Method and device for segmenting cerebral ischemia areas in diffusion-weighted images

Abstract

The invention provides a method and device for segmenting cerebral ischemia areas in diffusion-weighted images. The method comprises the following steps of: dividing diffusion-weighted images of a plurality of super-acute ischemic stroke patients into test images and training images; training a random forest model, a learning dictionary and a support vector machine model according to the trainingimages; carrying out initial cerebral ischemia area segmentation by utilizing the trained random forest model according to grey features of voxels in the test images; determining a sparse encoding matrix of a local image block feature vector of each voxel in connected regions on the basis of the trained learning dictionary; and classifying each connected area by utilizing the trained linear support vector machine model according to a package feature of each connected area, and deleting the connected areas, in which non-ischemic tissues are located, from a first initially segmented image so asto obtain an optimal segmented image. According to the method and device, automatic recognition and segmentation of super-acute cerebral ischemia areas can be solved, and the ischemia area segmentation precision is improved.

Description

technical field The invention relates to the technical field of image processing, in particular to a method and device for segmenting cerebral ischemic regions in diffusion weighted imaging images. Background technique Cerebrovascular disease has become the number one cause of death in my country. More than 60% of cerebrovascular diseases are ischemic strokes, and the only proven effective intervention is hyperacute thrombolysis. Ischemic stroke is mainly diagnosed by magnetic resonance imaging, the most sensitive of which is diffusion-weighted imaging. Diffusion-weighted imaging includes T2-weighted images, diffusion-weighted images (DWI) and calculated apparent diffusion. Coefficient image (Apparent diffusion coefficient, ADC). The methods of cerebral ischemic region segmentation mainly include the method of cerebral ischemic region segmentation based on fuzzy mean, the method of cerebral ischemic region segmentation based on gray histogram divergence, the method of cer...

AI Technical Summary

Problems solved by technology

Due to the progressive state of cerebral ischemia in the hyperacute phase, the diffusion-weighted imaging image presents complex changes. Segmentation, because this method relies on the edge intensity of the cerebral ischemic region, it may miss the ischemic region; when the gray level distribution in the cerebral ischemic region is not uniform and the contrast with the surrounding tissue is low, the histogram-based divergence method is used The cerebral ischemic region segmentation method based on measurement and the cerebral ischemic region segmentation method based on ADC and DWI gray scale constraints cannot effectively segment this type of region

Images