Diffusion tensor imaging of acute mild traumatic brain injury in adolescents

Mild traumatic brain injury (MTBI) is defined as brief loss or alteration of consciousness (i.e., <30 minutes) and post-traumatic amnesia (PTA) of short duration (i.e., <24 hours) following either blunt head trauma or an acceleration/deceleration of the freely moving head and is frequently associated with an absence of intracranial abnormality on head CT performed in the Emergency Department (ED).¹ Adolescents are frequently affected by MTBI as the incidence of TBI-related ED visits is 661.1 per 100,000 in the 15- to 19-year age range.² This is likely an underestimate of the true incidence of MTBI in this age range as an unknown number of cases never present to an ED for treatment.

In spite of normal head CT findings, reports of postconcussion symptoms, including cognitive (e.g., poor memory and concentration), somatic (e.g., headache, dizziness), and affective (e.g., depression, irritability) symptoms, are common following MTBI.³ Although these symptoms can resolve within weeks to a few months, they persist in a minority of cases,⁴ diminishing quality of life for these patients.^5,6 Although neuropathology studies of MTBI in fatal noncranial injuries have implicated a continuum of traumatic axonal injury (TAI) in regions such as the corpus callosum (CC), internal capsule, and cerebral peduncles,^7,8 and neuropsychological studies have demonstrated reduced interhemispheric transfer of information,⁹ conventional MRI has generally failed to reveal evidence of TAI in patients with MTBI. This dissociation during the initial weeks and months post-MTBI between relatively normal conventional imaging results in patients who exhibit cognitive dysfunction and postconcussion symptoms has remained enigmatic for clinicians and investigators.

Diffusion tensor imaging (DTI) is an imaging technique acquired on a standard MRI scanner that has been shown to be far more sensitive to white matter injury than conventional MRI.^10,11 DTI, which reflects myelin integrity in vivo, is based on the characteristic of myelin sheaths and cell membranes of white matter tracts that restricts the movement of water molecules. Consequently, water molecules tend to move faster along nerve fibers rather than perpendicular to them. This characteristic, which is referred to as anisotropic diffusion and is most commonly measured by fractional anisotropy (FA), is determined by several factors including the thickness of the myelin sheath and of the axons, and the distribution of directions, and density of white matter fiber tracts. FA ranges from 0 to 1, where 0 represents completely isotropic diffusion (free diffusion) and 1 represents the most anisotropic diffusion, i.e., diffusion restricted to one direction. In addition, radial diffusivity (RD) denotes the extent of diffusion perpendicular to the direction of maximal diffusivity, which presumably includes diffusion through intracellular and extracellular space perpendicular to the predominant orientation of the axons. The apparent diffusion coefficient (ADC) is the overall average measure of diffusion. FA and ADC are considered proxies for white matter integrity.¹²

Pathologic processes that alter the microstructure such as loss or disorganization of fibers associated with breakdown of myelin and downstream nerve terminals,¹³ neuronal swelling or shrinkage, and increased or decreased extracellular space, could affect both diffusion and anisotropy. An initial DTI study of five adults with MTBI who were scanned within 24 hours postinjury indicated that FA was reduced in normal-appearing white matter, including regions highly susceptible to shearing effects of TAI such as anterior and posterior CC, and the anterior and posterior limbs of the internal capsule.¹⁰ These findings were corroborated by another cross-sectional DTI study of 20 patients with acute and 26 patients with chronic MTBI using region of interest (ROI) analysis.¹⁴ Reductions in FA have also been confirmed in the CC as long as 5 years post-TBI of varying severity.^14,15 Validity of DTI in adult TBI has been supported by a positive correlation of FA in the internal capsule and splenium with the Glasgow Coma Scale (GCS)¹⁶ score, i.e., patients with less severe TBI had higher FA, reflecting preservation of white matter.¹⁷

Most previous DTI studies of MTBI have not excluded patients with lesions on CT, thus complicating the impact of disparate pathologies on DTI results. Concurrent assessment of postconcussion symptoms has also been neglected in previous DTI studies. The time from injury to imaging has also been inconsistent across DTI studies, thus also complicating interpretation of the findings. Consequently, we studied changes in white matter integrity using DTI in relation to neurobehavioral findings of patients with MTBI.

Because of the known vulnerability of the CC in TBI, we sought to determine if FA, ADC, and RD were altered in the CC in the acute phase of MTBI in adolescents with GCS scores of 15 and negative CT findings. DTI measures were obtained within 6 days postinjury and analyzed in relation to postconcussion symptoms and general emotional distress. We hypothesized that DTI abnormalities would be detected in the acute phase of injury and related to the severity of postconcussion symptoms.

METHODS

Imaging acquisition.

