG biomarker: a subtle slurred terminal R wave in lead aVL. This pattern is associated with higher risk of sudden cardiac death, likely linked to myocardial fibrosis.

Molecular classification of central nervous system (CNS) tumors via methylation profiling offers the highest diagnostic precision but is expensive, time-consuming, and not widely available. Standard H&E histology slides, by contrast, are routine and globally accessible.

Hetairos is a new AI model that predicts 102 methylation-based CNS tumor subtypes directly from digital H&E slides. Trained and validated on over 11,000 slides from 9,606 patients across 11 centers on four continents, it achieves strong accuracy — especially on high-confidence predictions (0.87–0.88) — and significantly outperforms board-certified neuropathologists when limited to histology alone.

A new AI framework turns routine retinal imaging into a powerful noninvasive window for simultaneous detection of multiple metabolic and endocrine disorders.

The system combines quality-aware modules with frozen foundation models to deliver accurate, interpretable predictions from standard eye scans in under 30 seconds.