A method for auxiliary diagnosis of cervical histopathology based on attention mechanism

Abstract

The invention provides an auxiliary diagnosis method for cervical histopathology based on an attention mechanism, which comprises reducing the whole glass slide image of cervical histopathology according to the proportion r to obtain a low-resolution image; building and initializing an attention module and a classification module; The rate image is input into the attention module, and the attention score is calculated; the position of the pixel with high attention score in the image is obtained, and the set size k blocks are intercepted; the small block is input into the training classification module, and the features of the small block with high score are extracted , get the feature vector, train the support vector machine classifier, input the trained support vector machine classifier, and realize the classification. The diagnostic efficiency and accuracy of the original method were improved. The lesion area in the positive cervical histopathological full-slide image is located to avoid the interference of the negative part in the positive cervical histopathological full-slide image to the classification model. Using attention mechanism to detect suspicious lesions in cervical histopathological whole slide images.

Description

technical field [0001] The invention relates to the field of automatic screening and analysis of pathological images of cervical tissue, in particular to a method for auxiliary diagnosis of cervical tissue pathology based on an attention mechanism. Background technique [0002] Cervical cancer is a common female malignancy. Early detection and early treatment are essential to reduce the cure rate and mortality of cervical cancer. Precancerous lesions of cervical cancer are also known as cervical intraepithelial neoplasia (CIN). In order to assess the severity of cervical precancerous lesions, precancerous lesions are divided into 3 grades (CIN1, CIN2, CIN3). [0003] Cervical histopathological diagnosis is the gold standard for cervical diagnosis and the final step in the diagnosis of cervical cancer. However, histopathological diagnosis relies entirely on the professional knowledge and personal experience of the pathologist for evaluation. This work is time-consuming and l...

AI Technical Summary

Problems solved by technology

[0007] The present invention provides a method for auxiliary diagnosis of cervical histopathology based on the attention mechanism, which solves the problems of low efficiency and low accuracy in the existing automatic classification method of cervical histopathological whole slide images existing in the prior art. Improvements were made on the basis of the method, and an attention mechanism-based auxiliary diagnosis method for cervical tissue pathology was provided, which reduced the workload of pathologists, improved diagnostic efficiency, and provided doctors with objective and accurate diagnostic results.

Images