A Machine Learning Approach to Assess Changes in Scoliosis

This paper presents a machine learning approach that can be used to evaluate the validity of the results obtained with an automated system to measure changes in scoliotic curves. The automated system was used to measure the inclinations of 141 vertebral endplates in spine radiographs of patients with scoliosis. The resulting dataset was divided into training and test set. The training set was used to configure three classifiers: a support vector classifier (SVC), a decision tree classifier (DT) and a logistic regression classifier (LR). Their performance was evaluated on the test set. The SVC had an accuracy of 86% discriminating Good Results (those in which the error was less than 3°) from Bad Results. This accuracy was better than that of the LR (76%) and DT (68%). The differentiation between Good and Bad Results using the proposed machine learning approach was achieved successfully.