Medical image automatic segmentation method based on deep learning

Abstract

The invention relates to a medical image automatic segmentation method based on deep learning, and provides a solution for challenges such as complex imaging background, lack of shape features, intensity specificity and position priori in a tumor or organ segmentation task, so that automatic positioning and segmentation of a tumor or an organ are realized. Specifically, a segmentation strategy based on a region of interest is designed, a sparse target detection module is used to automatically position and classify tumors or organs, and mask branches are used to finely segment the region of interest. In addition, boundary segmentation is fused into mask segmentation to obtain a finer segmentation result. The invention aims at solving the limitation of a conventional semantic segmentation method in tumor or organ segmentation tasks, and solving the problem of working efficiency of radiologists to a certain extent and reducing manual wrong segmentation caused by personal deviation and clinical experience by realizing full-automatic tumor or organ segmentation.

Description

technical field [0001] The invention relates to the fields of computer vision and medical image analysis, in particular to an automatic segmentation algorithm based on medical images. Background technique [0002] Tumor or organ segmentation is an important basis in the quantitative analysis of medical images. Many follow-up tasks, including quantitative assessment of tumors or organs, tumor staging, auxiliary diagnosis, radiotherapy, etc., require accurate segmentation results. However, manual segmentation of tumors or organs requires a combination of images, clinical information, and background knowledge to accurately localize tumors or organs. Fully automatic segmentation greatly reduces the workload of radiologists to a certain extent, and reduces human errors in segmentation due to individual deviation and clinical experience. [0003] The application of automatic segmentation techniques based on deep learning can greatly facilitate the study of tumors or organs. Sinc...

AI Technical Summary

Problems solved by technology

Since medical images contain many modalities, imaging parameters, and tumor or organ sizes, it brings the following challenges to the automatic segmentation of tumors or organs: the imaging background is complex and the target lacks position prior; the target lacks shape features, and the model is difficult to adapt to it. High variability in scale and shape; target lacks intensity specificity

Images