Multi-view collaborative segmentation for prostate MRI images

Abstract

Prostate delineation from MRI images is a prolonged challenging issue partially due to appearance variations across patients and disease progression. To address these challenges, our proposed collaborative method takes into account the computed multiple label-relevance maps as multiple views for learning the optimal boundary delineation. In our method, we firstly extracted multiple label-relevance maps to represent the affinities between each unlabeled pixel to the pre-defined labels to avoid the selection of handcrafted features. Then these maps were incorporated in a collaborative clustering to learn the adaptive weights for an optimal segmentation which overcomes the seeds selection sensitivity problems. The segmentation results were evaluated over 22 prostate MRI patient studies with respect to dice similarity coefficient (DSC), absolute relative volume difference (ARVD) and average symmetric surface distance (ASSD) (mm). The results and t-Test demonstrated that the proposed method improved the segmentation accuracy and robustness and the improvement was statistically significant.

A propagation-DNN: Deep combination learning of multi-level features for MR prostate segmentation

“…Computer vision with object segmentation [21][22][23] and saliency detection [24][25][26][27]) has progressed markedly in recent years and so many investigators applied transfer learning from general images to biomedical images for segmentation. Graph cut based methods [32][33][34][35][36] use proper graph to construct the graph-cut cost function. Various cues are employed to comprehensively measure the edge weights on the graphs, including spatial information [28], pixel intensity [28,29], Gabor filtered feature [30], prior shape knowledge [31] and image gradient [32].…”

“…Various cues are employed to comprehensively measure the edge weights on the graphs, including spatial information [28], pixel intensity [28,29], Gabor filtered feature [30], prior shape knowledge [31] and image gradient [32]. Zeng et al [33] proposed multi-kernels to collaboratively cluster biomedical image data to aid the segmentation task. The fixed parameters for balancing the cost function, however, need careful tuning, which limits the robustness of these methods across different images.…”