Support vectors machine-based model for diagnosis of multiple sclerosis using the plasma levels of selenium, vitamin B12, and vitamin D3
Seyed Sajjad Sharifmousavi, Matia Sadat Borhani
Abstract
Background: It has been reported that some nutritional factors can play a significant role in immune system regulation and the etiology of immune-mediated diseases such as multiple sclerosis (MS). However, there have been no studies on the relationship between nutritional factors and MS as a primary diagnostic method. Methods: Plasma samples of 99 MS patients and 81 healthy people as the control group were assayed using atomic absorption spectroscopy and the chemical autoanalyzer method to determine the amount of selenium and vitamins (B12, D3), respectively. To find a diagnostic model, three types of supervised machine learning methods, i.e., support vector machine algorithm (SVM), decision tree (DT), and K-nearest neighbor (KNN) were used. Results: The diagnostic model based on the support vector machine method had high accuracy (98.89%), sensitivity (98.98%), and positive predictive value (98.98%). Moreover, the SVM had a true positive rate of 99.9%. Conclusion: The results of this study showed that the SVM algorithm with Radial Basis Function Kernel could be effective in obtaining a good MS diagnostic method based on the plasma levels of selenium, vitamin B12, and vitamin D3 regarding DT and KNN. This method is non-invasive and cost-effective and can be used as a primary screening method for the diagnosis of MS.