Medical Image Segmentation Method Based on 3D Dynamic Edge Insensitivity Loss Function

Abstract

The invention belongs to the technical field of medical image processing, in particular to a medical image segmentation method based on a 3D dynamic edge insensitivity loss function. The model of the present invention adopts a dynamic edge insensitivity loss function, and the design principle of the loss function is: in each iterative training process, the farther the edge is from the wrong pixel point, the more sensitive the network is to it, and the greater the penalty weight is. This can reduce untrustworthy edge sensitivity, reduce the impact of edge uncertainty on the model from different expert-labeled data, and improve the generalization performance of the model for medical image segmentation. At the same time, the U-net architecture based on the attention mechanism is adopted. By fusing the weight distribution of the adaptive feature map of the attention module, the deviation of the network model is small, thereby reducing the impact of labeling noise on model learning and improving the generalization of the medical image segmentation model. accuracy and precision.

Description

technical field [0001] The invention belongs to the technical field of medical image processing, and in particular relates to a medical image segmentation method. Background technique [0002] The task of medical image segmentation is an important and essential work in clinical problems. With the rapid development of artificial intelligence, a large number of excellent algorithms have emerged, and the indicators of many challenge datasets have been continuously refreshed. It is currently recognized that the precise segmentation of edges is very important to improve the accuracy of segmentation, so there have been a lot of work to focus on the edge segmentation of medical images, and adding additional network branches to learn boundary features is one of the mainstream methods. There are also some works that add attention to edges in the loss function to achieve better segmentation results. [0003] The premise of these algorithms is that the ground-truth annotated edge segm...

AI Technical Summary

Problems solved by technology

When more attention is paid to the edge features of the region of interest in the limited training samples, although the indicators on the limited training samples can be continuously improved, the risk of over-fitting of the model is also getting higher and higher, which will inevitably affect the general performance of the segmentation model. Chemical

Images