Myelin quantification using ultrashort echo time magnetization transfer ratio in a mouse model of traumatic brain injury

Abstract

Background and Purpose:

This study aims to assess the potential of ultrashort echo time imaging-based magnetization transfer ratio (UTE-MTR) in detecting demyelination in mice with mild traumatic brain injury (mTBI) caused by an open-field low-intensity blast (LIB) injury model.

Methods:

This study included 30 male C57BL/6 mice, approximately 8 weeks old, sourced from Jackson Laboratories in Bar Harbor, ME, and conducted under institutional guidelines. The mice were divided into the mTBI group (n=15) and the sham control group (n=15). All animal experiments followed the approved protocols for the Care and Use of Laboratory Animals and Animal Research. The mTBI group underwent the open-field LIB injury. Behavioral tests were conducted to assess motor activity and anxiety-like responses. UTE-MT imaging was performed using a 3T Bruker system to measure UTE-MTR from two UTE-MT datasets with saturation powers of 1500° and 500°, and two frequency offsets of 2 and 50 kHz, respectively. Luxol fast blue (LFB) staining was performed to evaluate myelin content. The mean UTE-MTR values for regions of interest centered at the medial section of the corpus callosum were computed. The behavioral tests, LFB myelin staining, and UTE-MTR values were compared between the two groups using the independent T-test. P values <0.05 were considered significant.

Results:

The mTBI mice demonstrated decreased motor activity and increased anxiety-like response over sham controls. The mTBI mice also showed significantly lower UTE-MTR values (0.399±0.007 vs. 0.393±0.005; P<0.05) and reduced LFB myelin staining (0.848±0.324 vs. 1.145±0.260; P = 0.048) over sham controls.

Conclusion:

The significantly lower UTE-MTR values in the corpus callosum of mTBI mice are consistent with reduced LFB myelin staining, indicating that UTE-MTR can detect myelin loss and associated alterations in motor and anxiety domains post-LIB exposure.

Introduction

Mild traumatic brain injury (mTBI) is characterized by an indirect or direct blow to the head through a nonpenetrating mechanism. It is subsequently followed by loss of consciousness and/or alterations to the mental state ^1,2. An estimated 42 million people throughout the world suffer from mTBI or concussion every year, and up to 1.5 million people in the US sustain a mTBI; young men are most frequently affected ^3,4. While most people recover totally from mTBI in 12 weeks or less, as high as 60% of people have post-concussive syndrome, which includes physical, cognitive, and emotional symptoms that cause damage lasting for months or years ⁵. Around 90% of individuals with treated traumatic brain injury (TBI) have mTBI ^6,7. Conventional magnetic resonance imaging (MRI) methods are frequently ineffective in identifying anomalies in cases of multiple neurodegenerative and neuroinflammatory illnesses, while being widely employed in the clinical diagnosis of these conditions ^8,9. Accurate diagnosis is difficult since most people with mTBI do not show typical imaging findings, such as hyperintense lesions.

Myelin is a vital structure that speeds up the transmission of neural signals in both the central nervous system (CNS) and the peripheral nervous system (PNS). Changes in myelin are strongly linked to neurological diseases including mTBI ¹⁰. Loss of the myelin sheath has been observed in many animal and human brain specimen studies of mTBI ^10–12. Therefore, quantitative imaging assessment of myelin can be a crucial diagnostic tool for evaluating mTBI. In conventional MRI techniques, signals are from intracellular and extracellular water, myelin water, and cerebrospinal fluid (CSF) ^13–15. Recently, more advanced imaging techniques like myelin water fraction mapping ¹⁶ and diffusion tensor imaging ¹⁷ have been proposed. However, these methods allow for indirect imaging of myelin and may have limited specificity to myelin changes. Direct imaging of ultrashort T2 components may have higher specificity than indirect imaging techniques. It also removes certain complications in the analysis from the intermediate effects occurring as a result of interaction between water and myelin ¹⁸. However, it is technically challenging to directly image myelin due to its fast signal decay. The semisolid myelin sheath has very fast transverse relaxation, with T2 or T2* on the order of hundreds of microseconds or less ^18–20.

