Longitudinal assessment of magnetization transfer ratio, brain volume, and cognitive functions in diffuse axonal injury

Abstract Background Diffuse axonal injury (DAI) is a frequent mechanism of traumatic brain injury (TBI) that triggers a sequence of parenchymal changes that progresses from focal axonal shear injuries up to inflammatory response and delayed axonal disconnection. Objective The main purpose of this study is to evaluate changes in the axonal/myelinic content and the brain volume up to 12 months after TBI and to correlate these changes with neuropsychological results. Methods Patients with DAI (n = 25) were scanned at three time points after trauma (2, 6, and 12 months), and the total brain volume (TBV), gray matter volume, and white matter volume (WMV) were calculated in each time point. The magnetization transfer ratio (MTR) for the total brain (TB MTR), gray matter (GM MTR), and white matter (WM MTR) was also quantified. In addition, Hopkins verbal learning test (HVLT), Trail Making Test (TMT), and Rey–Osterrieth Complex Figure test were performed at 6 and 12 months after the trauma. Results There was a significant reduction in the mean TBV, WMV, TB MTR, GM MTR, and WM MTR between time points 1 and 3 (p < .05). There was also a significant difference in HVLT‐immediate, TMT‐A, and TMT‐B scores between time points 2 and 3. The MTR decline correlated more with the cognitive dysfunction than the volume reduction. Conclusion A progressive axonal/myelinic rarefaction and volume loss were characterized, especially in the white matter (WM) up to 1 year after the trauma. Despite that, specific neuropsychological tests revealed that patients’ episodic verbal memory, attention, and executive function improved during the study. The current findings may be valuable in developing long‐term TBI rehabilitation management programs.


INTRODUCTION
Traumatic brain injury (TBI) is a public health and socio-economic problem worldwide (Roozenbeek et al., 2013). In the United States, around 2.8 million new cases are diagnosed each year, with approximately 57,000 deaths (Frieden et al., 2015). In addition, 5.3 million people are estimated to live with a TBI-related disability in the United States, accounting for around 2% of the population (Akin et al., 2020;Frieden, 2015).
Diffuse axonal injury (DAI), a type of lesion characterized by microscopic axonal damage scattered throughout the encephalic parenchyma secondary to acceleration-deceleration forces, is one of the most common pathological mechanisms of TBI (Adams et al., 1989).
DAI has been almost universally demonstrated in fatal TBI and is estimated to be the mechanism most likely responsible for many cognitive deficits resulting from moderate to severe TBI (Skandsen et al., 2010).
Although magnetic resonance imaging (MRI) is more sensitive for the detection of axonal shear injuries than computed tomography (CT) and the currently preferred modality for assessing parenchymal involvement in DAI (Liu et al., 2014), it significantly underestimates the brain alterations. Furthermore, up to now, there is no well-established MRI sequence that can accurately evidence the severity of DAI or precisely monitor the progression of the post-traumatic brain damage (Cole et al., 2018;de La Plata et al., 2007).
The need for alternative MRI techniques that reflect the complex underlying pathophysiological processes in DAI and allow reliable follow-up of the brain damage has grown after experimental evidence of delayed cerebral changes after TBI (Irimia et al., 2012;Mamere et al., 2009). Examples of these techniques are the magnetization transfer imaging (MTI) and the brain volumetry.
The magnetization transfer ratio (MTR), a quantitative measurement obtained from the MTI, has been described as a reliable marker for monitoring myelinic and axonal content in the gray matter (GM) and white matter (WM) after TBI Gareau et al., 2020).
Clinical studies have described a reduction in the MTR in the normalappearing white matter (WM) up to years after TBI. However, few of these studies include moderate and severe TBI, and in these, the MTR quantification is limited to small regions of interest (ROI) (Kumar et al., 2003;Mamere et al., 2009). In addition, they include a heterogeneous group of subjects with cortical contusions, hematomas, and ischemic injuries under the more general term "TBI". Up to our knowledge, there are only two studies in patients exclusively with DAI Sinson et al., 2001), and none of these include neuropsychological assessment.
The number of studies describing volume change of various neuroanatomic structures in the chronic phase after TBI is higher, and progressive encephalic volume reduction is well-established (Bigler, 2001;Cole, 2018;Tate, 2000). Nonetheless, the association of DAI with other intracranial lesions is also observed in most studies, and few are restricted to subjects exclusively with DAI (Ding et al., 2008;Moen et al., 2012;Warner et al., 2010).
Finally, DAI damages crucial networks between the cortex and the deep white matter structure. This injury leads to cognitive deficits, especially non-specific memory impairment that can involve the encoding and retrieval of verbal and visual information equally (Levin, 1990;Palacios et al., 2013). Cognitive symptoms are more severe just after an injury and may only be present for a short time (Li et al., 2006;Zaninotto et al., 2017). Given the importance of verbal learning and memory in education and academic achievement, a significant number of studies have looked into the effects of TBI on verbal memory, but few have explored visuospatial memory after brain injury (Shum et al., 2000;Zaninotto et al., 2017).

