Relationship between post-traumatic amnesia and white matter integrity in traumatic brain injury using tract-based spatial statistics

This study used tract-based spatial statistics to examine the relationship between post-traumatic amnesia (PTA) and white matter integrity in patients with a traumatic brain injury (TBI). Forty-seven patients with TBI in the chronic stage and 47 age- and sex-matched normal control subjects were recruited to the study. Correlation coefficients were calculated to observe the relationships among the PTA duration, white matter fractional anisotropy (FA) values, and mini-mental state examination (MMSE) results in the patient group. Both before and after Benjamini–Hochberg (BH) corrections, FA values of 46 of the 48 regions of interests of the patient group were lower than those of the control group. The FA values of column and body of fornix, left crus of fornix, left uncinate fasciculus, right hippocampus part of cingulum, left medial lemniscus, right superior cerebellar peduncle, left superior cerebellar peduncle, and left posterior thalamic radiation (after BH correction: the uncinate fasciculus and right hippocampus part of cingulum) in the patient group were negatively correlated with PTA duration. PTA duration was related to the injury severity of eight neural structures, each of which is involved in the cognitive functioning of patients with TBI. Therefore, PTA duration can indicate injury severity of the above neural structures in TBI patients.

In the current study, we investigated the relationship between the PTA duration and white matter integrity in TBI patients by undertaking TBSS.

Methods
Subjects. Forty-seven patients with TBI (22 men, 25 women; mean age 45.21 ± 17.68 years; range 20 ~ 79 years) and 47 age-and sex-matched normal control subjects (22 men, 25 women; mean age 40.60 ± 12.49 years; range 22 ~ 74 years) with no history of neurological/psychiatric disease were recruited to this study. The 47 patients with TBI were recruited according to the following inclusion criteria: (1) first-ever TBI, (2) age 20-79 years, (3) DTI scans obtained during the chronic stage (> 4 weeks after onset), and (4) no previous history neurologic/ psychiatric disease. Those TBI patients with a continuous PTA state were excluded. Table 1 lists the demographic and clinical data of the patient and control groups. No significant difference in age or sex distribution was observed between the patient and control groups (p > 0.05). All procedures were performed following relevant guidelines and regulations. This study was performed retrospectively and conducted in accordance with the recommendations of the institutional review board of Yeungnam University Hospital. All of the patients and control subjects provided signed, informed consent and the institutional review board of Yeungnam University hospital approved the study protocol (ethical approval number: YUMC-2019-06-032).

Cognitive function evaluation.
A mini-mental state examination (MMSE) was performed to assess the subject's cognitive function. MMSE evaluates orientation, memory, attention, calculation, visuospatial, and language abilities 15 . The maximum score is 30 points, and a score of 24 or lower indicates that a patient possibly has cognitive impairment 16 . The reliability and validity of the MMSE are well-established 15 . The MMSE score was obtained at the same time of DTI scanning (average of 7.43 ± 5.47 months after the onset of TBI).
Diffusion tensor imaging. The DTI data were acquired at an average of 7.43 ± 5.47 months after head trauma onset using a 1.5 T Philips Gyroscan Intera (Hoffman-LaRoche, Best, Netherlands) with 32 non-collinear diffusion sensitizing gradients by performing single-shot echo-planar imaging. For each of the 32 noncollinear diffusion sensitizing gradients, 65 contiguous slices were acquired parallel to the anterior commissureposterior commissure line 17 . The imaging parameters were as follows: acquisition matrix = 96 × 96, reconstructed to matrix = 192 × 192 matrix, field of view = 240 mm × 240 mm, TR = 10,398 ms, TE = 72 ms, parallel imaging reduction factor (SENSE factor) = 2, EPI factor = 59 and b = 1000 s/mm 2 , NEX = 1, slice gap = 0, and slice thickness = 2.5 mm 17 . Eddy current correction was performed to correct image distortion and head motion effects by using a diffusion registration package (Philips Medical Systems). Analysis of diffusion-weighted imaging data was performed by using tools within the Oxford Centre for Functional Magnetic Resonance Imaging of the Brain Software Library (www. fmrib. ox. ac. uk/ fsl).
Tract-based spatial statistics. Functional MRI assessment tools included in the FMRIB Software Library (FSL) were used to perform the data analyses. A previously described method was used to generate the fractional anisotropy (FA) maps 14 . Voxel-wise statistical analysis of the FA data was performed using TBSS, as implemented in FSL 18 . A nonlinear registration algorithm (www. doc. ic. ac. uk/ ~dr/ softw are) was used to align the FA data for all subjects, obtained via FSL tools, to a template of average FA images (FMRIB-58) in Montreal Neurological Institute space. A mean FA image was produced and thinned to generate a mean FA skeleton representing the centers of all tracts common to the group members. A threshold was applied to a binarized mean FA skeleton at FA values > 0.2 before the resulting data were fed into the voxel-wise statistical analysis. The aligned FA data for each subject were then projected onto the mean skeleton, and voxel-wise cross-subject statistics were obtained to assess the differences between each group's FA values. The results were corrected for multiple comparisons by controlling for the family-wise error rate after performing threshold-free cluster enhancement. For further exploration of the data, the voxels identified by TBSS as showing substantial differences in each tract were selected, and the mean FA values for each subject were calculated. To summarize the FA data in a conventional neuroanatomical context and to identify relevant white matter tracts, mean FA values were calculated across the skeleton and within 48 regions of interest (ROIs), which were based on the intersections between the skeleton and the probabilistic Johns Hopkins University white matter atlases.  . Multiple comparisons were corrected by the Benjamini-Hochberg (BH) procedure to control the false discovery rate, which represents the proportion of significant outcomes that are likely to be false positives 19 . To confirm the presence of white matter damage in the patient group, a multivariate analysis of variance (MANOVA) was performed to determine differences between the patient and control groups' FA values using Pillai's trace that indicates a positive-valued statistic for contribution to the model 20 . The correlations between PTA duration and patient's FA value, as well as between MMSE score and FA value, were evaluated by undertaking Pearson's correlation analysis; p < 0.05 and false discovery rate cut off < 0.1 were considered statistically significant. Table 2 presents a comparison of the FA values of the patient and control groups. The FA values of 46 ROIs among the 48 ROIs showed significant differences between the patient and control groups (Pillai's trace = 0.712, p < 0.05) (Fig. 1A). The BH correction revealed that the FA values of 46 ROIs among the 48 ROIs of the patient group were significantly different with those of the control group (BH p < 0.05). Table 3  After the BH correction, the MMSE score was positively correlated with the FA value of the right hippocampus part of cingulum region. Fourth, both before and after the BH corrections, PTA duration was negatively correlated with MMSE score in the patient group. The FA value represents the degree of directionality of water diffusion and indicates the integrity of white matter microstructures, such as axons, myelin, and microtubules 21 . Thus, a decrease in the FA value of a neural structure represents a decrease in the neural structure's microstructural integrity [11][12][13] . Therefore, the study results showing lower FA values in 46 of the 48 ROIs assessed in the patient group both before and after the BH corrections indicate there is lower microstructural integrity in these neural structures, suggesting the presence of neural injury.