Nine of the 10 adolescents with MTBI underwent CT scanning during hospital admission on a 16-slice General Electric Lightspeed scanner (Waukesha, WI), using a standard clinical protocol with 5 mm contiguous axial slices covering the skull base to the vertex, performed without contrast at Texas Children's Hospital, Houston. One child underwent CT scanning at an outside facility, and the written report was obtained. In all cases, scans were reviewed by a board-certified neuroradiologist, and no abnormalities were reported for any subject. All subjects in both groups underwent MRI without sedation using a sensitivity encoding (SENSE) 8-channel head coil on a Philips 3.0T Achieva scanner (Philips, Cleveland, OH) at Texas Children's Hospital.

Conventional MR imaging.

A coronal T2-weighted fluid attenuated inversion recovery sequence was used (11,000 msec repetition time [TR], 120 msec echo time [TE], 2800 msec inversion time [TI], 5.0 mm slices, without SENSE) to detect any areas of abnormality in all subjects. For this sequence, a 230 mm field of view (FOV) was used with an acquired voxel size of 0.9 × 1.25 × 5.0 mm and a reconstructed voxel size of 0.45 × 0.45 × 5.0 mm. Axial fast field echo (FFE) was also employed to detect any blood products (847 msec TR, 16 msec TE, 4.0 mm slices, 0.0 mm gap, SENSE factor = 2). For this sequence, a 230 mm FOV was used with an acquired voxel size of 0.9 × 1.13 × 4.0 mm and a reconstructed voxel size of 0.45 × 0.45 × 4.0 mm. T1-weighted and T2-weighted images were also acquired in both patient groups. Parameters for the three-dimensional turbo field echo T1-weighted sequence included a TR = 0.9 msec, TE = 3.1 msec, 1.0 mm slice thickness, FOV = 256 mm, acquired and reconstructed voxel size of 1.0 × 1.0 × 1.0 mm, SENSE factor = 2. Parameters for the three-dimensional turbo spin echo T2-weighted sequence included a TR = 5,000 msec, TE = 80 msec, 1.5 mm slice thickness, FOV = 256 mm, acquired voxel size of 1.0 × 1.0 × 1.0 mm and a reconstructed voxel size of 1.0 × 1.0 × 1.5 mm, SENSE factor = 2. One subject in the MTBI group was noted to have signal dropout with surrounding hyperintensity on the T2-weighted spin echo imaging which was interpreted as being consistent with hemorrhagic contusion with surrounding edema affecting the left greater than right frontal orbital region. Additional abnormalities interpreted as small areas of hemorrhagic contusion were noted on the gradient echo imaging in the fronto-orbital (left greater than right), left anterior temporal pole, and right posterior medial parieto-occipital region. No other subject in either group was noted to have trauma-related abnormalities on any of the conventional imaging sequences.

DTI acquisition.

Transverse multislice spin echo, single shot, echoplanar imaging (EPI) sequences were used (6,318.0 msec TR, 51 msec TE, 2.0 mm slices, 0 mm gap). A 224 mm FOV (RFOV = 100%) was used with a measured voxel size of 2.0 × 2.0 × 2.0 mm and a reconstructed voxel size of 1.75 × 1.75 × 2.0 mm, and a SENSE factor of 2. Diffusion was measured along 30 directions using an electrostatic gradient model²⁰ (number of b-value = 2, low b-value = 0, and high b-value = 1,000 sec/mm²). To improve signal to noise ratio, two acquisitions were taken and averaged using Philips diffusion affine registration tool.²¹ Each acquisition took approximately 4 minutes 50 seconds, and 70 slices were acquired.

Susceptibility weighted imaging (SWI) acquisition.

All subjects underwent SWI using a three-dimensional FFE sequence (50 msec TR, 40 msec TE, 2.0 mm slices, 0 mm gap, 64 or more slices depending on the size of the head, SENSE factor = 2). A 256 mm FOV was used with an acquired voxel size of 0.375 × 0.985 × 2.0 mm and a reconstructed voxel size of 0.5 × 0.5 × 2.0 mm. Post-processing of the images was performed using SPIN software (The MRI Institute, Detroit, MI) to enhance the signal loss caused by microscopic hemorrhage by multiplying the original magnitude contrast images by a normalized phase mask resulting in the final SWI images.²² The patient with MTBI with abnormalities on conventional imaging was noted to have a small area of abnormality in the right posterior medial region on SWI, which was interpreted as an additional hemorrhagic lesion. No other subject demonstrated abnormal findings on SWI.

DTI imaging analysis.

