Traumatic microbleeds persist for up to five years following traumatic brain injury despite resolution of other acute findings on MRI

Abstract

Objective:

The primary objective of this study was to track the incidence and progression of traumatic microbleeds (TMBs) for up to five years following traumatic brain injury (TBI).

Methods:

Thirty patients with mild, moderate, or severe TBI received initial MRI within 48 h of injury and continued in a longitudinal study for up to five years. The incidence and progression of MRI findings was assessed across the five year period. In addition to TMBs, we noted the presence of other imaging findings including diffusion weighted imaging (DWI) lesions, extra-axial and intraventricular hemorrhage, hematoma, traumatic meningeal enhancement (TME), fluid-attenuated inversion recovery (FLAIR) hyperintensities, and encephalomalacia.

Results:

TMBs were observed in 60% of patients at initial presentation. At one-year follow-up, TMBs were more persistent than other neuroimaging findings, with 83% remaining visible on MRI. In patients receiving serial MRI 2–5 years post-injury, acute TMBs were visible on all follow-up scans. In contrast, most other imaging markers of TBI had either resolved or evolved into ambiguous abnormalities on imaging by one year post-injury.

Conclusions:

These findings suggest that TMBs may serve as a uniquely persistent indicator of TBI and reinforce the importance of acute post-injury imaging for accurate characterization of persistent imaging findings.

Introduction

Traumatic brain injury (TBI) is among the major causes of mortality and disability worldwide (1). In the U.S. alone, TBI results in 2.8 million emergency visits, hospitalizations, and deaths each year, and an estimated 5.3 million persons are currently living with a TBI-related disability (2,3). Increasingly, TBI is seen not as a single event but as a progressive disease process with long-term physical, psychological, societal, and economic impacts (4). Therefore, long-term studies of the effects of TBI are critical for improving the lives of those living with the effects of injury.

Classification of TBI in both acute and chronic settings is important in determining treatment and overall outcome for persons with TBI. Injury severity and prognosis following TBI is typically based on the Glasgow Coma Scale (GCS) and clinical findings on computed tomography (CT) of the head. Given its near universal access and sensitivity for detection of life-threatening lesions requiring surgical intervention, CT remains the modality of choice for the evaluation of patients with TBI, despite its lack of sensitivity in detecting more subtle forms of injury (5). However, studies increasingly seek to find neuroimaging findings that may aid in the identification, classification, management, and ultimately treatment of TBI. Advanced magnetic resonance imaging (MRI) techniques allow for the identification of lesions not easily discernible on CT, and in recent years MRI has been used to examine sequelae of TBI such as diffuse axonal injury (6), traumatic meningeal enhancement (TME) (7), and edema (8). In addition, traumatic microbleeds (TMB) have emerged as potentially useful markers of TBI that can be identified early following injury and are observed across the spectrum of injury severity.

Although MRI can provide objective evidence of injury to the brain and vasculature early after TBI, MRI is rarely obtained in the emergency department. Little evidence exists to guide patient management and justify obtaining an MRI for clinical purpose. In the military combat setting, acute MRI may be impractical. Civilians and service members with persistent symptoms may receive their first MRI weeks after injury, if at all. Because many findings evident on acute MRI, such as extra-axial hemorrhage, resolve or decrease in conspicuity over time, an imaging biomarker that can be identified acutely and that also persists for years after injury would aid in diagnosis and prognosis of TBI. Furthermore, acute scans would aid in understanding the pathomechanisms of those markers of injury that resolve or evolve into nonspecific lesions such as encephalomalacia.

Microbleeds (MBs) are a common finding on MRI across many populations, including the elderly, patients with stroke, and even healthy individuals (9). In the aging population, the incidence of MBs ranges from 5–35% in large epidemiological studies (10). Deep MBs are traditionally associated with hypotensive vasculopathy, while lobar MBs are associated with cerebral amyloid angiopathy, and a higher load of either type of MB is associated with increased cognitive decline (11). In studies of patients with TBI, TMBs tend to be more superficially located, with the highest burden in frontal, temporal, and parietal lobes, although they are also seen in corpus callosum and deeper brain structures (10,12,13). Further, linear appearing TMBs detected on T2*-weighted imaging appear to be specific to TBI and may represent vascular damage induced directly by shear forces during trauma (7).