Study population
This was a 1-year prospective longitudinal study for which approval was obtained from the Institutional Review Board. In addition, written informed consent was obtained from all participants or their representatives. Out of the 225 consecutive patients with TBI admitted to the Emergency Department who were initially considered for this study, 186 were excluded for not meeting the epidemiological, clinical, or radiological eligibility criteria.
The inclusion criteria were the following: (1) age between 18 and 55 years, (2) a TBI history dating with no longer than three months before the inclusion of the patient in the study, (3) an admission Glasgow Coma Scale (GCS) score of 3−12, (4) clinical and computed tomography (CT) diagnosis of DAI, and (5) a Marshall score of I, II, or III (Marshall et al., 1992) based on the CT images. The exclusion criteria were the following: (1) presence of contusion(s) larger than 10 mm or that may impair automated brain volume segmentation, (2) a midline shift greater than 0.5 cm, (3) extra-axial fluid collection(s) with a compressive effect on the brain structures, or (4) any contraindications to brain MRI.
The remaining 39 patients who followed the abovementioned criteria were submitted to an initial neuropsychiatric (NP) evaluation and a brain MRI. Out of these, 13 abandoned the study in the followup period, and one patient died. Thus, the final sample included 25 patients that were prospectively evaluated at three time points: around 2 months after the trauma (1), 6 months after the trauma (2), and 12 months after the trauma (3). Due to the comprehension difficulties, mental confusion, and agitation typically seen in the subacute stage post-trauma, the patients were submitted to neuropsychological assessment only at time points 2 and 3.

Image acquisition
All data were acquired on a 3T MRI scanner (Phillips Achieva, Best, The Netherlands) with an eight-channel head coil. Besides the sequences detailed below, a routine protocol that included fluid-attenuated inversion recovery (FLAIR), T2-weighted images, and diffusion-weighted imaging (DWI) was also acquired.
The volumetric T1-weighted Fast Field Echo (3DT1-FFE) sequence was obtained in the sagittal plane with the following parameters: inver-

Image processing
The MTR was quantified using the FMRIB Software Library (FSL -Analysis Group, FMRIB, Oxford, UK; available at http://www.fmrib.ox.ac. uk/fsl/) version 5.0. For that, segmentation of the GM and WM was performed in the anatomical 3D T 1 using the FMRIB Automated Segmentation Tool (FAST). Then, the automatic segmentation generated by FAST had to be corrected using a more robust mask of the subcortical GM (amygdala, hippocampus, caudate, putamen, pallidum, and thalamus) obtained with the FMRIB's Integrated Registration and Segmentation Tool (FIRST) (Patenaude et al., 2011). From this process, we attained accurate T1-based volumetric images corresponding to the GM, WM, and total brain (TB) from each MRI study ( Figure 1).
Second, we used the PRESTO images for the exclusion of the microhemorrhages from the MTR map. In the MTI processing, blood content is a major inconvenience since it corrupts the results by decreasing the MTR, and ideally, should be excluded from the MTR quantification.
To accomplish that, we segmented the neural tissue from the PRESTO images into four different classes according to voxel intensity using FAST to obtain a mask containing only voxels with very low signals neuroimaging data that allows automated parcellation of the corti-cal grey and white matter and subcortical volumes (Fischl, 2012). Its longitudinal processing enables a temporally unbiased evaluation of an arbitrary number of time points by treating all inputs the same and generating a within-subject template via iterative alignment of all input images to a median image, using a symmetric robust registration method (Reuter et al., 2012). Due to the long-time processing of the pipeline's first step (recon-all command), GNU Parallel was used, which allows multiple jobs to be performed simultaneously, one per core of the used computer, in our case, six at a time (Tange, 2011). The mean time for recon-all part of the processing, which alone includes 31 steps, was ∼6 h per image; using GNU Parallel 4, processing results for six T1weighted images were obtained in this same time. Following FreeSurfer recommended best practices; we performed an individual inspection of the output files. For this study, the total brain volume (TBV), white matter volume (WMV), and gray matter volume (GMV) were measured.

