A Hybrid Advanced Machine Learning Stacked Model with Genetic Weighting for Accurate Breast Lesion Classification on Ultrasound Imaging

Objective This study aimed to develop and evaluate a hybrid advanced machine learning model in classification of breast lesion using ultrasound-derived features. Material and methods A retrospective study done with the compromise a total of 206 lesion dataset of breast ultrasound imaging was analyzed, with lesions classified according to the BI-RADS classification system. Several model algorisms were evaluated including conventional machine learning classifiers, including Logistic Regression, Support Vector Machine, Decision Tree, and Naïve Bayes, in addition to ensemble methods such as Random Forest and Gradient Boosting. A hybrid advanced machine learning stacked model with genetic weighting was also evaluated to enhance feature discrimination and robustness. Model performance was evaluated using multiple metrics, including accuracy, precision, recall, and F1-score, to appropriately class imbalance. Result The model achieves high performance compared with conventional and machine learning models with an overall accuracy exceeding 90% and F1-score above 85%. also demonstrated an improved classification consistency across BI-RADS categories, achieving F1-scores of 81.25% and 68.97% for BI-RADS 2 and 3, respectively, and showed enhanced robustness in intermediate and high-risk categories despite class imbalance. Conclusion the study conclude that the proposed hybrid stacked machine learning model with genetic weighting provides a reliable and robust approach for breast lesion classification using ultrasound imaging. By effectively addressing class imbalance and improving discrimination across BI-RADS categories, the model shows strong potential as a supportive decision-making tool in clinical breast imaging.