Diffusion Tensor MRI Application In Alzheimer's Disease Analysis

This thesis analyzed the diffusion tensor magnetic resonance imaging (DT-MRI)technique and presented a linear recognition method to recognize the brain's DT-MRIof Alzheimer's disease (AD) for computer aided diagnosis.Diffusion weighted images (DWIs) and T₂ weighted images (T₂WIs) wereobtained with a magnetic resonance scanner from two groups: a group of probableAD patients and a matched group of controls. Median filtering and wave-letdecomposing was performed in all images to increase signal noise ratio. DWIs wereregistered to corresponding T2WIs based on maximization of mutual information toget rid of head motion artifact before calculating diffusion tensor D. Mean diffusion(MD) and fractional anisotropy (FA) were educed from the D of every pixel. 5 fieldsof view (FOVs) of genu of corpus callosum (GCC), splenium of corpus callosum(SCC), hippocampus (HI), subcortical white matter in temporal lobe (SWMT) andsubcortical white matter in frontal lobe (SWMF) were segmented manually. Themean values of the MD and the FA of these 5 FOVs were calculated respectively.Significant differences were found in 6 parameters between the AD group andthe control group with t-test statistical method. These 6 parameters were the meanvalue of the MD of GCC and the mean values of the FA of all 5 FOVs. A recognizingmodel was trained with these 6 parameters. The cross validation method withleave-one-out was used to estimate the recognition rate, and the accuracy was 94.44%.Furthermore, this model recognized correctly other 4 subjects as a test group.The research results prove that the MD and FA values of some FOVs in ADpatient's brain are changed significantly and that DT-MRI can reflect theabnormalities of AD patients' brain. The recognition method can recognize ADthrough DT-MRI of AD characters and it may assist doctors to diagnose AD inclinical with further developing.