Background: Alzheimer's disease (AD) is the most common disorder of dementia, which has not been cured after its occurrence. AD progresses indiscernible, first destroy the structure of the brain and subsequently becomes clinically evident. Therefore, the timely and correct diagnosis of these structural changes in the brain is very important and it can prevent the disease or stop its progress. Nowadays, remark to this fact that magnetic resonance imaging (MRI) provides very useful and detailed information, and due to non-invasiveness, this method has been great interest to the researchers. The aim of this study was diagnosis of AD with MRI by support vector machine model (SVM).
Methods: This is an analytical and modeling research which done in School of Public Health, Kermanshah University of Medical Science, Iran, from February 2017 to November 2017. The data used in this study was a database named Miriad containing brain MRI of 69 individuals (46 Alzheimer's disease and 23 healthy subjects) that was collected at the central hospital in London. Individuals were categorized into two groups of healthy and Alzheimer's disease with two criteria: NINCDS-ADRAD and MMSE (as the golden standard). In this paper SVM model with three linear, binomial and Gaussian kernels was used to distinguish Alzheimer`s disease from healthy individuals.
Results: Finally, SVM model with Gaussian kernel, separated AD and healthy subjects with 88.34% accuracy. The most important Areas for Alzheimer were three Area: Right para hippocampal gyrus, Left para hippocampal gyrus and Right hippocampus. The clinical result of this study is to identify the most important ROI for the diagnosis of Alzheimer's by a clinical specialist. Experts should focus on atrophy in the three Areas.
Conclusion: This study showed that the SVM model with Gaussian RBF kernel can separated AD from healthy subjects by high accuracy. Based on results of this study, can make a software to use in MRI centers for screening AD test by people over the age of 50 years.
