A CNN-Based Framework for Bladder Wall Segmentation Using MRI

Accurate segmentation of the bladder wall is of great importance for any computer-aided diagnostic system for bladder cancer (BC) detection and diagnosis. In this paper, a deep learning-based framework is developed for accurate segmentation of the bladder wall using T2-weighted magnetic resonance imaging (T2W-MRI). Our framework utilizes 3D convolution neural network (CNN) and incorporates contextual information at the vicinity of each voxel to enhance the segmentation performance. The CNN soft output is refined using a fully connected conditional random field (CRF) to remove noisy and scattered predictions. Our pipeline has been tested and evaluated using a leave-one-subject-out (LOSO) on MRI data sets that were collected from BC patients. Our framework achieved accurate segmentation results for both the inner and outer bladder walls as documented by various metrics: Dice coefficient (DSC) and Hausdorff distance (HD). Moreover, comparative segmentation results using other segmentation approaches documented the superiority of our framework to provide accurate results for bladder wall segmentation.