Manifold Cervical Abnormality Categorization using Deep Learned Features and Support Vector Machine

Cervical cancer stands as a prominent contributor to female cancer mortality on a global scale, ranking high among the leading causes of death from cancer in women. Persistent HPV infection is the cause of cervical cancer, which continues to be a global health concern. The purpose of this research was to identify and comprehend different forms of cervical abnormalities and their consequences to develop the CAD diagnosis sytem. We used deep learning models (Squeezenet) combined with the Orange Data Mining Tool to extract features at the image level from 966 cluster cell images of Pap smear slides from the SIPaKMeD database. We employed SVM with different kernel functions and k-NN with different parameters for cervical cancers classification. Resluths shows that tha SVM with RBF kernel function achieved highest classification accuracy of 87% in identifying the cervical abnormalities. High value of 96% for MCC clearly indicates that the features extracted using squeezenet features has the ability to categorize abnormality. The k-NN with different parameters shows the better performance in identifying the anomalies with an accuracy of 82.7%. The SVM with RBF kernel is found to show comparatively high peformance in detecting the abnormalities in the images with recall and AUC values of 87% and 97.8%.