Hopkins verbal learning test (HVLT)
This test evaluates the episodic verbal memory and consists of a 12item word list which is read to subjects on three successive learning comprises 24 words, including the 12 target words and 12 confounding words, six semantically related, and six semantically unrelated (Benedict et al., 1998). The results are expressed as the total raw score.

Trail Making Test (TMT)
Assesses attention and executive function. The test consists of two parts (A and B), one with numbers and the other with numbers and letters. In Part A, the task is to connect the numbers in ascending order (1-2-3-4. . . ). In Part B, subjects are required to alternately connect numbers and letters in ascending order for the numbers and sequential order for the alphabet letters (1-A-2-B-3-C. . . ). The time to complete each part is used when analyzing performance (Spreen & Strauss, 1998). The results are expressed as the total raw score. test consists of the presentation of words, and the patient is asked to define them. In the matrix reasoning test, a matrix of abstract pictures with one picture missing is presented, and the patient has to choose an option that better suits the missing image.

Statistical analysis
Statistical tests were performed using the software RStudio ver-

Demographics
A total of 25 outpatients (mean age = 28.36 years, SD = 9.17 years) with a history of moderate or severe DAI according to the GCS score at admission (<13) completed the 3 MRI scans. The majority of these patients were male (n = 22), suffered from motorcycle accidents (n = 14), presented with severe head trauma (n = 16) and a Marshall score of II on the initial CT (n = 21). The mean time interval between trauma and hospital admission was 39 min (SD = 16.50). Table 1 provides basic demographic data and clinical characteristics.
One patient did not perform the neuropsychological assessment on time point 2, and three did not perform the NP assessment on time point 3. These four patients were excluded from the longitudinal NP evaluation and the correlation analysis between imaging findings and neuropsychological performance but were included in the imaging evaluation.   Abbreviations: GMV, gray matter volume; TBV, total brain volume.

F I G U R E 5
Scatter plot with regression line showing the magnetization transfer ratio (MTR) change over time for the gray matter (GM), white matter (WM) and total brain (TB). The x-axis represents the time from trauma in days, and the y-axis represents the mean MTR F I G U R E 6 Scatter plot with regression line showing the volume change over time for the gray matter (GM), white matter (WM) and total brain (TB). The x-axis represents the time from trauma in days, and the y-axis represents the volume     Table 4 summarizes the Pearson or Spearman correlation regarding the neuropsychological results. The results from neuropsychological assessment in time points 2 and 3 were independently correlated with the annualized atrophy and the MTR reduction rates.

Correlation of neuropsychological results and MTR or volume reduction
Our data show a significant negative correlation of TB MTR, GM MTR, and WM MTR reductions with the HVLT-I and HVLT-D scores in time point 2. There was also a significant positive correlation between TB MTR decline and the TMT-A score in time point 2 (R = 0.41, pvalue = .04). In addition, there was a significant negative correlation TA B L E 5 Pearson's correlation results (Normality = 0) and Spearman's correlation results (Normality = 1) between volume reduction and MTR reduction in each encephalic compartment Abbreviations: GM, gray matter; TB, total brain; WM, white matter.  Table 5 depicts the results of the Pearson or Spearman correlation between MTR decline and volume reduction. There was no association between volume reduction and MTR decrease for the TB, GM, or WM in any of the intervals studied. Table 4 shows the Pearson or Spearman results when the variance in the neuropsychological scores (between time points 2 and 3) was correlated to the MTR or volume decline during the same interval. The WM MTR reduction between time points 2 and 3 was negatively correlated with the IQ increase (R = −0.51, p-value = .01) and with the improvement in the TMT-A scores (R = −0.43, p-value = .04). The TBV atrophy throughout this period was significantly correlated with the variance in the TMT-A (R = 0.60, p-value = 0) and the ROCF copy test (R = 0.44, p-value = .04). In addition, there was a significant negative correlation between the variance in the ROCF recall test and the GM atrophy (R = −0.51, p-value = .01). There were no additional significant correlations between the reduction in volume or MTR and the variance in neuropsychological outcomes. Table 6 4 DISCUSSION

