PCG-cut: graph driven segmentation of the prostate central gland

Area of Science:

• Medical imaging
• Computational anatomy
• Oncology

Background:

• Prostate cancer is a leading cancer in men, necessitating advanced diagnostic and treatment tools.
• Accurate segmentation of the prostate central gland (PCG) in MR scans is crucial for effective biopsy and radiation therapy planning.

Purpose of the Study:

• To develop and evaluate an automated graph-based algorithm for segmenting the prostate central gland (PCG) in MR images.
• To assess the algorithm's potential for improving MR-guided biopsy and radiation treatment planning.

Main Methods:

• A graph-driven segmentation algorithm utilizing a spherical template and ray sampling was employed.
• The algorithm constructs a graph from user-defined seed points and computes a minimal cost closed set using polynomial time s-t-cut for boundary delineation.
• The approach was implemented as a C++ module in MeVisLab, with manual segmentations by expert radiologists serving as ground truth.

Main Results:

• The automated segmentation algorithm achieved an average Dice Similarity Coefficient (DSC) of 78.94 ± 10.85% when compared to manual segmentations.
• The graph-based method successfully delineated the boundaries and volume of the prostate central gland.

Conclusions:

• The proposed automated segmentation method demonstrates promising results for PCG delineation in MR scans.
• This automated approach can potentially enhance the efficiency and accuracy of MR-guided prostate interventions.