Coronary artery sequence blood vessel segmentation method based on space-time discriminative feature learning

Abstract

The invention relates to a coronary artery sequence blood vessel segmentation method based on space-time discriminative feature learning, which is used for carrying out blood vessel segmentation processing on a cardiac coronary artery angiography sequence image, and includes processing a current frame of image and several adjacent frames of images based on a pre-trained improved Unet network model, and obtaining blood vessel segmentation result of current frame image, wherein the improved Unet network model comprises a coding part, a jump connection layer and a decoding part, the coding part adopts a 3D convolution layer to perform time-space feature extraction, the decoding part is provided with a channel attention module, and the jump connection layer aggregates features extracted by thecoding part, thus obtaining an aggregation feature map and transmitting the aggregation feature map to the decoding part. Compared with the prior art, the cardiac coronary artery blood vessel segmentation method introduces the spatial-temporal features to perform cardiac coronary artery blood vessel segmentation, reduces the interference of time domain noise, emphasizes the blood vessel features,alleviates the problem of class imbalance in blood vessel segmentation, and has higher blood vessel segmentation accuracy.

Description

technical field [0001] The invention relates to the field of image segmentation, in particular to a coronary artery sequence blood vessel segmentation method based on spatio-temporal discriminative feature learning. Background technique [0002] According to the data of the World Health Organization, cardiovascular disease has shown a high incidence in recent years, and its high mortality rate ranks first among various malignant diseases, seriously threatening human life and health. Early screening of cardiovascular diseases is an effective means to reduce the incidence of cardiovascular diseases. Based on computer-aided diagnosis technology, it can assist doctors in fast and accurate diagnosis and treatment, greatly reduce the workload of doctors, improve the utilization efficiency of medical resources, and allow medical resources to cover more people. Blood vessel segmentation, as a basic step in computer-aided diagnosis, provides support for subsequent screening and diag...

AI Technical Summary

Problems solved by technology

[0005] First, due to the low contrast, blurred boundaries, spatially distributed noise interference, and occlusion of other tissues in coronary angiography images, a single image cannot provide sufficient information to distinguish vessel pixels from background pixels

Some of the existing literature ignores the use of time-domain information, and some simply introduce time-domain information, while ignoring the introduction of noise interference during the process of introducing time-domain information.

[0006] Second, in the process of providing context reference for the blood vessel segmentation of the current frame contrast image, we artificially introduce more time-domain information, but the time-domain information also inevitably introduces noise interference. The introduction of time-domain information On the one hand, it provides sufficient information for the reconstruction of the subsequent blood vessel segmentation map; on the other hand, it also introduces more information redundancy, which increases the computational burden of the GPU.

[0007] Third, since the number of foreground (vessel) pixels in angiography images accounts for about 5% of the total pixels, foreground segmentation will encounter serious class imbalance problems

In the class imbalance problem, it will make the network tend to judge the type of pixels with a small proportion as the type of pixels with a large proportion, so that the segmentation accuracy will decrease.

In the previous blood vessel segmentation models, most of them used cross entropy as the loss function, and did not consider the problem of class imbalance.

Images