Based upon our previously published DTI quantitative tractographic methods using Philips PRIDE software v4.0,²³ we calculated FA, ADC, and RD in the total CC (figure 1). To examine intrarater agreement, each case was measured twice; intraclass correlation coefficients (ICC) exceeded 0.98 for all DTI indices. Inter-rater agreement was also assessed by measurement of the CC by two different raters in half of the cases in both groups; ICCs again exceeded 0.98.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Diffusion tensor imaging (DTI) tractography illustrating the commissural fibers coursing through the corpus callosum in an uninjured control subject

Red indicates fibers oriented in a right to left direction, green reflects fibers oriented in an anterior-posterior direction, and blue reflects fibers coursing in a superior-inferior orientation.

RESULTS

Imaging.

Analyses revealed significant group differences for the whole CC on measures of FA (t¹⁸ = −2.24, p = 0.038), ADC (t¹⁸ = 3.09, p = 0.006), and RD (t¹⁸ = 2.90, p = 0.010). The MTBI group demonstrated increased FA and decreased ADC and RD relative to the control group. Figure 2 illustrates the relation of DTI indices FA and ADC in the CC for each patient with MTBI relative to an age- and gender-matched control subject.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Bar graphs illustrating the relation of diffusion tensor imaging indices fractional anisotropy (FA) and apparent diffusion coefficient (ADC) in the corpus callosum for each subject with mild traumatic brain injury (TBI) relative to an age- and gender-matched control subject

Increased FA (A) and decreased ADC (B) and radial diffusivity (not shown) were evident in 8 of 10 pairs, indicating the frequency of this pattern.

Relation between neuroimaging indices and behavioral data.

Spearman rho correlations were computed between DTI measures of FA, ADC, and RD of the total CC with postconcussion symptoms (RPCSQ) and general emotional distress (BSI). In the control group, FA did not correlate with postconcussion symptom severity on the RPCSQ (rho = −0.04, p = 0.90) or level of general emotional distress on the BSI (rho = 0.28, p = 0.43). Only a trend was noted for correlation between ADC and severity of postconcussion symptoms (rho = 0.57, p = 0.08), and no significant correlations between the ADC and emotional distress were found. No significant correlations were found between the RD and the level of postconcussion symptom severity (rho = 0.17, p = 0.63) or general emotional distress (rho = −0.09, p = 0.78).

However, in the MTBI group, FA of the total CC correlated with the severity of postconcussion symptoms (rho = 0.76, p = 0.01), and general emotional distress (rho = 0.76, p = 0.01). There were trends noted for correlations between ADC and postconcussion symptom severity (rho = −0.60, p = 0.07) and general emotional distress (rho = −0.57, p = 0.08). RD was correlated with postconcussion symptom severity (rho = −0.78, p = 0.008) and general emotional distress (rho = −0.75, p = 0.02). In each of these analyses, increased FA and decreased ADC and RD in the MTBI group were associated with greater symptom severity. Figure 3 illustrates the strong relation between FA and scores on the RPCSQ and BSI in the MTBI group.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Scatterplots illustrating the relation between FA and scores on the RPCSQ (A), and BSI (B) for both the MTBI group and the uninjured control group

FA = fractional anisotropy; RPCSQ = Rivermead Post Concussion Symptoms Questionnaire; BSI = Brief Symptom Inventory; ADC = apparent diffusion coefficient; MTBI = mild traumatic brain injury.

DISCUSSION

Our preliminary DTI findings show that despite normal CT findings, unremarkable conventional MRI findings in all but one patient, and a GCS score of 15, adolescents who sustained MTBI had increased FA and decreased diffusivity in the CC within 6 days postinjury. Cognitive, affective, and somatic postconcussion symptoms were all related to DTI indices of CC integrity including FA, ADC, and RD. Consequently, we infer that the DTI indices were sensitive to pathologic processes of MTBI that contributed to the postconcussion symptom severity of our patients.

In contrast to previous DTI studies of small samples of adult MTBI^10,14 which reported reduced FA and normal diffusivity as early as within 24 hours postinjury,¹⁰ we found that FA was increased and diffusivity (ADC and RD) was reduced within 6 days after injury. In comparison with our adolescent sample whose GCS scores were 15, the patients with MTBI studied in previous reports were older (mean age nearly 36 years) and had GCS scores which ranged between 13 and 15 indicating more variation in severity of impaired consciousness.¹⁰ Consequently, other pathologic processes such as TAI might have been more prominent in the patients in these studies. In addition, our patients were generally scanned more than 24 hours (but less than 1 week) postinjury, at an interval where different injury mechanisms may have been occurring.

