A novel feature extraction approach with VBM 3D ROI masks on MRI

Abstract

Alzheimer's disease is a neurological disorder that usually starts with aging. Alzheimer's disease is a serious health and economic burden on governments, along with an increase in elderly population in developed and developing countries. There is no known cause of this disease and there is no treatment. For this reason, early diagnosis of the disease, socioeconomic and psychological outputs and medical treatments are still a hot topic investigated in the world. Magnetic Resonance Imaging is one of the medical imaging techniques that show the progression of Alzhiemer in brain. Brain deterioration and volume loss of the disease first begins with memory regions and then spreads to other brain regions. If atrophy is observed and detected by manual methods, it may not be seen due to user dependency, operator error and inexperience. For these reasons, automatic, numerical and atlas-based methods are being developed for the observation and capture of neurological diseases. In this study, 99 Alzheimer patients and 99 normal control MR images were analyzed using Voxel Based Morphometry, one of the numerical methods of atrophy observations in Magnetic Resonance Imaging. Losses in the brain were then produced as three-dimensional binary masks. Using these masks, normalized segmented, modulated normalized segmented, and normalized images that were stripped from the non-brain structures were masked. Histogram based first order statistical features were extracted in the masked areas. The efficany of this technique was statistically compared between Alzheimer's and normal control. MR images have been downloaded freely from the OASIS database.