A Decision Support System Framework Based on Text Mining and Decision Fusion Techniques to Classify Breast Cancer Patients

Medical decision support systems (MDSS) are designed to assist physicians in making accurate decisions. The required data by MDSS are collected from various resources such as physical examinations and electronic health records (EHR). In this paper, an MDSS framework has been proposed to diagnose and classify breast cancer patients (DSS-BC). Medical texts reports (MTR) were embedded, and essential feature vectors combined with EHR were extracted using principal component analysis (PCA). A new method based on a fuzzy min−max neural network with hyper box variable expansion coefficient (FMNN-HVEC) was used to determine the molecular subtypes, and the feature vectors were clustered using deep clustering. Also, a new decision fusion algorithm called weighted Yager was proposed based on the F1-Score for each class. This algorithm proposed a mathematical decision fusion technique to determine the Breast Imaging-Reporting and Data System (BI-RADS) and molecular subtypes values with the accuracy of 95.12% and 89.56%, respectively.