CT cervical rib positioning method and system based on atlas registration

Abstract

The invention provides a CT cervical rib positioning method and system based on atlas registration, and relates to the technical field of image processing. The method comprises the following steps: a bone segmentation step: segmenting a bone structure from a chest CT image, and obtaining bone segmentation masks; a left lung and right lung division step: obtaining bone segmentation masks of a left lung and a right lung from the bone segmentation masks; a coarse registration step: manufacturing rib atlases of the left lung and the right lung, and then respectively registering the rib atlases of the left lung and the right lung to the bone segmentation masks of the left lung and the right lung; a fine registration step: obtaining a tibia mask of left lung bone segmentation and a tibia mask of right lung bone segmentation after coarse registration of the rib structure; a splicing step: splicing the left lung and the right lung, and carrying out logic OR operation on cervical rib masks of the left lung and right lung bone segmentation to obtain cervical rib segmentation masks of the whole lung, so as to complete the positioning of the cervical ribs. According to the invention, the workload of manual labeling can be greatly reduced, and the topological relation of a rib structure in a positioning result can be kept unchanged.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a CT cervical rib positioning method and system based on atlas registration. Background technique [0002] Computed tomography (CT) is the main method for diagnosing rib fractures and other bone diseases. The number of CT images for each patient is two to three hundred or even more. To diagnose rib fractures and other bone diseases, doctors often need to scan one by one. The CT images are inspected layer by layer, and the dynamic changes of each rib in different layers of CT images are tracked. [0003] In counting rib numbers, the cervical rib is an important marker structure. First of all, the cervical rib is the starting point of rib counting. After the position and orientation of the cervical rib are determined, the subsequent rib structure only needs to be numbered sequentially. In addition, among all the ribs, the cervical rib has the most special shape, ...

AI Technical Summary

Problems solved by technology

Such methods often rely on artificially designed feature training machine learning models such as the color or brightness difference between the rib area and the background area, which is difficult to guarantee robustness and limits its application range.

The segmentation algorithm based on supervised learning in the prior art, such as the manual labeling workload required for training data in deep learning, is relatively large, while the completely unsupervised segmentation algorithm, such as the threshold-based segmentation algorithm, has robustness and poor generalization

Images