Esophageal tumor region segmentation and model training method, device and electronic equipment

Abstract

The invention provides an esophagus tumor region segmentation and model training method, device and electronic equipment. The esophagus tumor region segmentation model training method includes: acquiring a computed tomography sequence of the esophagus, and performing vertebral body detection according to the acquired computed tomography sequence, so that the The computed tomography sequence includes a two-dimensional image slice subsequence and a two-dimensional image slice labeling subsequence obtained by labeling the two-dimensional image slice subsequence; the center of the vertebral body is used as the cropping seed point, and the two-dimensional image slice subsequence is obtained from the two-dimensional image slice subsequence. Cut out a 3D image block containing the esophagus region, and cut out a 3D annotated image block from the corresponding position in the 2D image slice labeling subsequence; take the 3D image block as the input of the initial esophagus tumor region segmentation model, combine the 3D annotated image block The initial esophagus tumor region segmentation model is trained to obtain the esophagus tumor region segmentation model. It can improve the segmentation efficiency of esophageal tumor region.

Description

technical field [0001] The present invention relates to the technical field of tumors, in particular to a method, device and electronic equipment for esophageal tumor region segmentation and model training. Background technique [0002] Esophageal cancer is one of the major culprits in cancer mortality and burden worldwide, and due to its high mortality rate, it has become an urgent public health problem. Radiotherapy (RT, Radiotherapy) is a relatively effective treatment for esophageal cancer, and its most basic and critical step is to determine the gross tumor volume (GTV, Gross Tumor Volume). However, due to the fuzzy boundary of the tumor region and the low contrast with the surrounding normal esophagus (there is no clear boundary between the tumor region and the surrounding tissue), the traditional image segmentation method is not suitable for esophageal tumor region segmentation. [0003] Fully Convolutional Networks (FCN, Fully Convolutional Networks), as a deep lear...

AI Technical Summary

Problems solved by technology

However, in this method of tumor region segmentation using a fully convolutional network, it is necessary to construct a coarse segmentation network model and a fine segmentation network model to achieve better segmentation performance, which makes the segmentation process complicated and the network parameters of the model are large. The coarse segmentation network model and the fine segmentation network model constructed by the decoder structure and the skip-connected U-Net network, the parameter amount of each network model reaches more than 60M, which not only makes the time required for training the network model longer, but the training efficiency is low , which also makes the trained coarse segmentation network model and fine segmentation network model take a long time to segment the tumor area, and the segmentation efficiency is low

Images