Synopsis

Keywords: Traumatic brain injury, Brain Connectivity, Gray Matter, White Matter

Traditional concussion self-reporting has significant limitations due to its subjectivity and inconsistency, therefore a more objective MRI-based approach is proposed in this research. A concussion population was assessed using the traditional PCSS method versus DTI anisotropy analysis and rsMRI complexity analysis. Concussed patients were compared to large age and sex matched datasets and a Z-transform was used on white and gray matter ROIs to identify injured areas in the cerebrum. A weighted mean equation was then used to compare Z-scores to the 7-point PCSS. Results confirmed sex differences in self reporting and shows promise as a future objective assessment tool.

Introduction

Concussions are a common and devastating condition that severely affects over 1.7 million people globally each year¹. Unfortunately, patients further suffer from poor diagnostic outcomes as current clinical evaluations are subjective and inconsistent. The goal of this study was to compare the post-concussion symptom scale (PCSS), a routinely used subjective patient self-reporting assessment tool², against an objective assessment measure based on diffusion tensor imaging (DTI) and resting state MRI (rsMRI) analysis, within a concussion population³.

Methods

Images were obtained using a GE MR750 Discovery 3T MRI scanner and a 32-channel RF receiver coil (General Electric Healthcare, Milwaukee, WI). Axial DTI was acquired using a dual echo EPI sequence (60 non-coplanar directions, 6 b=0s/mm² images, TE/TR=87/8800ms, b=1000s/mm², 122x122 matrix, 2mm thickness, FOV=244mm) while resting state BOLD data was acquired using an EPI sequence (FOV=22cm, 64x64 matrix, TE/TR=35/2000ms, 3mm thickness, and 175 temporal points). 26 patients who suffered a concussion were analysed (10 males (33±12 yrs) and 16 females (38±12 yrs)) where a traditional PCSS test (includes 22 symptoms) was administered prior to the scanning session². A minimum of 88 DTI and rsMRI dataset pairs of healthy age and sex matched controls (age matched ±2yrs) were collected through the IDA, NITRC, and HCP databases^4,5,6 to satisfy the Cochran formula for 95% sensitivity⁷. Injured and healthy control datasets were warped into the MNI152 space, eddy corrected, and brain extracted and further segmented into common region of interests (ROIs) for DTI (19 white matter ROIs) and rsMRI (91 gray matter ROIs), respectively^8,9. A homebuilt python processing pipeline was used to perform routine rank-2 tensor analysis (FA, AD, MD, RD) on DTI data and voxel-wise rsMRI temporal complexity based on the Hurst exponent^{3,10,11,12,13}. Statistical map distributions for each healthy control population DTI and rsMRI dataset pairs were created and verified through skewness and kurtosis tests which enabled the application of a Z-score approach to identify subject-specific abnormal ROIs in the concussion patient³. The Z-score magnitudes were then classified into a severity metric on a 6-point scale in order to be compared to the 6-point scaled PCSS². A synthetic MRI-based assessment score was then calculated using a weighted sum mean of the abnormally identified ROI Z-scores for each patient, given DTI and rsMRI Z-score magnitudes. A synthetic assessment score was also calculated for each symptom corroborating the ROIs function and associated pathology (i.e. optic radiation pathology is associated with the symptom of visual problems), (see Figure 1). The weighted mean equation can be explained by $$MRIScore=\frac{1}{2}\left(\frac{\sum(x_i*w_i)}{\sum(w_i)}+\frac{\sum(y_i*w_i)}{\sum(w_i)}{}^{} \right)$$ where the MRIScore represents the rounded MRI-Based symptom score, x, represents the actual Z-score of an abnormal DTI ROI, w, represents the weights of the Z-score, and y, represents the actual Z-score of an abnormal fMRI ROI. Total PCSS scores (i.e. the sum of scores) that were self-reported where then compared to the MRI-Based method (i.e. synthetic scores). Overall accuracy was evaluated with age and sex effects where concussion patients were divided into younger (under 35 yrs) or older (over 35 yrs), and male or female respectively.

Results

Each concussion patient self-reported symptoms using the PCSS and each had at least one DTI or rsMRI abnormality based on our MRI-based objective analysis. For each patient, there was a 15% accuracy in terms of symptom matching the self-reported PCSS with the MRI-based calculated scores. Given that the PCSS is highly subjective and inconsistent, exact symptom matching is notoriously difficult. When assessing the sum of scores (self-reported vs. MRI-based), overall, no trends were found (see Figure 2). However, the evaluation of age and sex yielded some interesting trends where young men (under 35 yrs) tended to underreport concussion symptoms (i.e. lower self-reported scores than the MRI-based scores) whereas older women (over 35 yrs) tend to overreport their symptoms (i.e. higher self-reported scores than MRI-based scores) (see Figure 3). This trend corroborates with traditional literature that outlines younger men, in professional athletic settings, tend to underreport their symptoms^14,15.

Conclusions

The MRI-based approach proposed shows tremendous promise as a future tool to assess concussion patients and their symptoms more objectively. Future work must be performed to identify how self-reporting and MRI-based findings change in a longitudinal fashion in these patients. Our results confirm that sex differences play a significant role in self-reporting and that cultural effects most likely also confound self-reporting results. Together, these factors clearly indicate why objective measures for concussion must be investigated.

Acknowledgements

Special thanks to my supervisors and mentors Dr. Michael D. Noseworthy, Dr. Dinesh Kumbhare, Dt. Stephan Ulmer, and Norm Konyer for their abilities and guidance. More thanks to my fellow Brain Trust colleagues Ethan Danielli, Ethan Samson, Neil MacPhee, Bhanu Sharma, Mahnaz Tajik, Lauren Anderson, Calvin Zhu, and Yves Pauli. Thank you for the financial support from McMaster University, the National Sciences and Engineering Research Council of Canada (NSERC), and the MITACS Accelerate program. Lastly, thank you to my beautiful wife, my family, and friends who have tremendously supported me along the way.