Subject Based Registration for Individualized Analysis of Diffusion Tensor MRI

• Published in: PloS one Vol. 10; no. 11; p. e0142288
• Main Authors: Suri, Asif K., Fleysher, Roman, Lipton, Michael L.
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 18.11.2015; Public Library of Science (PLoS)
• Subjects: Abnormalities; Accuracy; Adolescent; Adult; Aged; Algorithms; Alzheimer's disease; Alzheimers disease; Brain; Brain - diagnostic imaging; Brain - physiopathology; Brain Injuries - cerebrospinal fluid; Brain Injuries - diagnostic imaging; Brain Injuries - physiopathology; Care and treatment; Central Nervous System - diagnostic imaging; Central Nervous System - physiopathology; Cerebrospinal fluid; Diagnosis; Diffusion; Diffusion Magnetic Resonance Imaging - methods; Diffusion Tensor Imaging; Error analysis; Evaluation; Female; Humans; Magnetic resonance imaging; Male; Medical imaging; Medicine; Methods; Middle Aged; Nervous system diseases; Neuroimaging; Neurosciences; NMR; Nuclear magnetic resonance; Pathology; Patient outcomes; Radiography; Registration; Tomography; Traumatic brain injury; United States
• DOI: 10.1371/journal.pone.0142288
• ISSN: 1932-6203

Registration of subject and control brains to a common anatomical space or template is the basis for quantitatively delineating regions of abnormality in an individual brain. Normally, a brain atlas is chosen as the template. Limitations in the registration process result in persistent differences between individual subject brains and template, which can be a source of error in an analysis. We propose a new approach to the registration process where the subject of interest is the registration template. Through this change, we eliminate errors due to differences between a brain template and a subject's brain. We applied this method to the analysis of FA values derived from DTI data of 20 individual mTBI patients as compared to 48 healthy controls. Subject-centered analysis resulted in identification of significantly fewer regions of abnormally low FA compared to two separate atlas-centered analyses, with subject-centered abnormalities essentially representing the common subset of abnormal low FA regions detected by the two atlas-centered methods. Whereas each atlas-centered approach demonstrated abnormalities in nearly every subject (19/20 and 20/20), the subject-centered approach demonstrated abnormalities in fewer than half the subjects (9/20). This reduction of diffusion abnormalities observed using the subject-centered approach is due to elimination of misregistration errors that occur when registering the subject of interest to a template. Evaluation of atlas-centered analyses demonstrated that 9.8% to 13.3% of subject GM and CSF was misregistered onto the WM of the brain atlas, resulting in the observation of additional low FA clusters compared to the subject-centered approach. Without careful evaluation, these misregistrations could be misinterpreted as pathology. An additional benefit of the subject-centered approach is that diffusion abnormalities can now be visualized directly in the subject's anatomical space, rather than interpolating results from the brain atlas space, and can thereby enhance correlation with other components of an imaging protocol.