Related Experiment Video
Updated: May 21, 2026
Thermal Ablation for the Treatment of Abdominal Tumors
Published on: March 7, 2011
Boiling-induced susceptibility artifact correction for clinical 3D PRFS thermometry in hepatic microwave ablation
Dominik Horstmann1,2, Othmar Belker1,2, Daniel Düx1,2
1Department of Diagnostic and Interventional Radiology, Hannover Medical School, Hannover, Germany.
Introduction:
Boiling-induced susceptibility artifacts substantially degrade Proton Resonance Frequency Shift (PRFS) thermometry during hepatic microwave ablation (MWA), leading to systematic temperature errors and biased ablation-zone prediction. Existing volumetric correction approaches assume temporally static artifacts. To address this limitation, this study introduces a dynamic, frame-wise susceptibility correction for clinical 3D thermometry.
Methods:
In 10 patients (12 hepatic lesions) undergoing MRI-guided MWA at 1.5 T, fully reconstructed 3D PRFS temperature maps were retrospectively corrected. For each time frame, the expected thermal evolution was predicted by propagating the previously corrected temperature map using a calibrated Pennes' bioheat simulation; the residual between this prediction and the measured temperature was attributed to susceptibility artifacts. A dynamic bubble-support mask constrained Tikhonov inversion of the susceptibility distribution, followed by dipole-based synthesis and subtraction of the resulting temperature offset, was used to obtain a corrected temperature map. Thermometry-derived ablation zones (CEM43 ≥ 240min) were compared with post-contrast reference segmentations using Dice score (DS), sensitivity, precision, and mean surface distance (MSD) across complementary spatial evaluation regimes.
Results:
Susceptibility correction significantly improved volumetric ablation-zone agreement, with gains across all 3D metrics (ΔDS 6.2 ± 3.3%, Δsensitivity 3.3 ± 3.4%, Δprecision 8.0 ± 4.9%, ΔMSD -0.8 ± 0.6 mm). Effects were smaller in central ablation regions primarily affected by susceptibility-related temperature overestimation, whereas the largest improvements were observed in more peripheral regions dominated by temperature underestimation. Processing time per frame was 1.26 ± 0.12s.
Conclusion:
Dynamic, frame-wise susceptibility correction substantially improves the accuracy of clinical 3D PRFS thermometry during hepatic microwave ablation. Future work should focus on seamless integration into fully real-time clinical thermometry workflows.