Results
Regarding the correlation between the PTA duration and FA values in the patient group, the FA values of eight white matter neural structures that are related to cognitive function were negatively correlated with PTA duration [22][23][24][25][26][27][28][29][30] . The fornix is involved in episodic memory recall, and the uncinate fasciculus functions are related to learning and memory [22][23][24] . The cingulum has been reported to be an important neural structure for episodic memory and executive control 25 , and the superior cerebellar peduncle is involved in the social cognitive neural network, acting as a major pathway for cerebro-cerebellar connections. In contrast, the posterior thalamic radiation, including the optic radiation, is involved in visuomotor task performance 26,29 . Regarding the medial lemniscus, lesions at the thalamus and prefrontal levels can cause executive dysfunction because this area is involved in a pathway that reaches the prefrontal cortex via the thalamus 27,28,30 . The PTA-FA correlation results suggest that PTA duration in TBI is closely related to the injury severity of the above eight neural structures that are involved in cognitive function. In addition, the results showing positive correlations between MMSE score and FA values of the fornix, uncinate fasciculus, and cingulum in the patient group indicate relationships between cognitive ability and FA values in those neural structures. The revelation of a significant correlation between PTA duration and MMSE score in the patient group both before and after the BH corrections further indicates that PTA duration is associated with cognitive function, and that correlation is consistent with the correlation between PTA duration and FA value of the neural structures involved in cognitive function. Furthermore, after the BH correction, the www.nature.com/scientificreports/ Table 2. Comparison of the fractional anisotropy values of brain regions between the patient and control groups. *Significant differences between the patient and control groups; patient group < control group, p < 0.05. **Significant differences between the patient and control groups using a Benjamini-Hochberg (BH) correction; Values are means ± standard deviations. www.nature.com/scientificreports/ results that the uncinate fasciculus and hippocampus part of cingulum were negatively correlated with PTA duration, and the hippocampus part of cingulum was positively correlated with MMSE score, demonstrated that the uncinate fasciculus and hippocampus part of cingulum were more closely related to PTA duration and cognitive ability. However, the white matter neural structures that did not negatively correlated with PTA duration after the BH correction appeared to be lack of the sensitivity necessary to detect relationship with PTA duration 31 . Many studies have described relationships between PTA duration and neuropsychological outcomes in TBI [32][33][34][35][36][37][38][39][40][41] . These studies have focused on the association between PTA duration and cognitive outcomes, such as verbal, intelligence, learning and memory, speed of information processing, and perceptual or constructional skills. On the other hand, many DTI studies have reported a relationship between cognitive performance and white matter integrity in patients with TBI by examining TBSS results [42][43][44][45][46][47][48] . These studies have reported relationships between white matter integrity and various cognitive functions, such as memory, associative learning, execution, and social function. However, the present study is the first to demonstrate a relationship between PTA duration and subsequent white matter integrity in patients with TBI. Nevertheless, some limitations of this study should be considered. First, various evaluation methods can assess cognitive ability, but this study, because it was retrospectively conducted, only used the MMSE to indicate cognitive function. Second, although DTI data were obtained during the chronic stage, duration from onset to DTI scanning was heterogeneous. Further prospective studies using detailed neuropsychological testing and considering more homogeneous duration from onset to DTI scanning should be encouraged. Third, the TBSS approach reduces white matter tracts to a skeleton framework, thereby delineating the center of the tract and projecting onto that center only the highest value FA along that projection, resulting in a loss of information and the potential presence of artifacts 49,50 . Fourth, TBSS-based methods calculate statistics only for skeleton voxels, resulting in few statistical comparisons in less-relevant voxels. Therefore, the effects of family-wise error correction are reduced 50 .
In conclusion, this study examined the relationship between PTA duration and white matter integrity in TBI patients by undertaking TBSS-based analysis. In patients with TBI, PTA duration was related to injury severity in eight neural structures (especially, the uncinate fasciculus and hippocampus part of cingulum) involved in cognitive function. The results suggest that PTA duration can indicate injury severity of the above neural structures in TBI patients.