Multi-phase fusion organ segmentation method and device based on non-local attention mechanism

Abstract

The invention discloses a multi-stage fused organ segmentation method and device based on a non-local attention mechanism. According to the method, a global scale multi-phase data pair and a local scale multi-phase data pair are subjected to picture segmentation through a preset multi-phase fusion organ segmentation network model, so that a global scale segmentation result and a local scale segmentation result can be obtained; and the global scale segmentation result and the local scale segmentation result are fused to obtain a segmentation result of the organ to be segmented. According to the invention, the cross-phase non-local attention fusion mechanism module in the preset multi-phase fusion organ segmentation network model complements the features of the two phases according to the position relationship and depth relationship between the features of the two phases, and the complementation of the two phases is fully utilized to enable the features of the two phases to be effectively fused. A multi-scale segmentation framework is adopted to fuse a global scale segmentation result and a local scale segmentation result, global context information is effectively utilized, and the accuracy of organ segmentation is improved.

Description

technical field [0001] The present application relates to the technical field of multi-phase organ segmentation, in particular to a multi-phase fusion organ segmentation method and device based on a non-local attention mechanism. Background technique [0002] In the field of medical image segmentation, due to the limited capabilities of existing CT imaging techniques, it is often difficult to accurately locate the contours of organs in a single-phase CT scan. Different phases can emphasize different details of organ boundaries, therefore, referring to image data from different phases is an effective strategy to identify organ boundaries as completely as possible. For example, many guidelines clearly recommend the use of dual-phase CT contrast-enhanced imaging, including arterial phase and venous phase pancreatic scheme. The image data of the two phases are very important in clinical diagnosis, especially for cancer, the image data of the arterial phase helps to find the tum...

AI Technical Summary

Problems solved by technology

[0003] The existing multi-stage fusion segmentation method has the following two shortcomings, which lead to low accuracy of organ segmentation:

[0004] (1) The input of the existing multi-stage fusion segmentation method is either a 3D patch of a medical image or three adjacent slices. However, neither the 3D patch nor the slice can capture the powerful three-dimensional characteristics of medical images, nor can it be effective. Leverage global context information;

[0005] (2) Due to the inconsistency between the multi-phase image data, the multi-phase image data cannot be effectively fused

Images