Predictive Analysis for Breast Cancer: A Machine Learning Approach

Abstract

The late stage diagnoses of breast in Nigerian women poses a significant public health concern that often results from screening delays and diagnostic inefficiencies. This research presents the design and development of a machine learning powered system for diagnoses meant to aid laboratory personnel in early breast cancer prediction. The system utilizes readily available key clinical features such as age, gender, laterality, tumor shape, nature of aspirate, and family history to classify aspirates as either malignant(cancerous) or benign (Non-cancerous). Data sourced from a pathology lab in Kano served as the training set for both the classical and deep-learning models with the deep learning model attaining better performance (F1 Score: 88.31%, Accuracy: 91.11%). Early patient-prioritization and screening are made possible by this system hence improving diagnostic turnaround times and healthcare results and healthcare outcomes especially in resource-constrained areas, the solution includes an easy-to-use interface for the smooth integration into laboratory workflows.