Chest CT image lesion segmentation model training method and system

Abstract

The invention discloses a semi-supervised chest CT image lesion segmentation model training method and system, and the method comprises the steps of obtaining a sample CT image, and calculating the credibility of the sample CT image, the sample CT image comprising a labeled image and an unlabeled image; according to the credibility of the sample CT image, pairing of the CT image is carried out, an augmented CT image is generated, and the CT image is an image only containing a focus; training a deep learning network according to the augmented CT image to obtain a focus segmentation model; and optimizing the segmentation precision of the focus segmentation model based on a self-learning strategy of a teacher-student model. According to the semi-supervised chest CT image lesion segmentation model training method, lesion information is reserved, the advantages of a full-supervised loss function and a common semi-supervised loss function are integrated, a large amount of labor cost is saved, and the requirement for massive annotation data is avoided.

Description

technical field [0001] The invention relates to the technical field of lung CT image lesion segmentation, in particular to a training method and system for a chest CT image lesion segmentation model. Background technique [0002] The lesion segmentation of chest CT is of great significance for the diagnosis and treatment of new coronary pneumonia. Although the fully supervised algorithm can get a better segmentation effect, it requires a large amount of labeled data. However, the lesion labeling of chest CT requires professional radiologists, but such doctors are generally busy with work and have little spare time. In addition, due to the different sizes and shapes of lesions and the huge number of CT layers, labeling is very difficult. Therefore, there is an urgent need for a method that can train a segmentation model based on a small amount of CT data labeled with lesions. [0003] Therefore, the prior art still needs to be improved and improved. Contents of the invent...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a training method and system for a chest CT image lesion segmentation model, aiming at solving the lack of training and segmentation based on a small amount of CT data with labeled lesions in the prior art The problem with the method of the model

Images