Diagnostic Application of Automated Brain MRI Volumetry in Comparing Alzheimer Disease, Behavioral Variant Frontotemporal Dementia and Traumatic Brain Injury
Somayeh Meysami1, Cyrus Raji2, Verna Porter1, David Merrill1, Mario Mendez3
1Pacific Neuroscience Institute, Providence Saint Johns Health Center, 2Washington University in St Louis, 3UCLA and VA Greater Los Angeles Healthcare System
Objective:
To evaluate if application of automated brain MRI volumetry can improve delineation of neurocognitive disorders.
Background:
Assessment of regional structural differences with brain MRI volumetric quantification can improve diagnostic separation of multiple neurodegenerative and non-neurodegenerative disorders.
Design/Methods:
We utilized 136 participants with traumatic brain injury (TBI; n=40), early (EOAD; n=45) and late onset Alzheimer disease (LOAD; n=31) and behavioral variant frontotemporal dementia (bvFTD; n=20). Participants had brain MRI including 3D T1 and analyzed with an FDA cleared software program, Neuroreader. At the time of analysis, Mini-Mental State Exam (MMSE) scores were available on majority of participants (n=125). One-way ANOVA compared mean age, MMSE, and brain volumes across all four groups with Bonferroni correction for multiple comparisons. Volumes were adjusted for total intracranial volume (TIV). Discriminant analysis was done with leave-one-out cross validation on measured brain volumes to determine diagnostic delineation. Automated linear regression identified predictive features.
Results:
LOAD was the oldest compared to other groups (p<.001). There were no statistically significant differences in sex (p=.58) with 54.4% women. EOAD and LOAD had the lowest MMSE scores compared to TBI (p=.04 for EOAD and p=.01 for LOAD). Regional volume difference across groups showed lowest hippocampal volumes in LOAD (p=.005), low white matter volume in TBI (p=.007), lower parietal lobe volumes in EOAD (p<.01), and lower frontal lobe volumes in bvFTD (p<.001). Regional volume differences resulted in a correct classification of 89%, 79.4% after cross validation. Predictive features included caudate, amygdala, frontal, parietal, temporal lobar and total white matter volumes.
Conclusions:
Lower parietal lobe volumes in EOAD and hippocampal volumes in LOAD as well as frontal atrophy of bvFTD reflect known anatomical distribution of pathology in those conditions. White matter volume loss in TBI has been noted with that condition. Brain MRI volumetry may be beneficial in clinical practice.