Automatic delineation method of head and neck lymph nodes and drainage area based on deep learning

Abstract

The embodiment of the present invention provides an automatic delineation method of head and neck lymph nodes and drainage areas based on deep learning. This method uses the symmetry of the human body structure to divide the head area into left and right parts for training and prediction, thereby indirectly increasing the number of models. The amount of training data, while reducing the size of the deep learning model. The present invention uses a delineation strategy of step-by-step processing and optimization. Firstly, the deep learning model is used to realize the easily segmented lymphatic drainage area, and then the multi-task deep learning model is used to optimize the lymphatic drainage area and segment the lymph nodes, making full use of the relationship between the two and improve the segmentation accuracy of the two. The present invention implements an artificial intelligence (AI)-assisted contouring method in the radiotherapy planning workflow, which can effectively improve the consistency of the contouring of the work efficiency of medical workers, and improve the accuracy of radiotherapy for head and neck tumors.

Description

technical field [0001] The invention relates to the field of medical images, in particular to an automatic delineation method for head and neck lymph nodes and drainage areas based on deep learning. Background technique [0002] Due to the extensive submucosal distribution of lymphatic network in the head and neck, the risk of regional lymph node metastasis of head and neck tumors is high, and the probability of occult metastasis is as high as 30%. In the clinical radiation therapy of head and neck cancer patients, in addition to high-dose radiation to the primary tumor (GTV), prophylactic radiation to the lymph nodes (GTVn) and lymphatic drainage areas (CTVn) surrounding the tumor is also required. . At present, radiotherapy is one of the important treatment methods for head and neck tumors, and there are many important and delicate structures in the head and neck. Precise radiotherapy technology has important clinical significance for the quality of life of patients after...

AI Technical Summary

Problems solved by technology

The lymphatic drainage area has good contrast and boundaries on CT, but lymph nodes are difficult to distinguish on localized CT or enhanced CT for radiotherapy, especially easily confused with adjacent muscles and blood vessels, resulting in mis-outlined or missed outlines

[0003] At present, the boundaries of metastatic lymph nodes and lymphatic drainage areas are mainly drawn manually by doctors or image segmentation methods based on deep learning, in which the efficiency of manual drawing is low, the repeatability is poor, and it is heavily dependent on the experience level of drawing doctors and drawing standard references

In addition, for different types of tumors and tumors in different locations in the head and neck, the risk of lymph node metastasis is quite different, which brings greater challenges to the clinical delineation of metastatic lymph nodes.

[0004] However, the existing conventional image segmentation methods based on deep learning rely on image labeling and training with a large amount of data, and clinically obtained data usually have problems with incomplete labeling and inconsistent labeling schemes, resulting in the actual amount of training data available. rarely

Therefore, the existing deep learning-based technology is still unable to support the automatic delineation of low-contrast and irregularly distributed lymph nodes.

In addition, the lymphatic drainage area of ​​the head and neck covers a large range of images, and its delineation depends on the image information between layers. The general hardware configuration cannot support the training of segmentation models under large-scale 3D image input.

Images