Field Notes on Clinical AI

Long and short-form analysis of benchmarking methods, healthcare system constraints, and what responsible deployment requires in practice.

Technical report cover for the RadLE benchmark protocol

Benchmark Protocol

RadLE v1 Technical Report: Benchmark Design and Evaluation Protocol

A working technical note on the RadLE case mix, blinded expert review process, scoring rubric, and model evaluation setup used for radiology reasoning benchmarks.

Technical report cover for evaluation and reliability metrics

Evaluation Framework

RADAR and TRUST Metrics: Draft Reliability Audit Framework

Temporary report summary for reliability, attribution, and inter-reader agreement checks in agentic radiology report generation workflows.

Technical report cover for clinical AI data readiness

Data Infrastructure

Clinical AI Data Readiness Checklist for Imaging Collaborations

A practical checklist for imaging partners covering DICOM hygiene, de-identification, governance, annotation planning, and benchmark-ready dataset preparation.