Classification of Sodium MRI Data of Cartilage with Machine Learning and Logistic Regression

Guillaume Madelin¹, James S. Babb¹, Ravinder R. Regatte¹

Statistical learning algorithms, such as support vector machine (SVM), k-nearest neighbor (KNN), naive Bayes (NB) and discriminant analysis (DA), and logistic regression (LR), are compared for classifying subjects with and without osteoarthritis (OA) from sodium MRI data of articular cartilage at 7T. The best accuracy results are obtained with SVM and LR. SVM can classify the data with an accuracy of 78-80% by combining MRI measurements acquired with and without fluid suppression. LR generates a slightly lower accuracy (74-79%), but use only a single MRI measurement acquired with fluid suppression.