Page 61 - NNS 2014 Program & Abstract Book

Joshi, S.1, Merchant, R.E.6, Pham, D.L.1 , Roy, M.J.1,3, Shenouda,
C.N.1, vanderMerwe, A.J.1, Diaz-Arrastia, R.R.1,3
1Center for Neuroscience and Regenerative Medicine, Bethesda,
2NINDS, Bethesda,
3USUHS, Bethesda,
4NIH Clinical Center, Bethesda,
5University of Maryland Medical Center, Baltimore,
6Virginia Commonwealth University, Richmond
Inconsistency in diagnosis and medical documentation of TBI pres-
ents a significant obstacle to recruitment of patients for acute and
subacute therapeutic trials. Three unproven constructs were im-
plemented: variable recruitment ‘‘pathways’’ appropriate to each
population identified, a ‘‘TBI Evidence’’ assessment standardized
across populations, and research MRI in the acute period.
Acute TBI patients were recruited at four clinical sites (Bethesda
Suburban Hospital, Washington Hospital Center, Virginia Common-
wealth University Hospital, and University of Maryland Shock
Trauma Center). Sub-acute and chronic patients were recruited
through NIH Clinical Center or by telephone. A locally developed
TBI evidence assessment was used to classify subjects into groups
(Definite/Probable/Possible) by best available evidence supporting a
TBI diagnosis. Severity was classified according to DOD/VA clinical
practice guidelines, with the addition of ‘‘Complicated Mild’’ based
on imaging abnormalities.
During the one-year study period, 278 subjects were enrolled; 186
(67%) at acute sites, 46 (17%) at NIHCC, and 46 (17%) by telephone.
Median age was 47 (IQR 31-57) years, 40% were female, and 31%
belonged to racial minorities. Median time-from-injury to MRI was 59
(IQR 24-216) hours overall and varied between acute sites (26–69
hours), in contrast to NIHCC (8 years). Approximately one-half of
subjects with imaging at NIHCC and acute sites had TBI-related
abnormalities on CT or MRI. TBI severity across all pathways was
35% mild, 26% complicated mild, 22% moderate, 10% severe, and
7% unclassified. 68% of subjects had Definite TBI.
The multi-pathway approach, coupled with acute MRI, provides a
diverse pool of well-characterized TBI subjects for referral to inter-
ventional and observational studies.
Key words
classification, early MRI, multi-site, recruitment, screening
A2-14
CHARACTERIZING TBI RADIOLOGY READS USING THE
ANNOTATION AND IMAGE MARKUP PLATFORM
Magrath, E.R.1, Pham, D.L.1, Chou, Y.Y.1, Afzal, M.M.1, Rao, A.2,
Mongkolwat, P.3, Latour, L.4, Butman, J.A.5
1CNRM, HJF, Bethesda, United States
2Radiology, UCSF, San Francisco, USA
3Radiology, Northwestern University, Chicago, USA
4Stroke Diagnostics and Therapeutics, NINDS, Bethesda, United
States
5Radiology, UCSF, San Francisco, United States
6RAD&IS, Bethesda, United States
Standardizing clinical reads of diagnostic images acquired from TBI
patients is an important step towards enabling systematic character-
ization of radiological findings across multiple sites and research
studies. Annotation and Image Markup (AIM) software is a freely
available software package for performing computerized entry of
structured radiological interpretations while simultaneously allowing
visualization and annotation of the imaging data. In this work, we
report on the results of employing this software for the CNRM
Screening Protocol, a study involving MR and CT imaging collected
from 5 different sites. Inclusion criteria for this protocol allow for any
participant 18 years or older having symptoms of concussion, TBI,
post concussion syndrome, or post concussion disorder. Findings of
each patient were compared to the clinical reads.
Images were read by a neuroradiologist using a release of the AIM
workstation developed especially for TBI. A template was created
for entering the radiological interpretations based on an abbreviated
version of the NINDS CDE’s. The following radiological findings
were considered: 1) TBI, 2) microbleeds, 3) extraparenchymal
hemorrhage, 4) contusion, 5) diffuse axonal injury, and 6) en-
cephalomalacia.
A total of 276 scans (186 CT, 90 MRI) were evaluated and com-
pared to their clinical reads. The clinical comparison was obtained by
manually extracting relevant information from the clinical radiology
read available in the patient’s medical record. Overall, 208 reads
(75.4%) yielded identical results compared to the clinical reads, while
68 (24.6%) were different in one or more categories. Findings were
similar except for microbleeds, which was much higher in the AIM
reads (30% vs. 6%)
The AIM platform provides a convenient informatics solution for
structured radiology reads in TBI studies.
Key words
CT, imaging, MRI, TBI
A2-15
CONNECTOME-SCALE ASSESSMENTS OF STRUCTURAL
AND FUNCTIONAL CONNECTIVITY IN MILD TRAUMATIC
BRAIN INJURY AT THE ACUTE STAGE
Iraji, A.1, Chen, H.2, Ayaz, S.I.1, Kulek, A.1, Welch, R.1, O’Neil, B.1,
Haacke, E.M.1, Liu, T.2, Kou, Z.1
1Wayne State University, Detroit, USA
2University of Georgia, Athens, USA
Mild traumatic brain injury (mTBI) is difficult to diagnose in
emergency setting due to the negative findings in clinical imaging
in most mTBI patients. Several functional network alternations
have been reported in mTBI; however, the large scale network
alternations, particularly at connectome scale, is still unknown. In
this study, we adopted a novel approach to analyze both struc-
tural and functional network changes at connectome level in mTBI
patients.
Forty (40) mTBI patients and 50 demographically matched healthy
subjects were recruited in our local hospital. The patients were
scanned at the acute setting before their discharge. Both diffusion
tensor imaging (DTI) and resting state functional magnetic resonance
imaging (fMRI) data were acquired. In data analysis, a novel approach
called DICCCOL (dense individualized and common connectivity-
based cortical landmarks) was applied. Each DICCCOL node is a
functional landmarker with consistent white matter (WM) fiber con-
nection profile across individuals and thus preserves the same func-
tional role across individuals.
Among 358 nodes on DICCCOL templates, 41 nodes were identi-
fied as discrepant DICCCOL nodes. The WM pathways associated
with these discrepant nodes include corpus callosum, superior and
inferior longitudinal fasciculi, cincuglum, arcuate fasciculus, and
dorso lateral frontal white matter. Functional connectivity analysis of
the common 317 nodes showed 60 functional connectivities as the
most distinctive and discriminative features of our data to differentiate
patients from healthy controls, labeled as connectomic signatures.
A-29