Previous studies suggest that ultrashort echo time (UTE) type sequences with TEs of tens of μs can detect signal from myelin ^18–24. UTE imaging with long T₂ saturation has been proposed for direct myelin imaging ²¹. Adiabatic inversion recovery UTE (IR-UTE) has also been extensively investigated for more robust long T₂ suppression and selective myelin imaging ^{18,20,25–27}. Magnetization transfer is another option for myelin imaging. Magnetization transfer refers to the transfer of spin magnetization from the macromolecular protons to water protons, allowing for indirect assessment of macromolecules with restricted motion and extremely short T2 ^28,29. An effective and fast way to quantify the MT effect is to measure the magnetization transfer ratio (MTR). UTE-based MTR (UTE-MTR) potentially enables more efficient detection of myelin changes in the brain than conventional gradient echo (GRE)-based MTR ³⁰.

In this study, we evaluated the potential of UTE-MTR in detecting myelin changes in an open-field low-intensity blast (LIB) injury model of mTBI. Behavioral tests were conducted to assess motor activity and anxiety-like responses. Luxol fast blue (LFB) staining was performed to evaluate myelin content.

Methods

Open-field LIB injury model of mTBI

This study included 30 male C57BL/6 mice, approximately 8 weeks old, sourced from Jackson Laboratories in Bar Harbor, ME, and conducted in accordance with institutional guidelines. The mice were divided into the mTBI group (n=15) and the sham group (n=15). The researchers who were examining the experimental results were double-blinded in the investigation. All animal experiments followed the approved protocols for the Care and Use of Laboratory Animals and Animal Research. The mTBI group underwent a previously demonstrated reproducible open-field LIB murine model ³¹. The full protocol for the open-field blast model in mice is available in ³².

We did not evaluate for sex differences in our study design and have deliberately used only male mice since sex and gonadal hormones have been reported to influence neuronal behavior after TBI in mice, which indicates that the inclusion of both sex can increase the experimental variability ^33,34.

Figure 1 shows the experimental setup. Anesthetized mice were positioned prone, three meters away from a detonation of 350g of high-energy C4 explosive, which was 1 meter above the ground. The total reflected pressure was determined using a flush mount pressure sensor. The stand was made impervious to movement by using shock-absorbing padding. The incident pressure was measured using a free-field pressure sensor. The sedated mice were placed in animal cages in a prone posture, facing the explosive source with their head and body orientated longitudinally along the shock wave propagation direction. The mice were subjected to three blasts for three days in a row (one blast each day). The sham control mice received the same anesthetic treatments but were not exposed to LIBs. One week after the blast procedures, MRI scans were performed on the animals.

Figure 1.

Experimental setup for the mouse model of mTBI induced by the open-field low-intensity blast (LIB) exposure. A flush mount pressure sensor was used to measure the total reflected pressure. The stand was positioned 3 meters away from the LIB when the explosive was detonated. Shock-absorbing padding was used to prevent the stand from any movement. A free-field pressure sensor was used to measure the incident pressure.

Behavioral Study

An open field test was conducted to evaluate the exploratory behavior and anxiety of mice. It took into account the rodent’s aversion to brightly illuminated, large, and unknown environments ^35,36. Based on prior work, short-term behavioral deficits following LIB exposure were detected using an automated home cage monitoring system ³⁷. Behavioral tests were conducted on both the control and mTBI groups of mice. The tests began five days after the injury (5 DPI, counting from the last blast exposure). In the light phase, locomotor/exploratory activity and anxiety-like behavior were assessed in the open field arenas. The 40 by 40 cm open-field arena (Noldus Information Technology) was set up on a gray platform. The animals were kept in the middle of the square arena for ten minutes. After each trial, the box was cleaned with a 10% ethanol solution and allowed to air dry. With the use of EthoVision TX software (v14. Noldus Information Technology; Windows XP OS, Leesburg, Virginia, https://www.noldus.com/ethovision-xt), the movements of the animals were monitored and examined. To assess the activity tracks of each mouse and the remaining area of the arena, a 10 × 10 cm region was designated as the center zone. The overall length of time spent in the field’s various zones (the center vs. the periphery, as a gauge of anxiety-like reaction) and the total distance traveled were estimated.

