Classification of Cervical Biopsy Images Based on LASSO and EL-SVM

Abstract

Cervical biopsy (biopsy) is an important part of the diagnosis of cervical cancer. The artificial classification of biopsy images in diagnosis is difficult and depends on the clinical experience of pathologists. However, the classification accuracy of computerized biopsy tissue images with similar lesions is low, and the problem of incomplete experimental objects needs to be addressed. This paper proposes a method of cervical biopsy tissue image classification based on least absolute shrinkage and selection operator (LASSO) and ensemble learning-support vector machine (EL-SVM). Using the LASSO algorithm for feature selection, the average optimization time was reduced by 35.87 seconds while ensuring the accuracy of the classification, and then serial fusion was performed. The EL-SVM classifier was used to identify and classify 468 biopsy tissue images, and the receiver operating characteristic (ROC) curve and error curve were used to evaluate the generalization ability of the classifier. Experiments show that the normal-cervical cancer classification accuracy reached 99.64%, the normal-low-grade squamous intraepithelial lesion (LSIL) classification accuracy was 84.25%, the normal-high-grade squamous intraepithelial lesion (HSIL) classification accuracy was 87.40%, the LSIL-HSIL classification accuracy was 76.34%, the LSIL-cervical cancer classification accuracy was 91.88%, and the HSIL-cervical cancer classification accuracy was 81.54%.