H-SegMed: A Hybrid Method for Prostate Segmentation in TRUS Images via Improved Closed Principal Curve and Improved Enhanced Machine Learning

Prostate segmentation is an important step in prostate volume estimation, multi-modal image registration, and patient-specific anatomical modeling for surgical planning and image-guided biopsy. Manual delineation of the prostate contour is time-consuming and prone to inter- and intra-observer variability. Accurate prostate segmentation in transrectal ultrasound images is particularly challenging due to the ambiguous boundary between the prostate and neighboring organs, the presence of shadow artifacts, heterogeneous intra-prostate image intensity, and inconsistent anatomical shapes. Therefore, in this study, we propose a novel hybrid segmentation method (H-SegMed) for accurate prostate segmentation in TRUS images. The method consists of two main steps: (1) an improved closed principal curve-based method was used to obtain the data sequence, in which only few radiologist-defined seed points were used as an approximate initialization; and (2) an enhanced machine learning method was used to achieve an accurate and smooth contour of the prostate. Our results show that the proposed model achieved superior segmentation performance compared with several other state-of-the-art models, achieving an average Dice similarity coefficient, Jaccard similarity coefficient (Ω), and accuracy of 96.5, 95.1, and 96.3%, respectively.