3D Image Segmentation of MRI Prostate Based on a Pytorch Implementation of V-Net

Abstract

Prostatitis is one of the common diseases in adult males. At present, the MRI prostate volumetric medical image analysis mainly relies on the radiologist’s naked eye localization recognition diagnosis, which is easy to be misdiagnosed due to the fatigue of the doctor. Although the traditional technology uses the Support Vector Machine (SVM) to identify the image, it exists. Manual design features are incomplete and detection accuracy is not high. Recently, with the popularity and breakthrough of convolutional neural networks in the field of computer vision and medical image analysis, the author uses the MRI Prostate MR Image Segmentation 2012 combined with the Vnet method to exploit the depth of open source. Learning framework Pytorch, to achieve 3D image segmentation. The experimental results show that the Dice coefficient of the model on the training set reaches 0.9238 that is, the similarity between the image after cutting and the original image of the original label is 0.9238, and the error value is only 6.42, which is fast and automatic and accurate for intelligent target area painting, thus facilitating the doctor. Diagnosing medical imaging work is conducive to alleviating the imbalance of the proportion of doctors and patients in China.