Recent studies have identified TMBs as potential imaging biomarkers of TBI. Although microbleeds have been extensively studied in aging populations, the extent to which TMBs can be used to understand the long-term progression of TBI remains unclear, as there is little understanding of how acute TMBs progress over time and whether they remain visible on MRI years following injury. TMBs are thought to represent hemosiderin deposits that remain in macrophages following microhemorrhage (14). They are typically perivascular and have been closely associated with traumatic axonal injury (15). Very few studies have examined the occurrence of microbleeds in the context of TBI, and thus their clinical significance remains largely unknown. Lawrence et al. (16) examined TMBs in 13 patients with TBI within 24 h and again 7–15 days post-injury. They found that TMBs served as a marker of injury severity and that a reduction in TMB volume over the first 15 days post-injury was associated with an improvement in GCS scores. TMBs in the dorsal brainstem have also been correlated with long-term outcome (17), while the number and volume of TMBs in frontal, parietal, and temporal regions have been correlated with depressive symptoms one year post-injury (13). Other investigators have associated the presence of TMBs with specific cognitive functions including memory (18). In a study of 14 patients with severe TBI, both the volume and number of TMBs in deep brain structures was positively associated with individual scores on the Coma Recovery Scale – Revised (19). Recent work by our group in a mostly mild TBI population has established that TMBs are a distinct form of vascular injury, and patients with TMBs were twice as likely to have disability at 30 or 90 days post-injury compared to those without TMBs (20). Thus, a number of studies indicate that TMBs may serve as a biomarker of outcome following TBI.

Only a few studies have tracked the progression of TMBs longitudinally following TBI, and none have done so beyond two years post-injury to our knowledge. In an early study of the evolution of microhemorrhages in TBI, most lesions appeared less visible over a two-year period in patients who received serial MRI (21); however, that study used only 1.0T images, and it is unclear whether these findings apply to newer, high field MRIs in use today. It is possible that more sophisticated imaging techniques in high field scanners have a greater potential to detect the initial occurrence and long-term persistence of TMBs in patients with TBI. A more recent study indicates that TMBs may expand and converge with one another during the acute period, resulting in changes in the overall number and volume of lesions during the early post-injury period (22). Watanabe and coworkers (23) also report three cases of patients with microbleeds appearing within 2–3 h after TBI. Interestingly, acute microbleeds became obscured or disappeared between 2–6 days post-injury, reappearing again at 1–3 months, which the authors attribute to susceptibility differences as the microbleeds age. A recent rodent model of traumatically-induced microhemorrhages confirmed this finding; namely, that microhemorrhages remain present histopathologically but commonly disappear on MRI during the acute post-injury period (24). Thus, the timing of acute imaging is critical for determining the presence, size, and number of TMBs as well as their prognostic value.

In this study, we report the progression of neuroimaging findings in a cohort of 30 patients who underwent MRI acutely (within 48 h of injury) following mild, moderate, or severe TBI. Given that many imaging markers of injury decrease in conspicuity or evolve into nonspecific lesions over time, we sought to provide a comprehensive characterization of the progression of neuroimaging findings identified during acute MRI for up to five years following TBI. We compared the persistence of TMBs relative to other imaging biomarkers of TBI and suggest that TMBs may serve as a unique biomarker of injury that remains distinct and detectable for a longer period of time than other lesions identified through MRI.

Methods

Participants

This is a retrospective analysis of patients enrolled in one of two IRB approved protocols (NCT01132937, NCT01287156). All patients presented acutely to either a Level I or Level II trauma center in the Washington, DC metropolitan region with suspected TBI. Inclusion criteria required participants to be at least 18 years of age, suspicion of non-penetrating TBI, able to provide informed consent, and deemed medically safe to participate by a physician. Participants were excluded if they had a contraindication to MRI, were pregnant, or were considered psychiatrically unstable by an attending physician. Patients were subsequently enrolled in a longitudinal study of the progression of TBI (NCT01132898).

