Automatic iceball segmentation with adapted shape priors for MRI-guided cryoablation.

Area of Science:

• Medical Imaging
• Interventional Radiology
• Computational Pathology

Background:

• MRI-guided cryoablation is a minimally invasive treatment for kidney tumors.
• Accurate monitoring of the iceball's 3D configuration is crucial for effective cryoablation.
• Current methods for iceball segmentation can be time-consuming and subjective.

Purpose of the Study:

• To develop and evaluate an automated segmentation method for extracting the 3D iceball configuration during MRI-guided cryoablation.
• To assess the accuracy and robustness of the proposed method using intraprocedural MRI data.

Main Methods:

• Retrospective analysis of 63 timepoints from 13 kidney tumor cryoablation procedures using a 3 Tesla MRI scanner.
• Automatic iceball segmentation employing the graph cut (GC) technique with adapted shape priors, initialized with semiautomatically localized cryoprobes.

Main Results:

• High average Dice Similarity Coefficients (DSC) achieved, reaching 0.93 at later timepoints and an overall average of 0.92 ± 0.03.
• The method demonstrated significant accuracy improvement compared to approaches without shape prior adaptation (P=0.026).
• Segmentation results were unaffected by the number of cryoprobes, and average computation time was 20 seconds, suitable for intraprocedural use.

Conclusions:

• The developed automatic iceball segmentation method is highly accurate and robust.
• This technique is suitable for practical intraprocedural monitoring of MRI-guided cryoablation progress.
• The automated approach enhances the precision and efficiency of cryoablation monitoring.