Working Memory And White Matter Microstructure In Mild Traumatic Brain Injury
Sohae Chung1, Els Fieremans1, Xiuyuan Wang1, Charles J Morton1, Dmitry S Novikov1, Joseph F Rath2, and Yvonne W Lui1

1Center for Advanced Imaging Innovation and Research (CAI2R) & Bernard and Irene Schwartz Center for Biomedical Imaging, Department of Radiology, NYU School of Medicine, New York, NY, United States, 2Department of Rehabilitation Medicine, NYU School of Medicine, New York, NY, United States


Working memory is a critical cognitive function implicated after mild traumatic brain injury (MTBI). Here, we investigate the association between white matter microstructure and working memory in normal controls (NC) and MTBI patients, using diffusion white matter tract integrity (WMTI) and WAIS-IV subtests, respectively. For the NC group, significant correlations were observed in axonal water fraction (AWF; higher axonal density/myelination) and mean kurtosis (MK; greater tissue complexity) with letter-number sequencing (LNS). However, such relationships were not present in the MTBI group.


Working memory is at the core of many critical cognitive functions, responsible for holding, processing and manipulating information.1 Impaired working memory is associated with a range of neurological disorders2 including mild traumatic brain injury (MTBI).3 Little is known about whether and how working memory relates to underlying white matter (WM) microstructure. In this study, we investigate the association between WM microstructure and performance on a set of working memory tasks in healthy adults and patients with MTBI. We relate compartment specific WM tract integrity (WMTI) metrics4 derived from multi-shell diffusion MRI as well as diffusion tensor/kurtosis imaging (DTI/DKI) metrics to Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV)5 subtests tapping auditory-verbal working memory.


We studied 18 patients with MTBI (mean age, 30±7, range 22-45 yrs; 8 male) within a month of injury (mean, 17 days) and 15 normal controls (NC) (mean age, 31±7, range 19-45 yrs; 7 male). All subjects are right-handed and native English speakers. MR imaging was performed on a 3T MR scanner (Skyra, Siemens). Diffusion imaging was performed with 5 b-values (0.25,1,1.5,2,2,2.5ms/µm2) along with 6,20,20,30,60 diffusion encoding directions using multiband (factor of 2) echo-planar imaging for accelerated acquisitions. Three images with b=0 were acquired. For geometric distortion correction, an additional image with b=0 was acquired with reversed phase encoding direction. Other parameters were: FOV=220×220mm2, matrix=88×88, resolution=2.5×2.5×2.5mm3, slices=56, TR/TE=4.9s/95ms, BW/pixel=2104Hz, a GRAPPA factor=2. We calculated maps of WMTI metrics (axonal water fraction [AWF], intra-axonal diffusivity [Daxon], extra-axonal axial and radial diffusivities [De,a and De,r]), as well as DTI metrics (fractional anisotropy [FA], mean, axial, radial diffusion coefficients [MD, AD, RD]) and DKI metrics (mean, axial, radial kurtosis [MK, AK, RK]). Auditory-verbal working memory was assessed using two WAIS-IV subtests: 1) Digit Span (DS) which includes DS Forward [DSF], Backward [DSB] and Sequencing [DSS]; and 2) Letter-Number Sequencing [LNS]. We performed tract-based spatial statistics (TBSS)6 with age and gender as covariates to reveal possible correlations between working memory subtest scores and the diffusion metrics. The resulting statistical maps from TBSS were thresholded at p < 0.05 (corrected for multiple comparisons). Spearman’s partial rank correlation coefficients were also calculated for those significant voxels on the skeleton, adjusted for age and sex.


In NCs, there were several brain regions demonstrating statistically significant positive correlation between AWF and LNS, the most complex of the set of four working memory tasks, most notably in parietal WM, left greater than right; others areas included right posterior limb of internal capsule, left body of corpus callosum, left posterior/superior corona radiata, bilateral superior longitudinal fasciculus, left inferior longitudinal fasciculus, bilateral anterior thalamic radiation and left cingulum (Fig.1). A small region in the right frontal WM (inferior fronto-occipital fasciculus) also showed a significant positive correlation between MK and LNS. For the MTBI group, no significant positive correlations were found from TBSS. No other diffusion metrics showed area of significant difference surviving correlation for multiple comparisons. Furthermore, no significant correlations were observed with the three relatively simpler digit span (DS) tasks. Fig.2 presents the scatter plots of each significant metric (AWF, MK) and LNS, showing that higher AWF (Fig.2(left); r=0.89, p < 0.001) and higher MK (Fig.2(right); r=0.92, p < 0.001) are associated with better performance on the LNS for those significant voxels on the skeleton for the NC group (black circle), but no significant correlations were observed for the MTBI group on the same location (red triangle).


