Pathological feature recognition method of skin biopsy images based on deep learning

Abstract

The present invention relates to a method for identifying pathological characteristics of skin biopsy images based on deep learning, which uses a multi-layer stacked autoencoder to re-express the features of the biopsy images, and uses a series of convolutional neural networks to convolve the image features layer by layer and sampling to obtain an abstract feature expression of the original skin biopsy image; the features obtained by the multi-layer stacked autoencoder and the convolutional neural network are spliced, and finally a multi-channel neural network is used to complete the identification of pathological characteristics. The present invention extracts abstract concept expressions through a deep learning model, and has strong adaptability to factors such as image color difference, illumination, magnification, thereby greatly improving the accuracy of computer recognition of pathological characteristics of skin biopsy images.

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