Participants were evaluated and scanned initially within 48 h of head injury and then annually thereafter for up to 5 years in the longitudinal study. Additional interval imaging may have been obtained between acute imaging and 1 year follow-up depending on the participant’s acute enrollment study and when the participant enrolled into the longitudinal study. To be included in this analysis, subjects must have completed an initial and one-year visit with MRI. All additional time points were included for each patient based on availability. Participants were classified as having sustained a non-penetrating mild, moderate, or severe TBI as per the VA/DoD severity rating scale (25), including the clinical radiology interpretation of acute post-injury head CT. The scale classifies patients based on structural imaging abnormalities, duration of loss of consciousness, alteration of consciousness, post-traumatic amnesia, and Glasgow Coma Scale.

Magnetic resonance imaging data acquisition

Research MRI was obtained on four separate scanners depending on the site. Acute scans were conducted at two locations in the Washington, DC metro area: Suburban Hospital (General Electric Sigma 1.5T and Siemens Magnetom Skyra 3.0T) and Washington Hospital Center (Phillips Achieva 3.0T). All longitudinal scans were conducted at the National Institutes of Health (NIH) on a Siemens Biograph mMR 3.0T. At each site, a standardized protocol was used that included T1 and T2-weighted imaging, T2 fluid-attenuated inversion recovery (FLAIR), T2* – susceptibility-weighted imaging (SWI) and gradient recalled echo (GRE), and diffusion tensor imaging (DTI) with derived isotropic diffusion weighted images (DWI) and trace apparent diffusion coefficient (ADC). Initial MRI was obtained within 48 h of injury, then again at 7, 30, 90, and 180 days, and annually for up to 5 years following injury. Sequence parameters have been detailed in other studies that examined this cohort (7), however parameters relevant to this analysis are as follows: T2*-weighted GRE imaging parameters: field of view (FOV) 24 cm; repetition time (TR), 800 ms; echo time (TE), 12–20 ms; slice thickness, 3.5–7 mm; flip angle, 20–30 degrees; acquisition matrix, 256 × 192; scan time, 2 min. Diffusion imaging parameters were: FOV, 24 cm; TR, 5000–17000 ms; TE, 62–98 ms; slice thickness, 2–3.5 mm; acquisition matrix, 96 × 96; b values, of 0 and 1000 sec/mm², isotropically weighted; scan time, 5 min. FLAIR imaging parameters were: FOV 24cm; TR, 8000–9000 ms; TE, 92–140 ms; TI, 2200–2600 ms; slice thickness, 3–3.5 mm; acquisition matrix, 192 × 128; scan time 5min.

Longitudinal analysis of imaging biomarkers

MRI scans were read on-site at the time of scan by a clinical radiologist. Research reads were performed on scans at all time points by a board-certified neurologist blinded to subject identifiers and scan time points. DWI and ADC images were read for punctate and multifocal ischemic lesions defined as areas of hyperintensity. On T2*-weighted sequences, images were read for 1) microbleeds, defined as punctate or linear areas of hypointensity < 1 cm; 2) intraparenchymal hematoma (> 1 cm); 3) epidural hematoma (EDH) and subdural hematoma (SDH); 4) subarachnoid hemorrhage (SAH); and 5) intraventricular hemorrhage (IVH). On FLAIR, images were read for: 1) traumatic meningeal enhancement (TME) defined as focal meningeal hyperintensity after injection of gadolinium-based contrast agent on postcontrast sequences; 2) SAH defined as a hyperintense signal in the subarachnoid space; 3) contusions, defined as areas of hyperintensity co-localized with intraparenchymal hematoma > 1 cm identified on T2* -weighted sequences; and 4) presence of hyperintense signals consistent with either edema (on acute scan) and/or nonspecific injury (e.g., gliosis). The blinded neurologist used acute GRE images to locate MBs, which were then confirmed on SWI, as defined by an ovoid or linear hypointense focus < 1 cm inconsistent with blood vessels, contusion, artifact, or other MB mimics as described previously (26) and not co-localized with contusion. A lesion was considered an acute TMB if it was identified in initial scans obtained within 48 h post-injury. After initial reads, all scans previously identified with TMBs were revisited by the neurologist, still blinded to time point, and each MB flagged for longitudinal analysis.

