A Hybrid Genetic Algorithm with a K-Means++ Clustering Model to Accurately Determine BI-RAD Scoring for Breast Mammography Image Screening

Abstract

Recent studies have revealed that breast cancer is responsible for 25.84% of all cancer-related deaths. In Africa and Ethiopia, new breast cancer cases account for 29.46% and 31.85% of the total new cancer cases, respectively. Researchers are highlighting the need to improve false positive (FP) and false negative (FN) values in the confusion matrix to develop an effective early detection model for breast cancer. Artificial intelligence (AI) plays a crucial role in improving breast cancer diagnosis, as late detection significantly reduces survival rates. In Ethiopia, 72.56% of breast cancer diagnoses are in an advanced stage (95% CI; 68.46-76.65%), highlighting the urgent need for early detection. Furthermore, the lack of radiologists contributes to delays in the annual reading and classification of BI-RAD. It is important to note that most women diagnosed with breast cancer receive negative or non-invasive results. Mammography image data is obtained from the local diagnosis center where a senior radiologist provides recommendations on the diagnosis results and the next follow-up steps for patients. A retrospective research project focuses on screening breast mammography images to identify the level of BI-RAD. A computational intelligence framework has been developed that uses OpenCV for image classification and hybridizing genetic algorithms with K-Means++ clustering. The model accurately screens BI-RAD levels and provides results for patients based on the recommendations of senior radiologists, yielding significant results in the research article.