Deep learning-based skin biopsy image pathological characteristic recognition method

Abstract

The invention relates to a deep learning-based skin biopsy image pathological characteristic recognition method. The method includes the following steps that: a plurality of layers of stacked automatic encoders are utilized to re-express the characteristics of a biopsy image; a series of convolutional neural networks are utilized to perform convolution and sampling on the characteristics of the image layer by layer, so that an abstract characteristic expression of the original skin biopsy image can be obtained; the characteristics obtained by the plurality of layers of stacked automatic encoders and the characteristics obtained by the convolutional neural networks are spliced together; and the recognition of pathological characteristics is completed by a multi-channel neural network. According to the deep learning-based skin biopsy image pathological characteristic recognition method of the invention, abstract concept expression is extracted through a deep learning model, and therefore, the method has high adaptability to factors such as color difference, illumination and magnification factors, and therefore, the accuracy of a computer in the recognition of pathological characteristics of skin biopsy images can be greatly improved.

Description

technical field [0001] The present invention relates to a method in the technical field of image processing, in particular to a method for identifying pathological characteristics of skin biopsy images based on deep learning. Background technique [0002] With the wide application of information technology in various medical disciplines, the acquisition and processing of digitally stored medical images has become easier and easier, and more and more digital medical images have been generated rapidly. The characteristics of these images can be summarized as large data volume. , high resolution, large amount of information contained in it, fast growth rate, unstructured and its characteristics cannot be easily identified. Using the professional knowledge of various medical experts to extract the information of digital medical images requires a lot of labor costs. At the same time, the quality of the extracted information is affected by the subjective factors of experts, and it...

AI Technical Summary

Problems solved by technology

In recent years, in the computer-aided analysis of biopsy images, it has been found that the pathological characteristics reflected by biopsy images are a complex and abstract concept, and the features obtained by traditional computer graphics feature extraction methods are difficult to express these concepts. The effect of the recognition model built on the basis will also be affected

This method adopts the method of establishing an image feature dictionary based on histogram, which belongs to a shallow statistical feature, and it is difficult to express complex abstract concepts. Therefore, there will be large errors in the analysis of complex biopsy images, which will affect the effectiveness of the method.

Images