Ultrasound-Based Breast Cancer Classification Using Machine Learning: An Objective Multi-Criteria Decision-Making Approach

Abstract

The diagnosis of Breast Cancer (BC) is challenging for physicians. Common diagnosis methods are subjective, negatively affecting their accuracy. We aim to develop an objective method to detect the type of BC (malignant or benign) using Machine Learning (ML). Public ultrasound images were used and preprocessed. Six texture-based and three gradient features were extracted, and Support Vector Machine (SVM), Random Forest (RF), and Decision Tree (DT) were used. Accuracy, sensitivity, specificity, False Discovery Rate (FDR), Negative Predictive Value (NPV), Matthews Correlation Coefficient (MCC), F1-score, and Area Under the receiver operating characteristic Curve (AUC) of classifiers were reported. Then the weight of each metric was calculated using Distance correlated CRiteria Importance Through Intercriteria Correlation (D-CRITIC) and the Technique of Order Preference Similarity to the Ideal Solution (TOPSIS) was used to rank the models. RF had the higher performance than others.