Medical image synthesis and classification method based on conditional multi-discriminant generative adversarial network

Abstract

The invention discloses a medical image synthesis and classification method based on conditional multi-discrimination generative adversarial network. The method includes the following steps: 1. Segmenting the lesion area in a computed tomography (Computed Tomography, CT for short) image, and extracting interest in the lesion Region of Interest (ROIs); 2. Data preprocessing for the lesion ROIs extracted in 1. 3. Designing a Conditional Multi-Discriminate Generative Adversarial Network (CMDGAN) model architecture, using The images in step 2 are trained to obtain a generative model; 4. Use the generative model obtained in step 3 to perform synthetic data enhancement on the extracted lesion ROIs; 5. Design and train a Multiscale ResNet Network (MRNet for short) . The method proposed in the present invention can generate a high-quality synthetic medical image data set, and the classification network has a high classification accuracy rate for the test images, so that it can better provide auxiliary diagnosis for medical workers.

Description

technical field [0001] The invention relates to medical image synthesis and classification of conditional multi-discriminant generation confrontation network Background technique [0002] In recent years, relying on the powerful layered feature extraction capability of Convolutional Neural Network (CNN), the image classification effect has surpassed that of humans in many aspects. One of the main reasons is the use of large-scale labeled datasets to Train deep neural networks for computer vision tasks such as handwriting recognition. However, there is still a lot of room for improvement in some aspects, such as the classification of specific areas of medical images. One of the biggest challenges is that it is impossible to obtain a large-scale medical image dataset, and the deep neural networks trained with small-scale datasets The network will not be able to learn some features in medical images, which will make its classification effect poor. [0003] One of the biggest ...

AI Technical Summary

Problems solved by technology

[0006] In order to solve the problem that the existing network model does not perform well in medical image classification, such as the inability to obtain a large number of medical images, images without real labels, etc.

Images