TUM team simulates brain tumor mass effect with springs, ~100x faster than PDE inversion
maier_ak · x · 2026-10-15
A TUM-led team (with MCML, Zurich, and FAU) posted the preprint A Differentiable Brain Tumor Mass Effect Solver, replacing costly per-patient PDE inversion for modeling tumor-induced tissue deformation.
- Method: a regular grid over the brain, pushed outward by a potential derived from tumor concentration and held back by a spring lattice with tissue-specific stiffness; energy is minimized end to end with a single global parameter, no per-patient calibration. Runtime is 15 seconds, roughly 100x faster than per-patient PDE inversion.
- Results: on 134 BraTS patients it beats established physics solvers on peritumoural tissue overlap — Dice gain 2.36 vs 0.88 (in hundredths).
- Honest ceiling: measured against a registration ceiling of 8.05, that's about 30% of what's reachable. The authors' verdict: fast, plausible, and honest about the gap.
More from Research
- BigCell Generates Coherent Gigapixel Whole-Slide Pathology Images, Boosting Few-Shot Classification 18% — StonyBrook-CVLab · 2026-10-15
- RAESR Does Super-Resolution in Frozen DINOv3 Latent Space at 37ms per 512px Image — RTP-LLM · 2026-10-15
- NavGPT-3 Reaches Human-Level Navigation: 90.43 SR on RxR-CE vs Human 90.4 — Metacognition-AI · 2026-10-15
- SWEeper-Bench: Best of 15 Frontier Agents Solves Only 59% of Interactive Bug-Finding Tasks — zlab-princeton · 2026-10-15
- Tsinghua's Latent Compaction Compresses KV Cache 32× While Keeping 96.7% Retrieval Accuracy — Yingfa Chen · 2026-10-15
- ByteDance Seed's Cortex Distills Context into Parametric Memory, Keeping 88.9% Gains — ByteDance-Seed · 2026-10-15