MECAU-Net network-based medical image segmentation method

Abstract

The invention discloses a medical image segmentation method based on an MECAU-Net network. The method comprises the following steps: performing feature extraction on an imported medical image at a coding end of the MECAU-Net network by adopting a 2 * 2 even convolution module; constructing a 4 * 4 even-number convolution channel parallel to the 2 * 2 even-number convolution module, extracting image information by adopting the 4 * 4 even-number convolution channel, and directly transmitting the extracted information to a trunk part of a coding network to fuse feature information in different receptive fields; the method comprises the following steps: symmetrically filling a to-be-segmented feature map in each layer, transmitting obtained information to a corresponding main body network in a splicing manner for next pooling, and expanding a receptive field of an even convolution kernel while eliminating pixel offset caused by even convolution. According to the method, more abundant information can be extracted, meanwhile, extra overhead is hardly increased, and a better segmentation result is obtained by utilizing calculation overhead as little as possible.

Description

technical field [0001] The invention belongs to the technical field of image segmentation, and in particular relates to a medical image segmentation method and system based on a MECAU-Net network. Background technique [0002] Deep learning has achieved very successful applications in many scenarios such as image classification, semantic segmentation, and object detection. Medical image segmentation has received extensive attention from researchers because of its unique application scenarios. The data needs to be labeled by professionals, and it is relatively difficult to obtain data sets, which increases the difficulty of medical image segmentation. With the rapid development of deep learning, researchers have designed various networks to improve the performance of medical image segmentation. Ronneberger et al. proposed the U-Net network, which used the U-shaped network structure for the first time to segment medical images, and achieved good performance. Its unique U-shap...

AI Technical Summary

Problems solved by technology

[0004] The invention with the patent No. CN113920108A discloses a training method for training the U-Net model for processing cell images. The U-Net model can accurately segment cell images, and the contour edges after segmentation are clear, which effectively improves the accuracy of segmenting and / or counting cell images, especially mixed types of cell images, and effectively solves the problems in the prior art. The high cost of image annotation, the difficulty of cell image segmentation, and the complexity of cell counting operations

Images