A novel registration‐based algorithm for prostate segmentation via the combination of SSM and CNN

Abstract

Precise determination of target is an essential procedure in prostate interventions, such as prostate biopsy, lesion detection, and targeted therapy. However, the prostate delineation may be tough in some cases due to tissue ambiguity or lack of partial anatomical boundary. In this study, we proposed a novel supervised registrationbased algorithm for precise prostate segmentation, which combine the convolutional neural network (CNN) with a statistical shape model (SSM). Methods: The proposed network mainly consists of two branches. One called SSM-Net branch was exploited to predict the shape transform matrix, shape control parameters, and shape fine-tuning vector, for the generation of the prostate boundary. Furtherly, according to the inferred boundary, a normalized distance map was calculated as the output of SSM-Net. Another branch named ResU-Net was employed to predict a probability label map from the input images at the same time. Integrating the output of these two branches, the optimal weighted sum of the distance map and the probability map was regarded as the prostate segmentation. Results: Two public datasets PROMISE12 and NCI-ISBI 2013 were utilized to evaluate the performance of the proposed algorithm. The results demonstrate that the segmentation algorithm achieved the best performance with an SSM of 9500 nodes, which obtained a dice of 0.907 and an average surface distance of 1.85 mm. Compared with other methods, our algorithm delineates the prostate region more accurately and efficiently. In addition, we verified the impact of model elasticity augmentation and the fine-tuning item on the network segmentation capability. As a result, both factors have improved the delineation accuracy, with dice increased by 10% and 7% respectively. Conclusions: Our segmentation method has the potential to be an effective and robust approach for prostate segmentation.

Simultaneous Hip Implant Segmentation and Gruen Landmarks Detection

“…Integrating segmentation and regression of the shape model parameters was utilised in other medical imaging domains, either using two parallel steps-one for predicting shape parameters and the other one for predicting the segmentation map such as in prostate segmentation in MRI image [11] or by combining the two steps in one pipeline [12] [13]. Regression of the SSM parameters and the distance map to segment the left ventricle was developed in [12], while [13] predicted shape coefficients and pose parameters to compute the coordinates of the landmark points that approximated the final segmentation.…”

“…The approaches that combine the shape priors as regularization terms in the loss function of the segmentation network are based on either using the landmarks distance [11] [25] or on the shape parameters [13] [12]. The normalized distant maps for the constructed contour from SSM parameters were combined in the segmentation as a parallel step to a network that generated the probability maps for prostate segmentation [11], or as the initialization step of the segmentation [25]. A stage-wise regression model is proposed in [13] that initially predicted the centre location of the prostate and subsequently incorporated shape parameters and rotation vector predictions.…”