Automatic drawing method for organs endangered by whole-body multi-part radiotherapy

Abstract

The embodiment of the invention provides an automatic drawing method for organs endangered by whole-body multi-part radiotherapy. According to the method, rigid registration is carried out on different patients based on a point cloud registration method, registration is carried out on a large number of patient's endangered organ sketching results to obtain a priori knowledge graph of the distribution probability of the endangered organs of each part, mapping of the probability distribution of the endangered organs on a to-be-processed medical image is realized through a point cloud registration method, priori probability distribution is obtained, and training of a convolutional neural network model is guided. Therefore, the convolutional neural network model can improve the precision of automatic sketching by utilizing anatomical structure associated information which is difficult to learn by the model, and avoids common wrong sketching. Compared with a traditional registration technology based on image gray, the adopted point cloud registration technology has the advantages of being higher in speed and smaller in noise influence.

Description

technical field [0001] The invention relates to the field of medical images, in particular to a method for automatically delineating organs at risk for radiotherapy of multiple parts of the whole body. Background technique [0002] Radiation therapy is one of the important treatments for malignant tumors at present. Accurate radiation therapy technology can greatly improve the survival rate of cancer patients, reduce patients' radiotherapy complications, and improve the quality of life of patients. Contour delineation in radiotherapy is the basic guarantee for precise radiotherapy. The accuracy of contour delineation directly determines the reliability of radiotherapy dose distribution. Wrong delineation may lead to serious radiotherapy accidents and endanger the lives of patients. At present, the delineation of contours in clinical radiotherapy is generally obtained by manual delineation, which has low efficiency and poor repeatability, and is heavily dependent on the exper...

AI Technical Summary

Problems solved by technology

However, the existing deep learning methods only consider the gray level of the image to be segmented, and do not consider the correlation between the anatomical structures, such as relative positional relationship, shape characteristics, etc., resulting in over-segmentation or organ interlacing in actual segmentation. This kind of segmentation error will greatly interfere with clinical radiotherapy work, and its segmentation error exceeds the normal anatomical structure relationship, and doctors or physicists are still prone to omissions in the review process

Images