Physics-based Fisher-KPP stroke model beats rCBF thresholding: Dice 0.46 vs 0.25
maier_ak · x · 2026-09-11
A physics-based stroke model embeds the Fisher-KPP reaction-diffusion equation into a patient's 3D MRI grid, with diffusion D(x) derived from the Tmax perfusion map and a logistic reaction term driving cell death.
On 29 ISLES 2017 patients it achieved mean Dice 0.46±0.24 and AUC-ROC 0.90±0.10, beating standard rCBF thresholding (Dice 0.25, AUC 0.70). Adding the diffusion term raised Dice by 0.12 over a reaction-only variant.
Related event: Physics-Based Reaction-Diffusion Model Improves Stroke Lesion Prediction(2 posts)→
More from Research
- LoMa and RoMa v2 to be presented at ECCV 2026 along with Quick ViTs — ducha_aiki · 2026-09-11
- UMO: unified in-context framework turns text2motion model into multi-task motion foundation — Michael_J_Black · 2026-09-11
- Clay Institute Changes Navier-Stokes Equation Status from Unsolved to Active — felpix_ · 2026-09-11
- First AI4AI Survey Maps Why AI Can't Yet Reliably Improve AI: The Composition Gap — 新智元 · 2026-09-11
- Independent kernel-level verification of OpenAI's Navier-Stokes Lean proof: all four builds pass — pvaa · 2026-09-11
- ACL Caps Authors to Save Reviewing, as Kyunghyun Cho Proposes ScholarCoin Token Economy — kchonyc · 2026-09-11