Tools for the delineation of regional MR diffusion abnormalities
in individual patients: Where have we been? Were should we be
going?
Tools for the delineation of regional MR diffusion abnormalities
in individual patients:Where have we been?Where should we be
going?Michael L Lipton, MD, PhD, FACR
Two aims:Take stock of the dMRI literature on TBI.Make a case
for patient specific identification of dMRI abnormalities.Aim 1:
Where have we been?The overwhelming consensus of these studies is
that low white matter FA is characteristic of TBI.
Should we expect this convergence?Many studies Many
variations
>115 StudiesMild-Severe TBIAcute-Chronic TBIVaried
techniqueAcquisitionAnalysisVaried outcome measuresIf they were
includedRelatively few longitudinal studies (but growing).What then
do we know?Low FA is typical of TBI regardless of details.Certain
brain regions are susceptible to TBI.Low FA is associated with
typical adverse TBI outcomes.The prognostic role of dMRI remains
uncertain.
What is missing?Efficacy for prognosis.Will more of the same
type of longitudinal studies be revealing?Will more sophisticated
dMRI measures help?Turnkey techniques usable in the
clinic.Quantification in individual patients.An untenable
hypothesis?
Both drivers in this head-on collision will have injury at the
same brain locations!!A potential missing link?dMRI measurements
are typically determined in an unreasonable manner:a priorilarge
ROINo accounting of interindividual variation.Patient specific
delineation of dMRI abnormalities is needed.
Aim 2: The case for individualized measurement**:Necessary for
clinical use.Arguably the more appropriate approach for
research.
**dMRI measures are extracted using individualized techniques.
Studies of efficacy, etc. employ these measures at the group
level.What has been done: 12 published studieshistograma priori ROI
analysistractographyVoxelwise 1 vs. many T-testVoxelwise
Z-scoreAnalysis of individual measures from group studiesROI,
tractography>100 Case reportsRequirements for individualized
detection of dMRI abnormalitiesStable dMRI measureComparable
normative dataExcellent co-registrationMetric for
quantificationThreshold for abnormality
These steps are ROI-agnosticAn approach to individualized
detection (voxelwise)Enhanced Z-score Microstructural Assessment
for Pathology (EZ-MAP)Regression adjustment of dMRI dataVoxelwise
Z-scoreBootstrap resampling of reference varianceThresholding and
clustering**Kim, et al., PLoS ONE 2013**Lipton, et al. Brain
Imaging and Behavior 2012
Patient APatient BPatient CEZ-MAP: Three mTBI
Patients13Presentation in the clinic
Can you do this in real life?Normative dataData qualityData
consistencyQuantitative analysisValidationQuality assurance
Not for the faint of heart.Need for accessible approaches.
Courtesy: Roman Fleysher, PhDMight dMRI be a better test than
the literature suggests?Most studies are based on group-level
identification.It is highly unlikely that injury mechanism is
uniform across patients.If injury variability leads to varied
spatial distribution of pathology, simple group-wise comparisons
may be very insensitive and poor prognostic tools.their anatomical
location does not always converge. This lack of convergence is not,
however, surprising, given the heterogeneity of brain
injuries****Shenton, et al. Brain Imaging and Behavior
2012Individualized dMRI measures outperform group-level
identification26 mTBI patients/40 controlsNormal CT; sMRI, SWI,
etc.3T DTI