Our finding of a reduction in ADC likely reflects trauma-induced cytotoxic edema and inflammation,^27,28 transient effects that occur during the subacute period after TBI.²⁸ Synthesis of the TBI and DTI literature suggests that cytotoxic edema and inflammation may result from ion homeostasis failure and membrane dysfunction. Acute and subacute reduction in ADC using diffusion weighted imaging in human studies have also been reported,^28,29 though these studies have typically focused on patients with moderate to severe TBI. In other causes of inflammatory reaction,²⁹ reduced ADC is considered an indication of cytotoxic edema.^30,31

Though previous reports using DTI in TBI have generally revealed decreased FA and increased measures of diffusivity including ADC and RD which were ascribed to diffuse degenerative changes such as Wallerian degeneration, axonal collapse, and myelin degeneration,^{10,14,17,23,32,33} these reports have generally examined or included more severe cases of TBI after longer postinjury intervals than our study. We found that FA was increased and ADC and RD were decreased within 6 days postinjury, findings we attribute to subtle cytotoxic edema associated with compressed intracellular space between CC fibers, thus restricting diffusion in a more uniform direction. Our findings of increased FA are consistent with a single case report of increased anisotropy in an infant with possible nonaccidental injury imaged with DTI within 24 hours after TBI which preceded the development of overt cerebral swelling.³⁴ In nontraumatic disorders including acute cerebral ischemia or infarction, elevated FA and decreased ADC have been reported to reflect basic alterations in cerebral microstructure.^35,36 Furthermore, in animal models of secondary trauma-related ischemia, cytotoxic edema was related to decreased ADC values in the postacute time period, presumably related to disruption of energy metabolism and failure of membrane pumps.³⁷

The time course of cytotoxic edema after TBI has not been definitively established, though human models have suggested that cytotoxic edema may reach maximal levels at 24 to 48 hours postinjury.³⁸ In our study, the pattern of increased FA and reduced measures of diffusivity was present in 8 of 10 adolescents with MTBI relative to their matched normal control subjects. Interestingly, the two patients who had the opposite pattern of reduced FA and increased diffusivity had the longest intervals between injury and scanning (4 and 6 days, respectively). It is possible that these early findings of increased FA and decreased diffusivity are evident only in the first few days following injury, but further investigation into the mechanism of injury and time course related to these findings is necessary.

We focused on the CC in this study because of its vulnerability to TAI³⁹ and robustness in measurement of both FA and diffusivity in this structure^11,23 as the largest white matter fasciculus. In a detailed biomechanical examination of 22 sports concussions in National Football League players using real-time image reconstruction of the brain to model brain deformation, the CC was observed to be one of the structures to receive the highest strain concentration, and the degree of strain concentration in the CC was associated with removal from play.⁴⁰ Consistent with the vulnerability of the CC, MRI investigation of pediatric TBI encompassing a wide range of severity⁴¹ has demonstrated a high frequency of CC lesions, especially in the splenium. In more severe injury, postmortem studies have consistently demonstrated the CC as a structure most frequently damaged, with early axonal damage noted as early as 35 minutes postinjury.⁴²

In the acute phase of injury there are a multitude of pathologic effects, some of which are transient in nature in milder head injury. These pathologic processes include disruption of axonal filaments that provide axonal ultrastructure, disruption of myelin integrity, excitotoxic biochemical reactions, localized inflammatory responses including macrophage and cytokine activation, and edematous reactions.^13,43 All of these pathologic responses may follow different time courses of expression, with some leading to permanent changes in cell function or actual neuronal death. How each of these neuropathologic factors independently or in combination relates to DTI findings is incompletely understood at this time. Furthermore, the extent of the impact of each of these factors in the immature brain vs the mature brain still warrants further investigation.

The primary limitations of the current study include the small number of subjects, the cross-sectional design, and the absence of an additional later follow-up assessment. Despite the balance between the groups on age, gender, ethnicity, premorbid intellectual functioning estimate, and education, replication of these findings in a larger sample is required. In addition, histopathology in experimental studies is required to confirm that cytotoxic edema and inflammatory responses are mechanisms of injury that account for these DTI findings. Future studies should examine additional brain regions and determine whether there are areas of the brain more prone to the injury apparent using these DTI indices. This methodology should also be explored in broader age ranges as the effects of MTBI on FA and diffusivity may be age-dependent. Finally, longitudinal studies are needed to determine how changes in DTI indices are related to recovery of postconcussion sequelae and objective measures of cognitive functioning. Nonetheless, this study demonstrates that acute DTI changes in the CC after MTBI are related to postconcussion symptoms. In the past, the neuropathologic underpinnings of human MTBI have been difficult to study due to the inability of older imaging modalities to reveal injury effects. Advanced MRI-based DTI methods may enhance our understanding of the neuropathology of TBI, including MTBI. Additionally, rapidly evolving techniques such as DTI may prove more sensitive than conventional imaging methods in detecting subtle, but clinically meaningful, changes following MTBI and may be critical in refining diagnosis, therapeutic intervention, and management of MTBI.

ACKNOWLEDGMENT

The authors thank Donna Mendez, MD, for assistance in patient recruitment and Stacey K. Martin for assistance in manuscript preparation. They also thank Douglas H. Smith, MD, and Andrew L. Alexander, PhD, for comments on an early draft of this manuscript.