Premium
P1‐402: AUTOMATIC COMPUTATION ON NEUROIMAGING AND TISSUE‐BASED BIOMARKERS FOR PREDICTING ALZHEIMER'S DISEASE
Author(s) -
Sandhya P.
Publication year - 2018
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2018.06.411
Subject(s) - voxel , artificial intelligence , discriminative model , pattern recognition (psychology) , neuroimaging , positron emission tomography , computer science , grey matter , voxel based morphometry , magnetic resonance imaging , nuclear medicine , medicine , neuroscience , psychology , radiology , white matter
Background:The study aim to maximally utilize multimodal neuroimaging and tissue based biomarker model for predicting Neurodegenerative pattern that characterizes the Alzheimer’s disease. A novel brain tissue-based method of feature selection is proposed for detecting the Alzheimer’s disease (AD) in positron emission tomography (PET). Methods: The transformation of MRI into the L*a*b* color space to have a good effect on image contrast. By statistical measurements, a subsequent multivariable threshold segmentation is performed to find a measurable detection region of grey matter (GM) voxel intensities in the L* histograms (DRGMVI-L*), where the scale model of MRI in the L* histograms, which provides an interface between the boundary GM volume and the group phenomena of interest quantitatively for detecting AD. These areas are further used as a mask to select and weight the most discriminative voxels to construct a classification model. We then describe a method to fuse the features computed from different image modalities based on the weights assigned by the individual Support Vector Classifiers during the training process. The method presented here has been applied to multimodal image containing both functional (18F-FDG PET) and structural (Magnetic Resonance) data. Results:Results reveals that on average, the group of normal control (NC) exhibits a larger volume of GM voxel significantly than those two subject groups with AD (the group of AD with CDR 1⁄4 2.5 and the group of mild cognitive impairment (MCI) with CDR 1⁄4 0.5 or 1), with likelihood ratios of NC/MCI/AD at 1:0.9241:0.8539 and 1:0.9716:0.8930 in the anatomical transverse and sagittal sections, respectively. Group of NC appears to have a signaling pulse of GM phenomenally, as compared with AD andMCI groups with approximately uniform across concave in the DRGMVI-L*; this proves to be unique biomarkers of disease in differentiating subjects with AD-dementia from the NC group, and may further develop to assist in identifying difficult cases of MCI. Conclusions:Upon the crossvalidation, the results prove that the proposed automatic computation technique improves the performance with a hit rate of 0.826 as compared with existing results in the detection of AD in MRI.