Liver segmentation method based on spatial multi-scale U-net and superpixel correction

Abstract

A liver segmentation method based on spatial multi-scale U-net and superpixel correction comprises the following steps: firstly, enhancing the contrast of a liver image by using a window modulation algorithm, and suppressing noise interference through Gaussian filtering; secondly, segmenting the liver preprocessing image by using a cavity space pyramid pooling U-net model to obtain a liver preliminary segmentation result; thirdly, acquiring an over-segmentation result of the liver by utilizing a morphological expansion algorithm; and finally, correcting the liver preliminary segmentation imageby applying an FSLIC-E superpixel algorithm, and obtaining the accurate edge of the liver. According to the liver segmentation method, the respective advantages of ASPP-U-net and FSLIC_E superpixel algorithms are combined, so that a problem of relatively poor robustness of U-net for liver image segmentation is solved, and the liver segmentation method can be better applied to automatic segmentation of the liver image, so as to solve a problem of low segmentation precision of a traditional network for the blurred liver image, and realizes high segmentation precision.

Description

technical field [0001] The invention belongs to the technical field of image processing and pattern recognition, in particular to a liver segmentation method based on spatial multi-scale U-net and superpixel correction. Background technique [0002] In the initial diagnosis of liver disease, medical images have been used as the basis for doctors' diagnosis and treatment to initially judge the severity of liver disease. In order to assist doctors in diagnosing and formulating treatment plans for patients with liver diseases, it is necessary to accurately segment the liver area. At present, computer vision combined with medical imaging research has become a hotspot in the field of intelligent medical care. Liver segmentation technology based on medical imaging can obtain information such as liver size and geometric shape, thereby assisting doctors in initial diagnosis and treatment. [0003] At present, researchers have proposed a large number of liver segmentation algorithm...

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a liver segmentation method based on spatial multi-scale U-net and superpixel correction, which can be better applied to the automatic segmentation of liver images, and solve the problem of blurring caused by traditional networks. The problem of low segmentation accuracy of liver images has the characteristics of high segmentation accuracy

Images