Once the TMBs had been individually flagged, the medical image processing, analysis, and visualization (MIPAV) application was used to align, co-register and overlay scans for comparison over multiple time points. The initial scan with the flagged imaging biomarker was processed first, then the midsaggital line alignment was used to align the image vertically. The scan from the comparison time point was then registered to the aligned image with the optimized automatic registration. The two transformed images were then loaded onto the same frame, windowed and leveled, and compared by phasing back and forth between images to assess the progression of the flagged injury. If all bleeds initially marked were present on both initial and follow-up scans, the TMB was flagged as “persistent.” If the bleed was no longer visible in any capacity, the image was flagged as “resolved.” If the initial MRI had no microbleed but the follow-up time point had been marked positive for MB by the blinded neurologist, the image was flagged as “new.” Individually flagged TMBs were counted and labeled with the locations left hemisphere (LH), right hemisphere (RH) and cerebellum (CB). If TMBs were present in multiple regions, all regions were documented. Patients who presented with closely clustered and branching microbleeds too dense to quantify were documented as having 5+ bleeds.

Results

Participants

A total of 30 patients from both the acute and longitudinal TBI protocols were included in this analysis. Demographics for the participants are shown in Table 1. There were no individuals with assault or exposure to blast as their mechanism of injury. During the acute time period, 20 scans were completed at Suburban Hospital (n = 16 on a General Electric Sigma 1.5T and n = 4 on a Siemens Magnetom Skyra 3.0T); n = 10 patients were scanned acutely at Washington Hospital Center on a Phillips Achieva 3.0T. All 30 patients had longitudinal scans conducted at the NIH on a Siemens Biograph mMR 3.0T.

Table 1.

Demographic and clinical characteristics of study participants.

Sex (M, F)	23, 7
Mean Age (SD)	46.4 (16.5)
Mean Years of Education (SD)	15.7 (2.9)
Median GOSE at 1 year (Q1-Q3)	7 (6–8)
Median NBSI at 1 year (Q1-Q3)	8 (2.3–19.5)
Median hours to initial scan (Q1-Q3)	23 (18.8–34.9)
Ethnicity, N (%)
White	22 (77%)
Black/African American	3 (10%)
Asian	1 (3.3%)
Latino	1 (3.3%)
Multiple	3 (10%)
TBI Severity, N (%)
Mild	7 (23%)
Moderate	19 (63%)
Severe	4 (13%)
Mechanism of Injury, N (%)
Fall	15 (50%)
Direct Impact	11 (37%)
Acceleration/Deceleration	4 (13%)
Assault	0
Blast	0
Cause of Injury, N (%)
Incidental Fall	17 (57%)
Road Traffic Accident	10 (33%)
Other	3 (10%)

GOSE = Extended Glasgow Outcome Scale; NBSI = Neurobehavioral Symptom Inventory

Biomarkers of TBI

Table 2 summarizes the incidence and progression of imaging biomarkers of TBI in our sample. Twenty-eight (93%) patients presented with imaging biomarkers of TBI on acute MRI at the initial time point. By DoD/VA standards, the severity classification of these patients based on initial injury presentation and acute head CT ranged from mild to severe. The two patients with no evidence of injury on MRI in this study were both categorized as having mild TBI based on their clinical presentations and absence of findings on acute head CT. Seventeen patients had evidence of DWI lesions at the initial scan, all of which were resolved by the 1-year follow-up. Twenty-four patients had extra-axial hemorrhages at the initial scan, only 6 (25%) of which remained observable with 5 (20.8%) presenting with nonspecific encephalomalacia and the remaining 13 (54.1%) showing resolution at 1-year follow-up. Seven patients presented with contusions (defined as intraparenchymal hematomas greater than 1cm in diameter in association with surrounding edema on FLAIR), and 5 patients presented with intraventricular hemorrhages, with all injuries showing resolution at the 1-year follow-up. Sixteen patients showed hyperintensities on FLAIR imaging consistent with evidence of contusion and/or edema on their initial scan. Of those patients, 4 (25%) had their MRI findings resolve by 1-year follow-up, 10 (62.5%) presented with persistent hyperintensity on FLAIR imaging and the development of encephalomalacia, and 2 (12.5%) had encephalomalacia in the absence of persistent hyperintensity on FLAIR. Sixteen patients presented with evidence of TME, with 10 (63%) resolving at 1-year follow-up.

