Experimental Framework for Development of a Hybrid Convolutional Neural Network Model for Early Detection of Breast Cancer

Abstract

Breast cancer is the second-largest cause of female cancer globally, causing over 8.15 million deaths in 2016. In Nigeria, it accounts for 26% of all cancer cases, with 70% of women at advanced stage. Despite lower mortality rates and high five-year survival rates, it remains the fifth-highest disease burden globally. Factors such as inadequate knowledge, poor healthcare systems, treatment costs, and fear hinder early detection and lower survival rates. Early detection and proper treatment before surgical procedures is essential for improving patient prognosis and saving lives. Early detection of breast cancer in Nigeria remains suboptimal due to high costs, limited availability, and lack of skilled personnel. Current diagnostic models and algorithms are developed using developed datasets, which may not accurately reflect the demographic, genetic, and socio-economic characteristics of Nigerian patients. Traditional approaches rely on image-based technologies and clinical evaluations, which may not be feasible in resource-limited settings. This research aims to develop an improved model for early detection using Nigerian-specific datasets, integrating imaging data with clinical and demographic information. The model will be validated using Nigerian datasets, compared with existing methods, and developed as a user-friendly web-based platform for widespread use in low-resource settings. The proposed architecture aims to develop a robust Hybrid Convolutional Neural Network Model for early breast cancer detection using mammographic images from Nigerian patients, reducing computational costs while maintaining high accuracy. The results of the framework will be evaluated for accuracy, sensitivity, specificity, ROC-AUC and F1 score.