Filter-Based Image Enhancement Effects on CNN-Based Decision Support System for Cervical Cancer Diagnosis

Le résumé fourni par la source

Cervical cancer is a type of cancer that begins in the cervix and spreads over time, typically associated with human papillomavirus (HPV) infection. This type of cancer can be prevented or treated with routine screening and early detection. The Pap smear test plays a crucial role in the early detection of cervical cancer, as regular testing increases the chances of detecting abnormalities early and starting treatment, thus preventing the disease from progressing. However, the interpretation of Pap smear images can be clinically challenging because it requires the identification of subtle differences in cell morphology and structural features, which can lead to subjective interpretations and requires evaluation by an experienced pathologist for an accurate diagnosis. This study aims to evaluate Pap smear images used in the diagnosis of cervical cancer using CNN-based decision support systems. The study examines the impact of five different filtering techniques (Blur, Bilateral, Gaussian blur, Median Blur, and Gaussian Blur) on the performance of NasNetLarge CNN model. These filters are used to reduce noise in the images and make cell structures more distinct. The performances of this model was analyzed by applying mentioned filtering techniques and evaluating their effects on diagnostic accuracy. The findings of the study reveal how filtering techniques and selected CNN model perform in the evaluation of Pap smear images. The most effective filter type was determined to be the Gaussian blur filter. The model achieved accuracy of 86.79%, sensitivity of 86.81%, specificity of 96.70%, and an F1-score of 86.78%• The results of this study demonstrate the significant role of filtering techniques and CNN model in the evaluation of Pap smear images. These findings highlight the potential of artificial intelligence and image processing techniques in the early diagnosis of cervical cancer and suggest that the use of these technologies in clinical applications can improve diagnostic accuracy.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.