Table 2.

Incidence and progression of MRI findings of traumatic brain injury from acute scan (<48 h) to one year post-injury.

^*Some findings were particularly likely to resolve across the population within a year

^**Encephalomalacia can be difficult to identify without an acute scan for comparison, but in our study was observed in a significant number of patients

Notes: DWI = diffusion weighted imaging; TME = traumatic meningeal enhancement; TBI = traumatic brain injury

Microbleeds

Table 3 summarizes the incidence, location, and progression of TMBs in our sample. MBs were found in 20 (66.7%) patients in the current study. MBs not present during acute scans were identified in two patients; therefore, a total of 18 patients (60%) presented with evidence of bleeds at their initial scan which were subsequently classified as TMBs. Seventeen patients with acute TMBs had at least one other biomarker of injury. In two patients, new MBs not present at the initial acute scan were observed during 90-day (patient 19) and 1-year (patient 20) follow-up. For both participants exhibiting new MBs, acute scans had been performed at 1.5T, while follow-up scans were performed at 3T. Due to participant attrition, not all patients received MRI throughout the entire 5-year study. However, for those patients with TMBs identified at their initial scan, 15/18 (83%) demonstrated persistence at one-year follow-up and all subsequent time points. Only 3 (17%) patients with TMBs identified during initial scan showed resolution at one year. Four patients presented with closely clustered and branching microbleeds too dense to quantify and were documented as having 5+ bleeds.

Table 3.

Incidence, location, and progression of traumatic microbleeds from the acute period up to five years following traumatic brain injury.

Notes: TMB = traumatic microbleed; LH = left hemisphere; RH = right hemisphere; CB = cerebellum; * “+” indicates a cluster of bleeds difficult to quantify

In patients with at least one MB at any time point (n = 20), lesions were observed in the left hemisphere only in 9 (45%) patients, in both hemispheres in 6 patients (30%), and in the right hemisphere only in 4 (20%) patients. A single cerebellar TMB was observed in one patient. Across severity, at least one TMB was observed in 2 (28.6%) patients with mild TBI (n = 7), 15 (78.9%) with moderate (n = 19), and 3 (75%) with severe injuries (n = 4). Figure 1 shows a representative progression of TMBs across 5 years in a single patient. All representative images are 3T. Although there appeared to be partial changes in the volume and number of TMBs over time, this was not quantified due to the small sample size of this study.

Figure 1.

Representative images of the progression of traumatic microbleeds (TMBs) in a single patient from the acute period (<48 h) to 5 years post-injury at 3T. The microbleed cluster persists almost completely across the 5 year timespan, although there appears to be a reduction in overall volume of the TMBs, which was not quantified.

Note: 48h and 30d are GRE sequences, while the remaining images are SWI sequences.

Discussion

The results of this study indicate that TMBs are a uniquely persistent neuroimaging biomarker of TBI. Although TMBs were not the most common finding during initial MRI, they were the most persistent evidence of injury observed one year post-injury. Eighty-three percent of TMBs identified in acute scans were persistent at 1-year follow-up. Furthermore, in those patients completing follow-up in years 2–5 post-injury, acute TMBs were persistent and visible on all subsequent images. In contrast, other biomarkers of injury were much more likely to resolve or evolve nonspecifically during the first year following injury. For example, patients identified with acute intracerebral hemorrhage and edema largely presented with persistent areas of encephalomalacia and FLAIR hyperintensities after one year. Without an acute scan as a reference, these lesions become indistinguishable from one another in retrospect, with the exception of focal FLAIR abnormalities which can be identified as microbleeds through colocalized hypointensities on GRE/SWI. Overall, these findings suggest that TMBs may serve as a uniquely persistent biomarker of TBI and reinforce the need for acute imaging as a reference in the identification of chronic brain injuries.