In the evening 5 DPI, the LightSpot test was used to assess anxiety-like behaviors in response to unpleasant spotlight stimulation during the dark phase ³⁷. The test was also used to examine spontaneous behaviors in a home-cage setting. A yellow light-emitting diode light (2000 lux, similar to the light of a typical overcast day in the absence of heat generation) was introduced without causing specific pain or discomfort in order to investigate each mouse’s response to mildly unpleasant stimuli. Without any human involvement, the LightSpot test adhered to pre-calibrated experimental methods. The sham controls and the mTBI mice were housed in PhenoTyper home cages (Model 3000; Noldus Information Technology, Wageningen, The Netherlands) to measure their spontaneous behaviors ^38,39. The PhenoTyper home cages (L = 30 × W = 30 × H = 35 cm) consisted of four semitransparent Perspex walls, an opaque Perspex floor, a shelter, food and water stations, and a top control unit. Each mouse was given its own bedding (Bed-o’Cobs, Laboratory Animal Bedding; The Andersons Laboratory Animal Bedding, Maumee, OH), as well as unlimited access to food and drink. EthoVision XT software (v14; Noldus Information Technology) managed a round-the-clock, infrared-sensitive automated video-tracking system that captured mouse actions at a sampling rate of 15 frames per second (FPS).

After the acquisition of data in the dark period, the LightSpot program began at 7:00 PM and an average spotlight ranged for one hour, from 7:15 PM to 8:15 PM. For meta-analysis, data were uploaded to the Sylics, Bilthoven, The Netherlands, web-based AHCODA-DB. The movements were divided into two categories: lengthy movements, like moving from one place to another, and non-lengthy movements, like twisting or rearing against the wall. When a lengthy arrest segment or a shelter stay that went above the brief shelter visit threshold occurred, activity bouts ended.

MRI Data Acquisition

Following the behavioral testing, the mice were taken to the VA San Diego Healthcare System’s Radiology Service unit for MRI with an average DPI of approximately 10.8±1.5, ranging from 9 to 13. Mice were euthanized, and their brains were extracted for MRI. The UTE-MT sequence was implemented on a 3T Bruker system equipped with a gradient strength of 450 mT/m strength and a slew rate of 4200 T/m/s. The mice brains were individually scanned using a 10 mm surface coil for signal reception and an 82 mm volume coil for transmission. The mice were sacrificed, and their brains were scanned right after extraction.

The following scan parameters were used: TR = 80 ms, TE = 0.026 ms, flip angle (FA) = 8°, number of spokes per MT preparation pulse = 13, bandwidth = 35 kHz, slice thickness = 0.43 mm, number of slices = 84, FOV = 11×11 mm², matrix size = 84×84, acquired voxel size = 0.13 × 0.13 × 0.43 mm³, MT power = 1500°,500°, two frequency offsets of 2 and 50 kHz, scan time of 4 min 30 sec for each frequency offset, total scan time of 9 min.

Due to the time-sensitive nature of the brain tissue samples and the risk of degradation with prolonged scans, additional averages were not feasible. Instead, the protocol was optimized to balance signal to noise ratio (SNR), resolution, and acquisition time, achieving adequate imaging quality without compromising sample integrity. Additionally, denoising techniques were applied, successfully reducing noise and further enhancing image quality.

For the repeatability assessment, a single mouse brain sample was scanned twice with repositioning.

MRI Data Analysis

The UTE data processing code was written in MATLAB R2021a (MathWorks, Natick, MA, https://www.mathworks.com/products/matlab.html). The region of interest (ROI) was outlined on the medial section of the corpus callosum (MCC). As myelin alterations have been noted in the medial corpus callosum (MCC) region in earlier research⁴⁰ this region was chosen for MRI data processing. Because of its comparatively larger structure, this region can also yield more accurate MTR measurements, lowering error caused by the partial volume effect.

