Microsoft Proposes RadFusion for Threshold-Controllable Radiology Reports
erichorvitz · x · 2026-08-15
Microsoft Research introduces RadFusion, a framework enabling threshold-controllable radiology report generation. Existing models lack control over the sensitivity-specificity trade-off crucial for different clinical scenarios like triage vs. confirmation. RadFusion fuses a multi-label classifier with a VQA-based generator, using an LLM to align diagnoses with classifier decisions at a selected threshold. This conformance allows generated reports to be quantitatively evaluated via ROC analysis, aiding regulatory clearance and adapting to diverse clinical needs.
More from Research
- ReLU Paper Review: Solving the Vanishing Gradient Problem — burkov · 2026-08-15
- Memories persist even after half of brain connections wiped out, study finds — mtizard · 2026-08-15
- SPP Paper: Alignment from Token Zero improves robustness to jailbreaks — dhadfieldmenell · 2026-08-15
- AutoGaze reduces visual tokens by 4x-100x for efficient 4K video understanding — Cohere · 2026-08-15
- OpenAI's Unreleased 'Astra' Model Solves 10 Math Problems — thursdai_pod · 2026-08-15
- WanSong: Pure Diffusion Model for 5-Minute Songs — Crazy-Repeat-2006 · 2026-08-15