Although the small sample size of this study prohibits a detailed statistical examination of the relationship between TMBs, injury severity, and clinical outcome, a few patterns in the data are worth discussion. First, patients across the spectrum of injury severity displayed evidence of at least one TMB. Likewise, we found evidence of multiple TMBs in every injury severity. Consistent with previous studies that have identified TMBs in both mild (12) and moderate to severe (27) injury, this would seem to indicate that the development of TMBs is independent of injury severity. We also observed an association between the number of TMBs and the number of other imaging biomarkers of injury. Given that patients with mild TBI typically have the fewest imaging markers of injury, this would seem to indicate that the number of TMBs may be more closely associated with injury severity. Unlike previous studies that have observed a temporary disappearance of TMBs on MRI during the acute post-injury phase in both humans (23) and animals (24), we found no evidence of disappearance or reappearance of TMBs in our select population of patients who underwent serial MRI. Only well into the chronic post-injury phase did we observe the appearance of new microbleeds not observed acutely, suggesting these were non-traumatic in origin. Toth et al. (24) have suggested that the acute period of invisibility of TMBs may be approximately between 24 h and 7 d post-injury. In our study, the median time to acute imaging was 23 h, with patients scanned again 7 d post-injury. Thus, the timepoints used in this study may be optimal for avoiding this phenomenon altogether and further highlight the importance of early MR imaging in studies of TMBs.

Overall, the 60% incidence of TMBs observed across injury severity in this study is higher than what has been observed in other studies of patients with TBI. In patients with mild, moderate, and severe TBI, Toth et al. (28) observed an overall 37% incidence of TMBs, while Lawrence et al. (16) observed an overall 46% incidence. The higher overall incidence of TMBs is most easily explained by the large numbers of moderate and severe cases in this study, who represent 86% of the sample and are more likely to have other traumatic lesions that coincide with TMBs. Considering mild TBI only, previous studies by van der Horn et al. (12) and Wang et al. (13) observed TMBs in 28% and 19% of patients, respectively, while only 8% of patients with mild TBI in Toth (28) had TMBs. Griffin et al. (20) observed TMBs in 29% of patients with a GCS of 14 or 15. The 28% of patients with mild TBI having at least one TMB in the current study is similar to these studies; however, the small number of patients with mild TBI in our study (n = 7) limits the generalizability of these findings relative to other studies with a higher number of patients having mild TBI.

Previous studies have demonstrated that TMBs can be identified in the acute post-injury period following TBI, yet the persistence of TMBs in the chronic post-injury period has received little attention, and there is a paucity of information regarding the evolution of TMBs over time. Unlike studies examining MBs in aging populations, it is unclear whether TMBs change in number, volume, and visibility in the months to years following TBI. In a study of military service members with a history of TBI, Liu and colleagues (29) found that the number of microhemorrhages identifiable with MRI decreased over time. The sample included predominantly patients with mild TBI due to blast injury. Considering the results from our study, this suggests that injury mechanism and severity of injury may play a role in the evolution of TMBs over time. In support of this notion, another recent study in military service members exposed to blast injury found no evidence of cerebral microhemorrhage on SWI (30). Therefore, TMBs may provide insight into the distinct pathophysiological processes underlying different types of TBI. Our results appear to be more consistent with the evolution of MBs in the general aging population, where large studies have suggested that the number of microbleeds rarely decreases over time (31).

Traumatic injury to the vasculature is a near universal feature of TBI (32), and recent studies from our group and others underscore the importance of identifying TMBs that persist for years following TBI. Previously, researchers have associated TMBs primarily with axonal injury (15,33). However, Griffin et al. recently demonstrated, using MRI-guided pathology, that what appears as punctate lesions on MRI are iron-laden macrophages in the perivascular space, extending in a linear fashion over an area much larger than what is visible on MRI (20). In an acute study of patients with stroke and TBI, these linear appearing TMBs were only present in patients with TBI, indicating that they may be a unique form of traumatically-induced injury to the vasculature (7). Similarly, Andreasen et al. have found that TMBs seen after severe TBI are not co-localized with axonal pathology except in the midsagittal region (19). Recent animal studies also indicate traumatic microvascular injury is associated with progressive pathology and poor outcome (34,35). These studies, taken together with our finding that TMBs may persist for up to 5 years following injury, indicate that TMBs represent an ongoing pathophysiological process distinct from axonal injury that may be amenable to novel therapeutics targeting the vasculature.