The mean UTE-MTR values for the delineated ROI were computed for the two groups and were compared using the independent T-test. P values <0.05 were considered significant. MTR was calculated by the following equation :

$$MTR=\frac{{MT}_{off}-{MT}_{on}}{{MT}_{off}}$$ (1)

${\mathrm{M}\mathrm{T}}_{\mathrm{o}\mathrm{n}}$ and ${\mathrm{M}\mathrm{T}}_{\mathrm{o}\mathrm{f}\mathrm{f}}$ represent the UTE scans with MT saturation pulse power on and off, respectively. In this study, ${\mathrm{M}\mathrm{T}}_{\mathrm{o}\mathrm{n}}$ was chosen from the UTE-MT acquisition with MT pulse FA of 1500° and frequency offset of 2 kHz. ${\mathrm{M}\mathrm{T}}_{\mathrm{o}\mathrm{f}\mathrm{f}}$ data was selected from the UT-MT acquisition with MT pulse FA of 500° and frequency offset of 50 kHz, which has almost no MT effect-induced signal loss owing to the very high frequency offset. The UTE-MT images were very noisy due to the relatively short scan time and high spatial resolution needed to depict the small MCC region in the mouse brain. To address this limitation, a tensor-based Marchenko-Pastur principal component analysis was carried out to denoise the UTE-MT images ⁴¹. The tensor-based denoising extends the Marchenko-Pastur principal component analysis method to multidimensional data. While the former method is matrix-based, the latter method utilizing tensors can accommodate higher dimensional data. This strategy leverages redundancy in the 3D data and significantly improves the SNR, which was calculated as the ratio of the mean signal intensity inside a user-defined region to the standard deviation of the background noise.

Histological Analysis

After the MRI acquisition, the brain samples were immediately preserved in zinc-formalin for 48 hours at 4 °C. The brains were sectioned coronally at a thickness of 5 μm after being embedded in paraffin. Three slides were stained with LFB overnight at 60°C, followed by a quick differentiation step with 0.05% lithium carbonate. Slides were collected at the Bregma −1.0 mm to −1.5 mm level at intervals of 50 μm. Consecutive sections of the same sham brain were stained as controls for each batch to guarantee staining consistency across brains. Using HALO AI software (v3.6.4134, Indica Labs, Albuquerque, NM, USA, https://indicalab.com/halo-ai/halo-ai-apps/) the average optical density of the entire corpus callosum was measured.

We employed the following procedures to enable accurate optical density measurements across samples to compensate for variability from slide to slide (or batch to batch):

1. A healthy mouse (mouse No. 15) was selected as the control. Consecutive sections (n = 20) were prepared at the plane of Bregma −1.0 mm with a slice thickness of 5 μm. We assume the same myelin density for all 20 sections, which spans a whole thickness of 20×5 = 100 μm. In other words, the average myelin density in the MCC region of a healthy mouse remains consistent across 0.1 mm thickness, which is a reasonable assumption.

2. One section from this sample was used as a control slide for each batch of staining. Again, the average optical density (AOD) for each of the 20 control slides is assumed to be consistent, as they are from the same healthy brain region spanning 0.1 mm in thickness. Acknowledging potential inter-batch variations, even by the same operator, we calculated a ratio of the AOD of No. 15 samples stained across different batches compared to the value obtained in the first batch. During analysis, all values from other batches were divided by this ratio, allowing comparison across batches.

3. To reduce operational variability, we mounted the brain slides in approximately the same position on glass slides for consistent staining. Additionally, we strictly controlled the staining time in the LFB solution and the differentiation time in lithium carbonate using a timer, minimizing inconsistency across different batches.

Statistical Analysis

All statistical analyses was carried out using Prism software (GraphPad Software Inc., La Jolla, CA,https://www.graphpad.com/). The normality of the dataset was tested using the Kolmogorov–Smirnov test. The UTE-MTR values in the MCC were compared between the control and mTBI groups. The MCC region was chosen since myelin alterations have been documented in this location in prior research ⁴⁰.

An independent t-test (normal distribution) was used to examine the UTE-MTR differences between the two groups. P-values less than 0.05 were considered statistically significant for all variables.

For the behavioral data analysis, an independent t-test and two-way ANOVA were employed to compare the mTBI and control groups for velocity, movements, and time spent in the center during the open field test, as well as for time spent in the shelter and outside the shelter during the LightSpot test.

