Membranous Urethral Reconstruction and Segmentation in Multiplanar T2W MRI Images

Multiplanar T2-weighted (T2W) magnetic resonance imaging (MRI) plays a key role in preoperative planning of surgical procedures for treating prostate cancer. It enables assessment of relevant anatomical structures such as the prostate gland, bladder and urethra before surgery, thereby preventing postoperative complications, such as urinary incontinence, which occurs when the membranous urethra (MU) is damaged. Manual measurement of MU properties such as length, volume and average radius is challenging and time consuming. To automate the task, we propose a novel semi-automated method that requires only very sparse expert manual annotations as input, and uses image preprocessing and fusion of active-contour-based segmentation techniques to produce the MU mask. We present preliminary quantitative and qualitative experimental results on clinical MRI data demonstrating the potential of our method and its capacity to produce MU measurements in agreement with clinicians.