An automatic delineation method for drainage area and metastatic lymph nodes in nasopharyngeal carcinoma of the head and neck

Abstract

The invention discloses a method for automatically delineating the drainage area and metastatic lymph nodes of head and neck nasopharyngeal carcinoma, belonging to the field of medical image processing, which includes the following steps: S1: collecting and processing case data, wherein the case data includes DICOM image; S2: Establish and train the partition model of the lymphatic drainage area; S3: According to the segmentation network of the lymphatic drainage area obtained in step S2, predict each partition of the drainage area of ​​all DICOM data, and obtain the left and right sub-regions of each partition after post-processing Segmentation; S4: Establish and train the automatic lymph node segmentation network model; S5: Input the case data into the lymphatic drainage area segmentation model and the automatic lymph node segmentation network model in turn, and obtain the drainage area and lymph node segmentation results. The invention can reduce the burden of the doctor's target region delineation during radiotherapy, improve the efficiency of the doctor's delineation, and reduce the subjectivity of the doctor's delineation.

Description

technical field [0001] The invention belongs to the field of medical image processing, and in particular relates to an automatic delineation method for head and neck nasopharyngeal carcinoma drainage area and metastatic lymph nodes based on deep learning. Background technique [0002] The lymphatic network is widely distributed in the nasopharyngeal region of the human head and neck, and nasopharyngeal carcinoma has a high rate of lymph node metastasis. Radiation therapy, as one of the most important treatment methods for nasopharyngeal carcinoma, requires doctors to accurately outline the gross tumor volume (GTV for short) and the corresponding subclinical target volume (Clinical target volume) on CT images. , referred to as CTV). Among them, GTV refers to the gross tumor area seen in imaging examination or clinical physical examination, specifically including the primary tumor (GTVp) and metastatic lymph nodes (GTVn). CTV includes subclinical lesions with a certain proba...

AI Technical Summary

Problems solved by technology

[0004] When doctors manually delineate metastatic lymph nodes and neck drainage areas, on the one hand, they only rely on the CT images of the case, lacking reference to the delineation results of similar cases, which may lead to subjective delineation results, and there may be large differences in the target areas delineated by different doctors. Differences, which have adverse effects on the determination of subsequent treatment plans and the evaluation of treatment effects

On the other hand, it takes a long time for the doctor to draw manually, and the drawing time of a case can be as long as 1-2 hours, which will bring a heavy burden to the doctor

[0005] Due to the low contrast between lymph nodes and the background area and the blurred boundaries, most traditional segmentation methods are only suitable for certain scenes and cannot be used for various clinical images.

The neural network-based segmentation method usually only uses the information of CT images, lacks the guidance of the neck drainage area information, and because the volume ratio of the lymph node to the background area is only 1 / 100 or even smaller, the general neural network-based The accuracy of the segmentation method is not high enough, and it contains more false positive regions

[0006] In summary, the current automatic delineation algorithm for lymph nodes and drainage areas generally has problems such as low accuracy, inaccurate borders, and many false positive areas. The main reasons are: (1) In CT images, the difference between lymph nodes and surrounding normal tissue structures The contrast is low and the boundaries are blurred. (2) The volume and shape of different lymph nodes vary greatly. (3) The distribution and deformation of lymph nodes vary greatly among different cases.

Images