Multi-layer nuclear magnetic image classification method and system based on reinforcement learning brain-like reading images

Abstract

The invention belongs to the technical field of image processing, particularly relates to a multi-layer nuclear magnetic image classification method and system based on reinforcement learning brain reading images, and aims to solve the problem of poor performance of an existing model in multi-layer nuclear magnetic image classification. The method comprises the following steps: splitting a multi-layer nuclear magnetic image into single layers; encoding and sensing the t-th nuclear magnetic image through the feature encoding and prediction model to obtain low-dimensional depth features and a classification result; based on the low-dimensional depth features, obtaining an action instruction xt through an action strategy generation model; if xt is not 0, performing feature coding and prediction of t + xt layer images, and action instruction generation and judgment until the action instruction is 0; and taking the hierarchical classification result of the image with the instruction of 0 asthe classification result of the multi-layer nuclear magnetic image. According to the method, pixel-by-pixel analysis is carried out on each layer of nuclear magnetic image to ensure correct hierarchical classification, and the hierarchy which truly contributes to final decision can be accurately positioned, so that a more accurate multi-sample multi-classification task is realized.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a multi-layer nuclear magnetic image classification method and system based on reinforcement learning-like brain image reading. Background technique [0002] Magnetic resonance images contain a large amount of high-level and quantifiable information that has not been fully mined. These high-dimensional feature information can be used to construct accurate sample classification methods with the help of pattern recognition methods. However, due to the low interslice resolution of magnetic resonance images, there is information loss in the scanning axial information, which seriously affects the integrity of the response space information of the tissue imaging method, which in turn affects the accuracy of the classification model based on nuclear magnetic images. [0003] The current digital image processing and modeling methods usually first analyze each layer o...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above problems in the prior art, that is, the existing models cannot effectively distinguish and combine the semantic information of each layer of the multi-layer nuclear magnetic image only through simple statistical methods, so that the performance in the multi-layer nuclear magnetic image classification is not good. Problem, the present invention provides a kind of multi-layer nuclear magnetic image classification method based on reinforcement learning class brain image reading, and this classification method comprises:

Images