Assessing the suitability of automated registration and segmentation for dosimetry calculations in SIRT treatment planning.

Abstract

Background: Selective internal radiation therapy (SIRT) increasingly relies on accurate magnetic resonance imaging (MRI) to computed tomography (CT) registration and accurate liver and tumour segmentation, for effective pre-treatment planning. This study evaluates the impact of automatic registration and segmentation techniques on dosimetry calculations for SIRT treatment planning. It compares semi-automatic and automatic registration, as well as manual and automatic deep learning-based segmentation.

Methods: Pre-treatment data from 90 patients with hepatocellular carcinoma were analysed. The dataset consisted of contrast-enhanced T1-weighted MRI scans with manually delineated liver and tumour volumes of interest (VOIs), as well as single-photon emission computed tomography (SPECT)/CT scans with manually or semi-automatically delineated liver VOI. The clinical routine pipeline, which involves semi-automatic registration and manual/semi-automatic segmentation, was used as the baseline pipeline and compared to experimental pipelines that use intensity-based deformable automatic registration or deep learning-based automatic segmentation. Dosimetric accuracy was assessed via metrics such as the mean absorbed dose, the minimum dose received by 70% of the volume (D70), and inverted cumulative dose-volume histograms.

Results: Semi-automatic and automatic MRI to CT liver registration achieved comparable Dice scores of 92%. However, tumour registration varied significantly between registration methods yielding average Dice scores of 79%. Multimodal tumour segmentation approaches outperformed monomodal ones, achieving average Dice scores of 66.6 versus 62.5%. Using the baseline pipeline, the average tumour absorbed dose per patient was 115.6 Gy. Using the fully automatic approach, tumour absorbed doses differed from baseline values by an average of 6.6 Gy. Differences ranged from 0.3 Gy in the best case to 202.8 Gy when the automatic tumour segmentation markedly deviated from the manual delineation. Finally, it was found that a Dice score of at least 80% was required to avoid statistically significant differences in absorbed dose estimates between the clinical and automatic approaches.

Conclusion: Semi-automatic and automatic registration show equivalent performance, allowing for complete automation. Automatic segmentation demonstrates promising results, with approximately 40% of patients achieving tumour Dice scores above the 80% threshold. 30% of cases show intermediate performance (60-80% tumour Dice scores), while the remaining 30% are still challenging. Further refinement of segmentation methods is required to enhance dosimetric accuracy.