Nuclear Medicine

MARLIN: practical framework for incident learning systems in European nuclear medicine

MARLIN operationalised recommendations for incident learning systems across Europe during a 24-month mixed-methods project. Surveys and interviews showed wide variability in NM reporting criteria; only 11/23 countries had NM-specific criteria. The study identifies enablers—just culture, multidisciplinary teams, and standardised criteria—and practical guidance for ILS setup. Implementation will require coordination among competent authorities, professional societies, and clinical facilities. 1

Real-world LLM interventions show mixed clinical outcomes and poor reporting completeness

This systematic evidence map of 55 studies found most interventions used human-AI collaboration for decision support. LLM-assisted interventions showed benefits in psychological endpoints but diagnostic accuracy in randomized trials was variable (range 0.65–0.88). Reporting quality was suboptimal with mean CONSORT-AI adherence around 78.8%, and important omissions in data quality and error handling. The evidence is heterogeneous and insufficient to support routine autonomous LLM deployment in clinical care. 2

PSMA RLT: durable palliation and responses but limited long-term disease control

Lutetium-177 PSMA radioligand therapy yields high objective responses and rapid pain relief in advanced prostate cancer. Randomised trials show only modest survival benefit and near-universal eventual disease recurrence. Maximising durable benefit likely requires imaging-guided adaptive strategies, earlier-line testing, or novel isotopes and combinations. For now, PSMA RLT remains valuable for symptom control and selected patients, not a cure for disseminated disease. 3