A lesion detection system based on an active screening semi-supervised lesion detection network

Abstract

The invention belongs to the field of medical image processing and deep learning, and specifically relates to a lesion detection system, method and device based on an active screening semi-supervised lesion detection network, aiming to solve the problem of high training difficulty and low detection accuracy of existing lesion detection networks. The problem. This system includes: a detection module configured to obtain CT images to be detected and obtain detection results of lesions through a pre-trained lesion detection network; the lesion detection network is constructed based on the U‑Net convolutional neural network. The invention reduces the training difficulty of the lesion detection network and improves the detection accuracy.

Description

technical field [0001] The invention belongs to the field of medical image processing and deep learning, and in particular relates to a lesion detection system, method, and device based on an active screening semi-supervised lesion detection network. Background technique [0002] Although the convolutional neural network is regarded as one of the important tools in the field of modern medical image analysis, it has been successfully applied to various auxiliary medical cases many times. However, lesion detection based on CT images is a more difficult 3D object detection problem than the general 2D object detection problem. Due to the limited GPU memory, there are great challenges in directly generalizing 2D object detection methods to 3D cases. First, labeling 3D data is usually much more difficult than labeling 2D data, which may cause deep learning models to fail due to overfitting. Secondly, it is difficult for human eyes to accurately distinguish the exact positions of...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above-mentioned problems in the prior art, that is, in order to solve the problems of the existing lesion detection based on CT images, the detection network training is difficult and the detection accuracy is low due to the small number of marked data samples and the large error of artificially marked samples. In the first aspect of the invention, a lesion detection system based on an active screening semi-supervised lesion detection network is proposed, the system includes a detection module;

Images