Predicting Early-Stage Cervical Cancer With An Integrated Hybrid Method Of Deep Learning And Statistical Analysis

Abstract

Early detection of cervical cancer is crucial for effective treatment and improved survival rates. However, conventional diagnostic methods often rely on manual interpretation and are limited by variability in clinical assessments. This paper proposes an integrated hybrid framework combining deep learning techniques with statistical analysis to enhance the prediction of early-stage cervical cancer. The model utilizes a convolutional neural network (CNN) to automatically extract meaningful features from cervical cell images, while key clinical and demographic parameters are analyzed using statistical methods to identify significant predictors. By merging visual and non-visual data sources, the framework improves diagnostic accuracy and supports early medical intervention. The proposed hybrid system was evaluated using a publicly available cervical cancer dataset, incorporating both image-based and tabular features such as age, number of pregnancies, hormonal contraceptive use, and Pap smear results. The fusion of CNN outputs with logistic regression and ANOVA-based statistical analysis led to a robust prediction model, achieving high performance in terms of sensitivity, specificity, precision, and F1-score. Experimental results demonstrate that the integrated approach outperforms standalone methods, offering a reliable and interpretable solution for early-stage cervical cancer prediction. This research highlights the potential of combining deep learning with traditional statistical tools to support clinicians in making informed, data-driven decisions.