Comparison of Four Auto-Contouring Software for Prostate Cancer Structures

Read full text Summarise with VIRAT
Authors: Vannyat Ath, Sang Hyoun Choi, Kum Bae Kim, Jae Choon Lee, Dinh Chau Nguyen
Cited by: 0
References: 30

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Purpose: This study aimed to compare the accuracy of four auto-contouring models.Retrospective data from 10 patients with prostate cancer were used as the manual reference contours.Methods: Software-generated contours were evaluated using multiple metrics: the dice similarity coefficient (DSC) for spatial overlap; the 95th percentile Hausdorff distance to measure surface distance; relative volume difference for volumetric analysis; and centroid error to evaluate positional accuracy.Results: The femoral heads achieved the highest DSC values across all software, with AIS-1 demonstrating the greatest accuracy (0.97±0.01), reflecting the advantage of well-defined bone structure boundaries.AIS-2 performed optimally for bladder segmentation (DSC=0.94±0.02),whereas rectal segmentation demonstrated greater variability, although AIS-2 still yielded the highest performance (0.85±0.08).Improved results were observed for the anorectum structure, particularly for AIS-4 (DSC=0.91±0.02),likely due to more effective contouring of the junction between the rectum and anus.In contrast, anal segmentation posed the greatest challenge across all software, with consistently low accuracy across all metrics.Conclusions: The findings highlight the differential strengths of each software, with higher segmentation accuracy observed for bone structures, such as the femoral heads, than for soft tissues, such as the rectum and anus.Overall, AIS-1 demonstrated superior performance in femoral head and bladder segmentation, AIS-2 performed best in rectal segmentation, and AIS-4 excelled in anorectal segmentation.

Fields of Study & Subjects

Medical Image Segmentation Techniques Advanced Radiotherapy Techniques Radiomics and Machine Learning in Medical Imaging

Publication details

Journal / source
Progress in Medical Physics
Volume
37
Issue
1
Pages
8–18
Publication date
26 March 2026
Language
EN