New analytical and hybrid heat transfer models for thermal ablation procedures validated by MRI thermometry

This study proposes analytical and hybrid models for fast and accurate temperature field reconstruction in microwave ablation (MWA), laser interstitial thermal therapy (LITT), and radiofrequency ablation (RFA), aiming at future real-time clinical use.

The proposed approach combines spatial variable transformation and the Laplace transform for time-dependent terms, with finite difference techniques. A 1 mm isotropic grid represents the voxel network. To ensure accurate temperature representation, voxel-averaged temperatures are computed by integrating the solution of the bioheat equation, under spherical symmetry, over voxel bounds. To approximate the elongated ablation zone, the central circumference of the spherical model is repeated and incorporated into a hemisphere-based geometry. Simulated temperature fields are aligned with experimental MRI data using Advanced Normalization Tools (ANTs). All experiments were conducted ex vivo: MWA in bovine liver, and LITT and RFA in agar phantoms. Regions of interest (ROIs) include voxels with significant thermal variation. Heat source parameters are estimated by minimizing the quadratic difference between simulated and MRI-derived temperatures via a spatiotemporal objective function.

Across all modalities, over 83-98% of voxels presented RMSE $\le$ 1°C, with few exceeding 10°C. LITT showed the best overall agreement. Total simulation and alignment per repetitions required under 0.3 s, significantly below MRI repetition time, enabling potential intraoperative use.

Although approximate and not yet ready for in vivo clinical application, the proposed models offer fast, voxel-level temperature reconstructions. Their computational efficiency supports further development toward real-time monitoring and procedural adjustment during thermal ablation.