Cervical cell image identification method based on joint feature PCANet (Principal Component Analysis Net)

Abstract

The invention provides a cervical cell image identification method based on a joint feature PCANet (Principal Component Analysis Net). The method comprises the steps of S100, denoising cervical cell images; S200, establishing the joint feature PCANet for the denoised cell images to extract image features, employing the joint feature PCANet with a structure of two PCA layers, inputting the output of a first PCA layer and the output of a second PCA layer into a binary hash and histogram processing layer and jointing the obtained features as final output image features; and S300, identifying the cervical cell images. According to the method provided by the invention, the cell particular features are extracted through an artificially designed feature operator; three kinds of cells: normal, lesion and canceration cells can be effectively identified; the method is applicable to identification of other cell images; and the generalization ability of the method is high.

Description

technical field [0001] The invention belongs to the field of medical cell image processing, and relates to auxiliary recognition of cell images, in particular to a cervical cell image recognition method based on joint feature PCANet (Principal Component Analysis Net, PCANet). Background technique [0002] Early diagnosis of cervical cancer is of great significance for timely detection of cervical precancerous lesions, eradicating cancer in its infancy and saving the lives of female patients. It has been clinically proven that cervical cancer screening every five years can reduce the mortality rate of cervical cancer by 80%, and once every three years can reduce the mortality rate by 90%. Among the screening methods for cervical cancer, cervical smear screening is considered to be one of the most effective means of preventing cervical cancer, but it requires a lot of manual reading, which is affected by factors such as fatigue and experience, making early screening of cervica...

AI Technical Summary

Problems solved by technology

Among the screening methods for cervical cancer, cervical smear screening is considered to be one of the most effective means of preventing cervical cancer, but it requires a lot of manual reading, which is affected by factors such as fatigue and experience, making early screening of cervical cancer difficult. The accuracy of the accuracy is greatly limited, so the intelligent computer-aided interpretation of cervical cells has become increasingly important

[0003] At present, most of the cervical cell computer-aided identification is based on the specific features of cervical cells extracted by artificially designed feature operators. At the same time, preprocessing such as image segmentation is required before feature extraction. Due to the subjective influence of people, sometimes artificially designed feature operators The extracted features are not sufficient, which affects the recognition effect and the recognition accuracy is low, and the object of this type of method is highly targeted, making its generalization ability weak

Images