For the LFB analysis, statistical comparisons (an independent t-test) were performed specifically between the mTBI and sham groups to evaluate differences in optical density values. This approach was chosen to quantify the extent of myelin loss in the mTBI model, a key histological feature of interest.

Results

Figure 2 shows the behavioral assessment carried out on the mTBI and control mice. General behavioral tests and the mTBI LightSpot for the control and mTBI mice were observed from the home cage monitoring system. While there was no significant difference when the mice were placed in the spotlight for the first hour and a half, the mTBI mice spent more time within the shelter and less time outside, after the lighting phase. Between 10 PM and 1 AM, mTBI mice showed a preference for hiding, even in a dark setting. When the amount of time spent in (A) the shelter, (B) outside, and (C) in the feeder was plotted versus time, a 2-way ANOVA assessed for the period of 10 PM to 1 AM (blue shaded box), yielded a P value less than 0.05, showing a significant difference between the sham control and the mTBI mice.

Figure 2.

General behavioral tests and mTBI lightspot observed in the home cage monitoring system for the control and mTBI mice.. The time spent in the shelter (A), time spent outside (B), and time spent in the feeder (C) were plotted and a 2-way ANOVA was evaluated for the duration of 10PM to 1AM (blue shaded box). P<0.05 was considered significant.

Figure 3 shows representative UTE-MTR images for a control mouse. MT_on image acquired with FA of 1500° and power of 2 kHz is shown in column (A). The MT_off images were acquired with FA of 500° and power of 50 kHz as shown in column (B). The rows present the images before and after pre-processing using the tensor-based denoising algorithm ⁴¹. The SNR values for the UTE-MT_on images before and after denoising were 25.00±11.12 and 51.92±16.71, respectively, corresponding to a 2-fold increase in SNR. Near identical SNR enhancement was observed for the UTE-MT_off images. An average SNR improvement of 2.75-fold was achieved after applying the denoising filter. A mean SNR averaged across all the mice after denoising was observed as 88.61±12.21. Column (C) provides the UTE-MTR maps where the MCC region has a stronger signal indicative of higher myelin content, as shown by the yellow dashed box.

Figure 3.

UTE-MT imaging of a control mouse brain. Representative image slice of UTE-MT data with MT_on (A), MT_off (B), and the corresponding MTR map (C) before (1^st row) and after (2^nd row) denoising using the tensor-based Marchenko-Pastur principal component analysis. The region of interest, MCC is highlighted in the yellow dashed box and presents higher intensity signal than other brain regions, reflecting higher myelin content in MCC. UTE = Ultrashort echo time, MCC = Medial corpus callosum

Figure 4 shows representative images from a mTBI mouse. The MT_on images before and after the denoising process are presented in (A), while the MT_off images are shown in (B). The denoising process utilizes the redundancy of multidimensional data, thereby yielding substantial enhancement in the SNR as shown by the UTE-MT maps in (C). The SNR values for the UTE-MT_on images before and after denoising were 29.27±4.88 and 72.24±14.49, respectively. A nearly identical trend was observed for the UTE-MT_off images. An average SNR improvement of 3-fold was achieved using the denoising filter. A mean SNR averaged across all the mice after denoising was observed as 91.68±10.21. The quantitative values evaluated from the UTE-MTR maps were lower compared to the control mouse brain for the MCC region. This indicates that the mTBI mouse experienced sufficient myelin loss following trauma.

Figure 4.

UTE-MT imaging of an mTBI mouse brain. Representative image slices of UTE-MT data with MT_on (A), MT_off (B), and the corresponding MTR map (C) before (1^st row) and after (2^nd row) denoising with the tensor-based Marchenko-Pastur principal component analysis. The Medial corpus callosum region, marked by the yellow dashed box shows reduced signal intensity in contrast to the control mouse brain indicating myelin loss in the mTBI mouse.UTE = Ultrashort echo time

Figure 5 provides representative UTE-MTR maps (A,B) obtained from the brains of 8-week-old male control and mTBI mice, respectively. The MCC region, where the ROI was centered to get the mean MTR, is highlighted by the bounding box. The average and standard deviation of UTE-MTR values for the ROI are provided in (C). Significant differences (0.399±0.007 vs. 0.393±0.005; P<0.05) in MTR values were found between the control and mTBI mice as shown in (D). The MTR value in the mTBI brain was lower than that of the control brain, suggesting a reduced myelin content. To address reproducibility, we conducted a test where the brain sample was scanned twice with repositioning, measuring the ROI values for two repeated scans. We found the error to be 0.5%, corresponding to a difference of 0.001 between the two repeats.

