Automatic lung medical image segmentation method

Abstract

The invention belongs to the medical image segmentation technology field and discloses an automatic lung medical image segmentation method. According to the method, the pixel gray constraint and the growth distance constraint are acquired according to an initial growth seed point, and a lung lesion region is determined from the pixel gray constraint and the growth distance constraint, and the lunglesion region is smoothed to accurately acquire a segmented image of the lung lesion tissue. The method is advantaged in that the initial lung shape can be relatively excellently acquired, over-segmentation during subsequent adjustment (for example, recognizing the spine and the stomach cavity as the lung region) can be avoided, under constraints of an active shape model, the lung region on the X-ray chest can be accurately segmented to provide valuable data for clinical or computer analysis.

Description

technical field [0001] The invention belongs to the technical field of medical image segmentation, in particular to an automatic segmentation method for lung medical images. Background technique [0002] At present, lung segmentation can be roughly divided into two methods: rule-based reasoning and pixel classification. The methods used by rule-based schemes are (local) thresholding, region growing, edge detection, morphological operations, fitting geometric models and functions, dynamic programming, etc., and most lung region segmentation algorithms fall into this category. Pixel-based classification schemes attempt to classify each pixel in an image into an anatomical type (usually lungs and background, but in some cases more types are used, e.g. also heart, middle diaphragm and diaphragm). The classifiers used are various neural networks or Markov random field models, and various (local) features are used for classification, including grayscale, position, and texture me...

AI Technical Summary

Problems solved by technology

[0003] To sum up, the problems existing in the existing technology are: the existing segmentation algorithms for lung lesion tissue based on region growth all need to manually obtain the initial growth seed points, which results in inaccuracy; at the same time, the rule-based boundary detection method The flaw is that only the main lung boundary segments are given, and the complete lung outline is not given

Images