Machine Learning-based Automated Diagnosis of Breast Cancer from Mammography Images

Abstract

This paper presents a novel machine learning-based system for the automated diagnosis of breast cancer using mammography images. The proposed system employs a nature-inspired feature extraction algorithm to accurately identify and highlight the salient features within the mammographic scans. The features derived are highly representative and effective in distinguishing between malignant and benign cases, thus addressing the inherent complexity and variability within the breast tissue structures. To enhance the prediction accuracy, a hybrid decision tree and gradient boosting algorithm is introduced. The decision tree algorithm provides a transparent and interpretable model, facilitating easy understanding and justification of the diagnosis decisions. The gradient boosting algorithm further refines the model by iteratively correcting the errors of the decision tree model, leading to a substantial improvement in diagnostic performance. The proposed system was tested on a comprehensive dataset and compared with the existing state-of-the-art diagnostic tools. The results demonstrated significant improvements in terms of accuracy, sensitivity, and specificity, thus showing promise in aiding radiologists in making more accurate and confident diagnoses. This research paves the way for a more robust, reliable, and automated system in breast cancer detection, thereby enhancing the effectiveness of breast cancer screening and early detection strategies.