3D Image Segmentation Learning Network Training Method, Segmentation Method, Segmentation Device and Medium

Abstract

The present disclosure relates to a training method, a segmentation method, a segmentation device and a medium of a 3D image segmentation learning network. The 3D image contains the object of interest and the proportion of the object of interest in it is lower than a predetermined threshold. The training method includes: building a segmentation learning network based on the sequential joint construction of multiple dense blocks. The basic units in the dense block have dense connections. The unit is composed of batch normalization layer, RELU layer and convolutional layer; the processor, based on the training data set of 3D images, uses the loss function to train the segmentation learning network. Negative voxels are penalized. It can quickly and accurately segment irregular and small objects of interest with a learning network with a more compact structure and fewer parameters, and the training process of the learning network can solve the imbalance between samples and foreground and background, and try to avoid training samples lack of overfitting problems.

Description

[0001] cross reference [0002] This application claims priority to U.S. Provisional Application No. 62 / 675,765, filed May 24, 2018, the entire contents of which are hereby incorporated by reference. technical field [0003] The present disclosure generally relates to image processing and analysis. More specifically, the present disclosure relates to a training method of a learning network for 3D image segmentation, a 3D image segmentation method and a segmentation device, and a non-transitory computer-readable medium on which a corresponding program is stored. Background technique [0004] Cancer is one of the major diseases faced by human beings, and early detection of cancer can greatly increase the survival rate. However, early-stage tumors are often irregularly shaped and represent a small proportion of corresponding medical images. Take nasopharyngeal carcinoma (NPC) as an example, which is one of the most common cancers, accounting for 0.7% of all cancers. From 199...

AI Technical Summary

Problems solved by technology

The object of interest as the foreground only occupies a small part of the entire image, which will lead to serious foreground and background imbalance problems in the training of the learning network, which will lead to training difficulties

In addition, training samples are scarce due to the tediousness of accurately delineating tumor boundaries, which often leads to the overfitting problem of the learned network

Images