Related Experiment Video
Updated: Dec 18, 2025
04:25
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
3.5K
Segmentation of prostate zones using probabilistic atlas-based method with diffusion-weighted MR images
Dharmesh Singh1, Virendra Kumar2, Chandan J Das3
1Centre for Biomedical Engineering, Indian Institute of Technology Delhi, New Delhi, India.
Computer Methods and Programs in Biomedicine
|June 17, 2020
Summary
This study presents a novel semi-automated framework for segmenting the prostate gland and its zones using diffusion-weighted imaging (DWI). The method achieves high accuracy, improving computer-aided diagnosis for prostate cancer (PCa).
Area of Science:
- Medical Imaging
- Radiology
- Biomedical Engineering
Background:
- Accurate prostate segmentation is crucial for computer-aided diagnosis of prostate cancer (PCa) using diffusion-weighted imaging (DWI).
- Challenges in DWI segmentation include low signal-to-noise ratio and high variability in prostate anatomy.
- A semi-automated framework is proposed to simultaneously segment the prostate gland and its zones.
Purpose of the Study:
- To develop and validate a semi-automated framework for prostate and prostate zone segmentation using DWI.
- To assess the accuracy and stability of the proposed segmentation methodology.
- To evaluate the impact of partial volume (PV) correction on zonal segmentation performance.
Main Methods:
- The Chan-Vese active contour model and morphological opening were used for prostate gland segmentation.
- An in-house developed probabilistic atlas with partial volume (PV) correction algorithm was employed for peripheral zone (PZ) and transition zone (TZ) segmentation.
- The methodology was validated on a dataset of 18 patients and independently tested on the QIN-PROSTATE-Repeatability dataset (15 patients), using Dice similarity coefficient (DSC), Jaccard coefficient (JC), and accuracy.
Main Results:
- The proposed method achieved high segmentation accuracy for the prostate gland (DSC: 90.76% ± 3.68%), PZ (DSC: 77.73% ± 2.76%), and TZ (DSC: 86.05% ± 1.50%) on the primary dataset.
- Independent testing on the QIN-PROSTATE-Repeatability dataset yielded comparable results (Prostate DSC: 85.50% ± 4.43%, PZ DSC: 74.40% ± 1.79%, TZ DSC: 85.80% ± 5.55%).
- Statistically significant improvements (p<0.05) in DSC, JC, and accuracy were observed for both PZ and TZ segmentation with the PV correction algorithm.
Conclusions:
- The developed semi-automated segmentation framework is stable, accurate, and user-friendly for segmenting the prostate gland and its zones (PZ and TZ).
- The atlas-based segmentation with PV correction shows promise for integration into computer-aided diagnostic systems for PCa localization and treatment planning.
