Nuclear Medicine

LLM-assisted CDS in Kenyan primary care was safe but did not lower 14-day treatment failure

In 16 Kenyan primary care clinics, clinicians using an EMR with LLM assistance cared for 4,693 patients in the intervention arm and 4,654 in control. Treatment failure within 14 days occurred in 2.2% versus 2.0% (adjusted OR 0.77, 95% CI 0.55–1.08; P = 0.13). No serious adverse events were attributed to the LLM, supporting short-term safety in this low-resource setting. Any clinical benefit, if present, is probably modest and was not demonstrated for the primary outcome. 1

AI in genitourinary oncology shows rapid growth but few randomized trials

Systematic review identified 1,220 eligible AI studies in GU oncology, with rapid publication growth from 2013–2023. Most studies were retrospective (n = 962); only 4 randomized controlled trials were found across prostate, urothelial, and renal cancers. Two RCTs showed diagnostic advantages; two improved prognostication or planning, but risk of bias varied. Overall, AI shows promise in diagnostics and prognostication, yet high-quality randomized evidence remains limited. 2

Perspective: integrate multi-tracer PET/CT, liquid biopsy, and AI for stepped-care theranostics

Authors propose combining multi-tracer molecular PET/CT and liquid biopsy markers with AI/ML to individualize theranostic sequences. They emphasize targeting tumor microenvironment elements and expanding radiopharmaceutical options for better tolerability. The framework suggests a stepped-care algorithm to prioritize radioligand therapy, chemotherapy, or immunotherapy by molecular phenotype. This is conceptual and requires prospective validation before routine clinical adoption. 3