Machine learning in medicine: Performance calculation of dementia prediction by support vector machines (SVM)

Abstract

Machine Learning (ML) is considered as one of the contemporary approaches in predicting, identifying, and making decisions without having human involvement. ML is quickly evolving in the medical industry ranging from diagnosis to visualization of diseases and the study of disease transmission. These algorithms were developed to identify the problems in medical image processing. Numerous studies previously attempted to apply these algorithms on MRI (Magnetic Resonance Image) data to predict AD (Alzheimer's disease) in advance. The present study aims to explore the usage of support vector machine (SVM) in the prediction of dementia and validate its performance through statistical analysis. Data is obtained from the Open Access Series of Imaging Studies (OASIS-2) longitudinal collection of 150 subjects of 373 MRI data. Results provide evidence that better performance values for dementia prediction are achieved by low gamma (1.0E-4) and high regularized (C = 100) values. The proposed approach is shown to achieve accuracy and precision of 68.75% and 64.18%. Keywords: Machine learning, OASIS, Support vector machines, Kernel, Gamma, Regularization (C)