Automatic segmentation method for lung-area CT (Computed Tomography) sequence

Abstract

The invention discloses an automatic segmentation method for a lung-area CT sequence. The method is characterized by comprising the steps of 1) an image I of the lung-area CT sequence is input; 2) the image I is segmented via interactive region growth; 3) according to a segmentation result, an initial contour is obtained, and seed-point coordinates of adjacent images are calculated; 4) based on the seed-point coordinates of the image I, a present image II in the sequence is segmented via the interactive region growth in step 2; and 5) step 2, 3, 4 are repeated to determine whether all images in the CT sequence are segmented, and if no, the step 3 is turned to. The automatic segmentation method automatically calculates mapping of the seed-point areas by combining the image context object characteristic continuity of the sequence to realize sequence segmentation, thereby obtaining complete three-dimensional area data of the lung, and providing basis for VOI extraction and classification of a suspected lung tubercle for a CAD system.

Description

technical field [0001] The invention relates to image processing, and relates to the field of medical image processing, in particular to a method for automatic segmentation of CT sequences of lung regions. Background technique [0002] Lung parenchyma segmentation can narrow the search area for nodule detection, shorten the running time of nodule detection algorithms and improve the detection rate and accuracy. The central task of the stage. The difficulty of this step lies in the over-segmentation or under-segmentation of the adhesion nodules such as blood vessels and pleural membrane, which will cause the problem of the connection between the left and right lungs or the gap in the lung boundary. [0003] Algorithms commonly used for lung parenchyma segmentation are based on threshold, edge detection, and region merging growth, etc.: the threshold method is simple to implement, and the single-threshold and multi-threshold methods can effectively segment images with bimodal...

AI Technical Summary

Problems solved by technology

The difficulty of this step lies in the over-segmentation or missing-segmentation of the adhesion nodules such as blood vessels and pleural membrane, which will cause the problem of the connection between the left and right lungs or the gap of the lung boundary.

[0003] Algorithms commonly used for lung parenchyma segmentation are based on threshold, edge detection, and region merging growth, etc.: the threshold method is simple to implement, and the single-threshold and multi-threshold methods can effectively segment images with bimodal or multi-peak histograms. Threshold calculation There are also calculation methods such as iterative method, maximum inter-class variance, adaptive optimal threshold, etc., but it is difficult to obtain good segmentation results for images with stable distribution, close grayscale or large overlap, and often need to be combined with other enhancements and other processing. ; The essence of edge detection is to detect the edge according to the gray-scale jump gradient between neighboring pixels, and then perform segmentation and other processing. Considering the calculation method of the gradient, gradually introduce , Gaussian operator and other local differential or gradient operators, this method is more effective for edge detection, but often due to the existence of the gradient detection threshold, the target boundary may be intermittent, the boundary is not closed, and the edge is messy. The degree jump is sensitive to gradient information, but cannot eliminate the influence of noise, and the noise immunity is poor; the growth method based on region merging is based on the similarity of pixels, and there must be a starting point and benchmark for comparison, that is, seed points and merged growth Rules, this method can merge pixels of the same nature in the same area, and then divide the image into several different areas with different properties, which is simple and efficient, and it is clear at a glance. The disadvantage is that the degree of automation of the seed points is not enough, and a certain amount of human intervention is required.

Most of the existing research is based on a single image, which cannot realize serialization and automatic processing, requires a large amount of manual interaction, and cannot meet the requirements of CAD processing large data sequence images such as HRCT

Images