In this study, for the NC group, significant correlations were observed between AWF and LNS, and to a lesser degree between MK and LNS. Higher AWF has been previously shown in regions of higher axonal density and greater myelination both of which could contribute to greater efficiency in information processing.6 MK reflects degree of tissue microstructural complexity. Asymmetric correlation with left hemisphere microstructural measures is consistent with expected lateralization for tasks of auditory-verbal working memory tested in this case.7 The correlation in NCs was quite strong and highly statistically significant; however, none of these relationships persisted in the MTBI group. In MTBI, there is evidence of WM injury8 as well as altered patterns of brain activation after injury.9 These factors likely contribute to the loss of correlation between WM microstructure and performance on LNS task in MTBI subjects observed here.


The relationship normally present in control subjects between measures of white matter microstructure and a complex task of auditory-verbal working memory is not observed in patients after MTBI and may reflect known WM injury and changes in functional organization that occur after injury.


This work was supported in part by grant funding from the National Institutes of Health (NIH): R01 NS039135-11 and R21 NS090349, National Institute for Neurological Disorders and Stroke (NINDS). This work was also performed under the rubric of the Center for Advanced Imaging Innovation and Research (CAI2R, www.cai2r.net), a NIBIB Biomedical Technology Resource Center (NIH P41 EB017183).


  1. Baddeley A. Working memory: looking back and looking forward. Nat Rev Neurosci. 2003;4(10):829-839.
  2. Sandry, J. Working memory and memory loss in neurodegenerative disease. Neurodegener Dis Manag. 2015;5:1-4.
  3. Sandry, J., et al. Individual differences in working memory capacity predicts presponsiveness to memory rehabilitation after traumatic brain injury. Arch Phys Med Rehabil. 2016;97:1026-1029.
  4. Fieremans, E., et al. White matter characterization with diffusion kurtosis imaging. Neuroimage. 2011;58:177-188.
  5. Wechsler, D. Wechsler Adult Intelligence Scale. Fourth Edn. Pearson Assessment. 2008.
  6. Jelescu, I.O. et al. In vivo quantification of demyelination and recovery using compartment-specific diffusion MRI metrics validated by electron microscopy. Neuroimage. 2016;132:104-114.
  7. Thomason, M.E. et al. Development of spatial and verbal working memory capacity in the human brain. J Cogn Neurosci. 2009;21:316-332.
  8. Chung, S. et al. Characterizing white matter microstructural changes after mild traumatic brain injury based on diffusion white matter tract integrity (WMTI) and Shannon entropy. 25th ISMRM, Honolulu, Hi, USA. 2017;#4530.
  9. McAllister, T.W. et al. Brain activation during working memory 1 month after mild traumatic brain injury. Neurology. 1999;53:1300-8.


Figure 1. (Top) For the NC group, TBSS results show multiple areas of significant positive correlation (corrected p<0.05) between AWF and LNS, particularly involving left greater than right parietal WM. Other regions involved: right posterior limb of internal capsule (pLIC), left body of corpus callosum (BCC), left posterior/superior corona radiata (pCR/sCR), bilateral superior longitudinal fasciculus (sLF), left inferior longitudinal fasciculus (iLF), bilateral anterior thalamic radiation (aTR), and left cingulum (CGL). (Bottom) For the MTBI group, no significant correlations were found.

Figure 2. Scatter plots show strongly significant positive correlations (left) between AWF and LNS (r = 0.89, p < 0.001) and (right) between MK and LNS (r = 0.92, p <0.001) in normal controls (NC) (black circle). No such correlations were found in the MTBI groups (red triangle).

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)