Prostate Volume Segmentation in TRUS Using Hybrid Edge-Bhattacharyya Active Surfaces

Area of Science:

• Medical imaging analysis
• Biomedical engineering
• Computer-aided diagnosis

Background:

• Accurate prostate volume segmentation is crucial for radiation therapy planning and monitoring.
• Transrectal ultrasound (TRUS) is a common imaging modality for prostate assessment.
• Existing segmentation methods face challenges with low contrast and initialization.

Purpose of the Study:

• To introduce a novel hybrid edge and region-based parametric deformable model (active surface) for prostate segmentation in TRUS images.
• To enhance the accuracy and robustness of prostate volume segmentation in challenging imaging conditions.

Main Methods:

• Development of a new radial bas-relief-based edge detector for improved prostate edge detection in low contrast TRUS images.
• Integration of an edge-based force with a Bhattacharyya gradient flow-driven region-based force for parametric active surfaces.
• Implementation of a quasi-automatic initialization technique using radial edge detector response analysis for surface model warping.

Main Results:

• Validation on 36 TRUS images with manual delineations by expert radiation oncologists.
• The proposed hybrid model achieved state-of-the-art segmentation accuracy.
• Quantitative metrics confirmed the effectiveness of the hybrid framework.

Conclusions:

• The developed hybrid active surface model provides a significant advancement in prostate volume segmentation accuracy.
• The modular framework offers a valuable addition to the strategies for addressing prostate segmentation challenges in TRUS.
• This approach holds promise for improving clinical decision-making in prostate cancer management.