There are a few limitations to this study. We had a small sample size restricted by the patients enrolled in both TBI studies, and there was subsequent attrition of patients who did not return for all follow-up visits across the 5-year period. Our study also had a large proportion of moderate and severe TBI compared to the larger TBI population, as may be expected given the high frequency of imaging findings in our patients. In addition, acute and follow-up scans were acquired with different MRI machines at different sites, resulting in various parameters across sequences and differences in scan quality. Although this makes it difficult to compare imaging data from multiple testing sites, ambiguous images were assessed multiple times by a reader blinded to time point and patient identification, with consistent results. The inconsistencies in scan quality and resolution across 1.5T and 3T MR scanners may also affect the visibility of certain injuries over time. On the other hand, the fact that we were able to identify and track the progression of TMBs across multiple scanners and institutions makes us optimistic that similar techniques could be applied to a range of clinical settings. The NCAA-DoD Concussion Assessment, Research, and Education (CARE) consortium, which conducts large-scale, multi-site studies of athletes and military service cadets, likewise found high stability in MRI metrics used in their studies, with between-subject instability equal to or greater than between-site instability (36).

In any study of TMBs where pre-injury scans are not available, it is not possible to completely rule out the presence of premorbid MBs that could confound the interpretation of the effects of TBI. Microbleeds of non-traumatic origin have long been associated with normal aging (31), Alzheimer’s Disease (AD) (37), vasculopathy (38) and other nervous system disorders. However, the incidence of MBs observed in this study is much higher than what would be expected even in an aging population, suggesting that the MBs observed in our sample are most likely traumatic in origin. Furthermore, we made every effort to distinguish TMBs from those MBs of non-traumatic origin, including the exclusion of MBs that did not appear during initial scans.

Conclusion

The identification of TMBs in the early post-injury period that persist for years following injury reinforces the importance of obtaining advanced imaging sequences acutely following injury. Such early findings may play a critical role in the diagnosis, management, and stratification of patients with TBI. Acutely, the presence of TMBs may help with overall prognosis and identification of those patients requiring specialized care. Chronically, the presence of TMBs has the potential to serve as an indicator of long-term clinical outcome. Furthermore, while many acute imaging findings are resolved or unremarkable at one-year follow-up, our data suggests that TMBs are likely to persist for years in a majority of patients who show evidence of bleeds in acute scans. Therefore, the presence of TMBs on a chronic MRI may indicate the co-existence of other significant injury that could have been identified during an acute scan.

These results hold the potential for many future directions. Currently, it is unclear how the number, volume, and distribution of TMBs is associated with injury severity and long-term clinical and neuropsychological outcome. The presence of these small blood deposits in the brain years after injury indicates a risk of prolonged inflammatory reactions in the parenchyma with potentially harmful effects. Indeed, previous work in other patient populations has demonstrated direct pathological links between MBs and focal tissue injury, including the presence of β-amyloid and markers for cellular apoptosis (39). In imaging studies, the presence of cerebral MBs has been associated with widespread loss of white matter structural integrity (40). Recent work from our group indicates that the area surrounding TMBs is associated with cell loss, demyelination, and pathology consistent with ischemia, indicating that the parenchyma surrounding TMBs is injured (20). Further, the presence of baseline MBs in the aging population is associated with a 5-fold increased risk of developing new MBs at follow-up (31); future studies are necessary to determine whether similar pathologies exist in patients who exhibit TMBs following TBI. In other populations, the presence of MBs may identify subgroups of patients who are more susceptible to long-term cognitive decline (9). Thus, the identification of TMBs and their comorbid imaging biomarkers may lead to the identification of specific phenotypes of TBI that could provide insight into long-term outcome and improve clinical decision and patient stratification, ultimately leading to the identification of new therapeutic targets. Similarly, the resolution of more severe markers of injury, compared with the persistence of TMBs, raises the possibility of inherently different recovery processes at work in response to different types of injuries. Lastly, it is unknown to what extent the presence of TMBs may interact with the normal aging process, where the presence of non-traumatic MBs predicts the likelihood of developing future MBs, and the overall number of lesions generally increases with age.