Abstract
Cryoablation is a minimally invasive tumor treatment whose success depends on accurate procedure planning under patient-specific anatomical and biophysical variability. Computational models could facilitate planning by forecasting post-operative ablation zones. However, they require temperature-dependent tissue properties for highly accurate predictions, which are difficult to measure across cryogenic ranges. These properties could instead be inferred by personalizing such models to experimental data, yet phase-change-induced nonlinearities often require flexible function representations with potentially many tunable parameters, making conventional optimization methods loose tractability. We propose a differentiable physics framework to personalize a cryoablation bioheat model from temperature observations using gradient-based optimization. Tissue properties are represented as bounded B-splines and optimized end-to-end, with a curvature penalty to reduce non-physical oscillations. In synthetic bench test experiments, we evaluated the method on target functions of increasing complexity and compared it with BFGS andNelder-Mead.Differentiable-physics-basedoptimizationremainedeffective as the number of trainable parameters increased, achieving competitive or better recovery of tissue properties, while requiring substantially fewer function evaluations than the comparators. These results suggest that differentiable physics is a parametrically scalable approach for estimating non-linear thermal tissue properties in cryoablation modeling.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/DT4H_034.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=JwBv2AY1Uz
BibTex
@InProceedings{SamRau_Learning_MICCAISAT2026,
author = { Samanta, Raunak AND Raghunath, Adarsh AND Audigier, Chloé AND Matei, Teodor Ionut AND Frings, Oliver AND Maier, Andreas AND Meister, Felix},
title = { { Learning Temperature-Dependent Properties in Cryoablation with Differentiable Physics } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17275},
month = {pending},
page = {pending}
}