Figure 5.

Representative UTE-MTR maps obtained from 8-week male control (A) and mTBI mouse brains (B). The bounding box highlights the mid-section of the corpus callosum, where the region of interest was centered to calculate the mean MTR. The mean MTR value is significantly lower for the mTBI group than the control group, indicating reduced myelin content in mTBI brains as shown in (C). The average, standard deviation, and statistical significance (Independent T-test) were computed using the MTR value between the control and mTBI groups as plotted in (D).

“*” in panel D indicates a P-value lower than 0.05.

UTE = Ultrashort echo time

Figure 6 displays the LFB staining results from a representative sham control (A) and an mTBI mouse (B), respectively. The averaged optical densities evaluated in the MCC region for the two groups are shown in (C). The mean optical density for the sham control mice (1.145±0.26) is significantly higher than that for the mTBI mice (0.848±0.324) with P = 0.048. These results demonstrate significant myelin loss in mTBI mice due to the open field LIB injury.

Figure 6.

Representative LFB stained images from a control mouse (A) and an mTBI mouse (B), along with bar plots of the summarized optical density measurements from 15 controls and 15 mTBI mice evaluated in the MCC region (shown in dashed bounding box). The mean LFB optical density in the MCC region for the sham control mice is significantly higher than that of the mTBI mice (1.145±0.26 vs. 0.848±0.324; P =0.048).

“*” in panel C indicates a P-value lower than 0.05.

LFB = Luxol fast blue, MCC = Medial corpus callosum, μm = micro meters, a.u = arbitrary unit

The motor and anxiety measures were included in the study because behavioral changes are well-documented outcomes in mTBI models, including the LIB model. These measures are critical for characterizing the functional consequences of mTBI and complement the histological and imaging findings by providing a broader understanding of the injury’s impact. While these measures were not explicitly rationalized in the context of imaging, their inclusion aligns with the comprehensive evaluation of mTBI-related deficits and reflects the multifaceted nature of this condition.

Discussion

In this study, we utilized the UTE-MTR biomarker to evaluate myelin alterations in mice subjected to mTBI. The complete dataset comprised fifteen mice exposed to open-field LIB to induce mTBI, and fifteen mice that constituted controls without any LIB exposure. Behavioral assessments were conducted prior to the MRI scanning. Acute reductions in motor activity and elevated anxiety levels were observed in the mTBI group as opposed to the control group. The significantly lower UTE-MTR values in the corpus callosum of mTBI mice indicate the demyelination process associated with the open-field LIB injury, as confirmed by LFB myelin staining. The results show that UTE-MTR can be used as a potential biomarker for measuring myelin content in the corpus callosum, as well as for characterizing myelin damage in mTBI.

In the open-field test, the mTBI mice showed a much lower tendency to move while spending time in the center compared to the control mice. There was no change in the overall distance since the observed movement was faster, which is quite different from that noted in the single blast mTBI mice⁴⁰. We also observed that the mTBI mice tend to spend more time in the shelter portions of their life cycle, exhibiting less mobility in the center during the light and dark phases. After a modest level of explosive blast at approximately 46.6 kP, the mTBI mice group experienced acutely decreased motor activity, increased interest in exploring the environment and higher anxiety levels. These findings suggest acute structural changes due to the mTBI-induced alterations that are associated with post-injury fear and anxiety states^2,42.

Evaluation of UTE-MTR and its relation to the biophysical phenomenon can be attributed to the potential of UTE in direct estimation of signals from protons in the myelin phospholipid membranes. Since other sources of signals such as intra- and extra-cellular water have much higher T2 (>50 ms) ¹⁵, they should not induce any significant changes in myelin quantification using MTR. Direct imaging of semisolid lipids such as myelin has been shown to provide noninvasive biomarkers of neurodegenerative diseases ^27,43,44. Examining the efficacy of UTE-MTR is crucial in evaluating myelin changes in mTBI and other neurological conditions.

