HER2 image classification method and system based on convolution and residual networks

Abstract

The invention discloses an HER2 image classification method and system based on convolution and a residual network. The method comprises the steps: obtaining a plurality of HER2 images with annotations from a tissue microarray database, and taking the HER2 images with annotations as an initial data sample; respectively cutting each initial data sample by adopting a cutting function, and establishing an extended data sample set; training the convolution and residual error network by using the extended data sample set to obtain a trained convolution and residual error network as an HER2 image classification model; and inputting the HER2 image to be classified into the HER2 image classification model to obtain a classification result of the HER2 image to be classified. According to the HER2 image classification method, more HER2 images are obtained by utilizing the clipping function, so that the technical defect that the data volume of the HER2 images cannot meet the training requirementis overcome. The HER2 image classification is realized by utilizing the convolution and residual network, the over-fitting phenomenon of an existing neural network model is avoided, and the HER2 imageclassification method is realized under the condition that the data volume is relatively small. The HER2 images are automatically and accurately classified by using the neural network model.

Description

technical field [0001] The invention relates to the technical field of medical image classification, in particular to a method and system for HER2 image classification based on convolution and residual network. Background technique [0002] With the continuous development of social life, people have higher and higher requirements for classification of various images. With the development of high-resolution digital scanners, Whole Slide Imaging (WSI) is widely used, and WSI images can be processed by computer software for extensive analysis of complex cell and protein features. The Human Epidermal Growth Factor Receptor-2 (HER2) gene is an important gene, and its correct classification is a difficult task. Typically, breast tissue samples are assigned different HER2 scores depending on the degree and proportion of cell membrane staining. In the past, its evaluation was entirely based on manual observation. Not only was the work intensive, but the accuracy also largely relie...

AI Technical Summary

Problems solved by technology

For medical images, the amount of data is too small to satisfy the training of deep neural networks

Moreover, for the general convolutional neural network, as the number of network layers increases, the overfitting phenomenon is often very serious, and it is impossible to obtain a deep neural network model for medical image classification.

Images