Abstract Details

Title Analysis of Regional Brain Volumes After Traumatic Brain Injury and Correlation with Clinical Features

Topic Neuro Trauma, Critical Care, and Sports Neurology

Presentation(s) S2 - Neurocritical Care (1:55 PM-2:06 PM)

Poster/Presentation
Number 006

Background TBI is a prevalent worldwide health problem caused by a mechanical insult to the head, resulting in transient or permanent alteration to brain tissue and/or function. Standard neuroimaging with computerized cranial tomography (CT) and structural magnetic resonance imaging (MRI) is often unrevealing during the evaluation of patients with TBI, particularly those classified as mild TBI.

Objective Examine the value of quantitative analysis of regional brain volumes in the evaluation of traumatic brain injury (TBI).

Design/Methods Medical records and MRI imaging from 44 patients with TBI evaluated at a Level I trauma center were analyzed. Clinical notes were reviewed to assess reported symptoms and physical findings. Regional volumes from TBI subjects were derived using the software package Freesurfer image analysis suite that utilizes a T1-weighted structural scan to calculate volumetric information. A machine learning algorithm, random forests, was employed across volume measurements from 25 regions of interest (ROI) to determine the most important regions for classifying subjects based on clinical outcome and symptomology.

Results Quantitative analysis showed that basal ganglia volume showed the highest variable importance with regards to classifying subjects who exhibited symptoms of cognitive dysfunction (MeanDecreaseGini=1.067, MeanDecreaseAccuracy=5.966e-03). Left lateral ventricle volume was important in classifying subjects with motor and vestibular alterations (MeanDecreaseGini=2.037, MeanDecreaseAccuracy=2.92e-02). Left choroid plexus volume was the most important region for classifying subjects with sensation and somatic dysfunction (MeanDecreaseGini=0.271, Mean Decrease Accuracy=4.82e-03).

Conclusions

TBI damages brain tissue and subsequent volume loss which may result in clinical symptoms. This study showed that volumes of specific ROIs can be used to classify TBI subjects that fall into predetermined symptom categories. The current study is ongoing. It will be extended to a larger cohort to determine whether volume changes in specific ROIs can act as useful clinical biomarkers for chronic symptoms.

Authors/Disclosures
James J. Rock III PRESENTER	No disclosure on file
Kodiak Carb	No disclosure on file
	No disclosure on file
	No disclosure on file
Erika Silverman (University of Pennsylvania)	Ms. Silverman has nothing to disclose.
	No disclosure on file
Megan T. Moyer, NP	No disclosure on file
Danielle Sandsmark, MD	The institution of Dr. Sandsmark has received research support from NINDS. The institution of Dr. Sandsmark has received research support from BrainBox Solutions Inc. The institution of Dr. Sandsmark has received research support from Department of Defense.
Ramon R. Diaz-Arrastia, MD, PhD, FAAN (University of Pennsylvania)	Dr. Diaz-Arrastia has stock in BrainBox, LLC. Dr. Diaz-Arrastia has stock in Nia Therpeutics. The institution of Dr. Diaz-Arrastia has received research support from National Institutes of Health. The institution of Dr. Diaz-Arrastia has received research support from Department of Defense.

��ɫ��

��ɫ��