A recent study on the longitudinal evaluation of demyelination in MS mouse models used quantitative MTR as one of the biomarkers ³⁰. The study was conducted on a clinical 7T scanner. It was shown that, in the cortical areas, only UTE-MTR was able to achieve significant differences and provide the estimation of bound proton fraction, a validated metric proven to detect myelin content changes in cortical areas of mouse brains.

Cortical demyelination has been associated with cognitive decline and the advancement of various neurodegenerative diseases, although conventional MRI cannot detect it ¹⁰. White matter atrophy clearly shows that patients with moderate-to-severe TBI who survive a chronic post-injury stage have substantial white matter involvement ⁴⁵. Henceforth, it is imperative to create new quantification techniques to detect and identify atrophied white matter regions. Only very few imaging studies have been able to detect corticular alterations ^46,47. Monitoring demyelination using magnetization transfer imaging in an MS mouse model has been studied previously ⁴⁸. The field strength used for the study is 9.4T which is clinically not yet feasible. The data revealed that the MTR significantly decreased during demyelination, which is consistent with our results. Another study using human postmortem tissues showed that myelin content had an impact on MTR ⁴⁹. These findings suggest that MTR is a reliable biomarker for the myelination process.

We did not perform a direct quantitative slide-by-slide comparison between UTE-MTR values and histological optical density in this study. Instead, we only did group analysis, comparing UTE-MTR and optical density values between the mTBI and control groups. There are two reasons behind this strategy. First, the number of samples is too small. We only have 15 mice in the mTBI group, and 15 in the sham control group. Second, the changes in UTE-MTR are very small, on the order of ~0.006, or 1.5%. The small changes in UTE-MTR are likely due to little changes in myelin content associated with the mTBI model, where the open field low-intensity blast resulted in minor myelin loss, as demonstrated by prior studies^31,32. A larger group size is required to robustly detect slide-by-slide myelin changes. Considering our small sample size and the mTBI model, we employed a group analysis, which yielded statistically significant differences in UTE-MTR values and average optical density in the MCC regions. We are conducting a similar study with a much larger group size (n=40 per group) and a model that might produce greater changes in myelin. We plan to conduct a slide-by-slide correlation analysis between UTE-MTR values and LFB optical density values. We also explored the possibility of using internal normalization staining. However, no suitable method could be identified, as the AOD of other brain regions may vary due to technical operations or treatments, such as blasting. These potential variations limit the feasibility of internal normalization for this study.

There are several limitations to this study. First, the sample size of acquired mouse brains was relatively small, which could impact the statistical assessment, particularly the behavioral studies. Second, only ex vivo study was performed to minimize scan time. UTE-MT imaging of the mouse brain is subject to low SNR due to the requirement of high spatial resolution at a relatively low field strength of 3T. Much improved SNR was achieved for UTE-MT images by removing the skull and other body parts of the mouse. Even so, the final UTE-MT images still suffered from low SNR and required the use of tensor-based Marchenko-Pastur principal component analysis to improve the final image quality (i.e., SNR improved by a factor of 2). Third, UTE-MT modeling could provide more comprehensive information including macromolecular proton fraction and exchange rates, but was not investigated in this study due to the requirement of significantly longer scan time. Fourth, our findings imply that the open-field LIB damage model caused a slight loss of myelin. This mouse model would be the subject of a long-term investigation to shed light on demyelination and remyelination in mTBI and whether the UTE-MTR measurement can identify both. Fifth, it lacks direct applicability to civilian mTBI conditions, as the repeated blast exposures do not reflect typical scenarios like sports-related concussions where re-exposure is usually prevented.​ Future research should consider models that incorporate longer intervals between injuries to better simulate civilian circumstances.

Our results indicate that UTE-MTR can detect alterations of myelin composition in white matter of the mouse brain. Significant differences were observed between the mTBI and control mouse brains. mTBI may damage myelin, impairing cognitive functions. The UTE-MTR maps provide a novel assessment of myelin content, representing a significant advancement in the diagnosis and treatment monitoring of mTBI.