Key Highlights
•
A study using six large language models (LLMs) successfully screened unstructured clinical notes from 110 epilepsy patients, identifying surgical candidates with near-perfect accuracy. The analysis found that 45% of patients who met surgical criteria had never been previously considered for surgery, highlighting a major gap in current clinical practice.
Source →
•
Researchers developed a quantitative model to predict methotrexate-induced nephrotoxicity in over 700 patients with primary central nervous system lymphoma. The model identifies safe concentration thresholds at 24, 48, and 72 hours—9.71, 0.81, and 0.26 µmol/L respectively—to help clinicians prevent kidney damage.
Source →
•
Genomic newborn screening raises important questions about how long to retain sequencing data for research and clinical reuse. A new article in the European Journal of Human Genetics explores the ethical and practical considerations of data retention policies for these valuable datasets.
Source →
Stay curious. Stay informed —
Science Briefing