Correlation of the variation between neuropsychological results and imaging results
This study characterizes a progressive decline in the WMV and TBV volume up to one year after TBI, with an annualized TBV atrophy rate of 1.38% and a WMV rate of 3.61%. These values are significantly higher than the annual atrophy rates of up to 0.52% for the TB and 0.47% for the WM in normal individuals (Ge et al., 2002;Smith et al., 2007).
The annualized WMV atrophy rate matches the atrophy rate of around seven years of normal aging. In addition, the annualized TBV atrophy rate matches the 1.5% rate described in moderate and severe TBI, but not for the WM, in which the reported atrophy rate is around 1.4% (Ding et al., 2008;Sidaros et al., 2009;Trivedi et al., 2007).
We also describe a progressive decline of the mean TB MTR, GM MTR, and WM MTR, which reflects the Wallerian degeneration of axons, mitochondrial dysfunction, excitotoxicity, oxidative stress, and the apoptotic cell death of neurons and glia observed in the delayed phase after TBI (Ng et al., 2019). The interval of more significant WM MTR decline (between time points 1 and 2) preceded the one with greater WMV atrophy (between time points 2 and 3) and no correlation was found between MTR and volume reductions throughout the study. The relative independence of MTR decline and atrophy could be explained by pathological correlations that suggest that, despite atrophy being associated with MTR decline, demyelination is the primary driver of MTR reduction. (Foss et al., 2013). It could also be reinforced by histological research in rats with induced encephalomyelitis, which revealed no association between cortical demyelination and atrophy, implying that the correlation between myelinic loss and parenchymal loss is complex (Pomeroy et al., 2008).
Our follow-up also evidenced a significant improvement in the HTLV-I, TMT-A, and TMT-B scores between time points 2 and 3. In addition, improvement in the TMT-A scores showed a strong correlation with the WM MTR decline and TBV atrophy. The improvement in TA B L E 6 Correlation of the brain atrophy and MTR reduction (between time points 2 and 3) and the variance between the neuropsychological tests results TMT results agrees with a previous study that evidenced improvement in TMT-A and TMT-B after a year post-trauma (Farbota et al., 2012).

Test
After a brain injury, partial or complete recovery of specific cognitive domains can be attributed to both the restoration and relearning of lost capabilities and adaptation and compensating of spared functions (Nudo, 2013).
Previous research has revealed that patients with TBI often have an amnesic pattern characterized by verbal episodic memory deficits, particularly in late recollection of new material (Vakil, 2005). However, there is disagreement regarding the period of most significant cognitive recovery after a TBI; some believe it to be 6-12 months, while others suggest the interval between 1 month to 1 year after the trauma (Dikmen et al., 1987). Patients with severe TBI exhibited improved verbal episodic memory scores when tested six and twelve months after the event (Kersel et al., 2001). Nonetheless, even after recovery, patients are unlikely to regain their premorbid ability levels and continue to score worse than healthy people compared to normative data (Zaninotto et al., 2017).
In general, in this study, neuropsychological scores correlated more with MTR decline than with volume atrophy. In healthy individuals, studies have described a significant correlation between TB MTR and cortical GM MTR and performance on memory, executive function, and motor skills assessments (Seiler et al., 2014). Lower MTR in the normalappearing white matter has also been correlated with impairments in processing speed, executive function, and episodic memory in normal adults aged 50−90 years (Schiavone et al., 2009 The main limitations of this work include the relatively small sample size and the gender imbalance. The sample size was probably one of the main contributors for not finding a significant relationship between the time after trauma and MTR or volume decline, despite significant differences in the MTR and volume mean between time points 1 and 3. The gender imbalance is common in TBI studies since males are more frequently affected by TBI than females. However, studies investigating brain atrophy in TBI patients that used controls matched by gender found similar results to others that used unmatched controls (Ding et al., 2008;Farkas et al., 2007;Kim et al., 2008;Trivedi et al., 2007).
The inclusion of both moderate and severe TBI. Another limitation is inherent to MTI. Even though MTR is considered sensitive to alterations in the microstructural environment, its ability to reveal specific forms of underlying pathology requires further investigation, particularly in TBI, where multiple forms of pathology may co-exist (Pagani et al., 2008).
To our knowledge, this is the first longitudinal study to assess patients with DAI using MTR and volumetry. It is also the first one to obtain global MTR instead of in small ROI and to compare MTR and neuropsychological results in DAI. In addition, this work introduces a simple pre-processing methodology for automatic extraction of hemorrhagic content. Blood content is a significant inconvenience for many quantitative advanced MRI techniques, such as MTI and Diffuse Tensor Imaging (DTI). Therefore, excluding blood content should be a fundamental step in the processing pipeline to obtain reliable and reproducible results.

CONCLUSION
On a series of MRIs performed up to one year after DAI, this study shows gradual axonal/myelinic rarefaction and volume loss.
Despite a tendency toward worsening MRI quantitative measures, neuropsychological tests revealed that some patients had improved episodic verbal memory, attention, and executive function, suggesting that at least part of the morphometric changes result from a dynamic process in which reorganization prevails over the damage.