Method, system and electronic device for cerebrovascular segmentation

Abstract

The present application relates to a cerebrovascular segmentation method, system and electronic equipment. The method includes: performing multi-scale filter enhancement processing on the original blood vessel image data to obtain the enhanced blood vessel image data and the corresponding direction vector field; establishing a finite mixture model, and estimating the relevant parameters of the finite mixture model to obtain the class conditional probability P(y|x); calculate the initial label field of the blood vessel image data, and combine the initial label field and the corresponding direction vector field to form a new Markov random field; according to the Markov random field and Gibbs distribution The equivalence of the class prior probability P(x), based on the class conditional probability P(y|x) and the class prior probability P(x), through the maximum posterior probability and conditional iteration mode to obtain the blood vessel image data Vessel segmentation results. The present application can extract a more effective vessel candidate space, extract the vessel structure under low contrast, and make the obtained vessel network more complete.

Description

technical field [0001] The present application belongs to the technical field of medical image processing, and in particular relates to a method, system and electronic equipment for cerebrovascular segmentation. Background technique [0002] Up to now, a large number of cerebrovascular segmentation methods have been proposed, and some methods have been successfully applied to clinical diagnosis. Existing cerebrovascular segmentation methods are divided into two categories: skeleton-based and non-skeleton. The former segments and reconstructs vessels by first detecting the vessel centerline, while the latter directly extracts 3D vessels. More reports are based on non-skeleton segmentation methods, including: scale analysis method, deformation model method, tracking method, statistical model method and mixed model method. [0003] The blood vessel segmentation problem is regarded as a labeling problem, and all points in the angiographic data are divided into two categories, ...

AI Technical Summary

Problems solved by technology

[0007] However, the existing cerebrovascular segmentation methods still have the following defects: 1), the current method has low robustness, and the extracted blood vessels will have more fractures when the contrast is low; 2), the vascular network extracted by the current method is not complete , thus limiting the scope of clinical diagnosis and treatment

Images