Diffusion Tensor Imaging

DTI is a non-invasive imaging technique sensitive to the movement of water molecules in nervous tissue (Jones, Reference Jones2008). When unrestricted, water diffuses randomly, whereas it is hindered in its movement in highly organized tissue such as collections of nerve fibers composing white matter. As illustrated in Figure 1, the combination of axonal membrane and myelin coating form barriers restricting intra-axonal water dispersion (vertically oriented lines in the diagram). The flow and dispersion of water molecules are in a direction mostly perpendicular to the confining membrane wall. If degradation from trauma occurs within the axonal membrane and surrounding myelin, water dispersion is less restricted. When the level of damage in TBI reaches a threshold where the membrane becomes leaky, breaks down, or is literally sheared away, DTI detects the signal changes in water dispersion that reflect breaches in white matter integrity.

Fig. 1

(a) This diagram depicts unconstrained molecular movement of a water droplet when there are no membranes to hold the water within. (b) Vertically oriented axon with a membrane wall that constrains the direction of movement of intracellular water as it moves in the same vertical direction (see up-and-down arrows) as the constraining membrane. However, as shown in (c), if the membrane breaks down or is degraded, water molecules disperse in a similar manner to the unconstrained condition. FA=fractional anisotropy, ADC=apparent diffusion coefficient, MD=mean diffusivity. Figure was created by Geri R. Hanten, Ph.D., Baylor College of Medicine.

Some of the most frequently used DTI metrics useful in quantifying pathological change in white matter include the apparent diffusion coefficient (ADC) and fractional anisotropy (FA). Other often-used metrics are mean diffusivity (MD), axial diffusivity (AxD), and radial diffusivity (RD). ADC and MD measure the extent to which water is impeded in the tissue of interest. Pathologically high ADC or MD values may reflect unimpeded motion and are suggestive of decreased integrity of neuronal structure, whereas low values are thought to reflect cellular edema and inflammation. DTI involves the use of a tensor model to extract the directionality of water flow in axonal fiber tracts. FA, perhaps the most frequently used index in DTI studies, measures the extent to which water displacement is directionally dependent in its flow along the cell. High FA values are indicative of strong anisotropy, putatively associated with intact white matter microstructure. However, elevated FA beyond an expected normative value may be a sign of inflammation, where perturbation of white matter has the potential to stimulate an inflammatory response (Bramlett & Dietrich, in Reference Bramlett and Dietrichpress). AxD reflects water diffusion along the longitudinal axis of the tensor and, although controversial, is purported to be a measure of axonal structure and damage. RD reflects diffusion along the two non-primary axes of the tensor and is purported to be a measure of myelin integrity.

Since similarly oriented white matter fibers will possess similar water diffusion properties, aggregated axons traveling together may be extracted (Choudhri et al., Reference Choudhri, Chin, Blitz and Gandhi2014), a method referred to as DTI tractography. This method represents a key neuroimaging tool for network identification (Catani & Thiebaut de Schotten, Reference Catani and Thiebaut de Schotten2008) and allows traumatic axonal injury in TBI to be visualized as an interruption or degradation of fibers within specific white matter tracts (see Figure 2).

Fig. 2

(a) Dorsal view of an age-matched control whole-brain tractography image compared to (b) a patient with severe TBI. The white arrows point to the corpus callosum, which is intact in the control but with major tract loss in the TBI patient. Likewise, a mid-sagittal T1 MRI view shows an intact corpus callosum in the control subject (c) and a withered, atrophied one in the TBI patient (d). More dramatically, DTI tractography of the corpus callosum shows extensive loss of callosal tracks in the TBI patient compared to the control.

Although DTI represents a great improvement in terms of sensitivity to traumatic axonal injury, there is currently no one accepted standard data analysis protocol, making cross-study comparisons challenging. DTI metrics can be derived from several methods including ROI-based methods, programs such as tract-based spatial statistics (TBSS) that allow for whole-brain analysis, and tractography. After data collection, DTI data are subjected to an array of processing steps before they are interpretable. In the pre-processing pipeline, decisions need to be made regarding how to deal with eddy current artifacts and subject motion, although improvements in imaging hardware over the last decade have reduced concerns of motion sensitivity and signal distortions for DTI data (Tournier et al., Reference Tournier, Mori and Leemans2011). With regard to post-processing, there are different methods for estimating the tensor and deriving quantitative measures, which may influence the results (for a review, see Jones & Cercignani, Reference Jones and Cercignani2010). As such, consideration of the analytic methods used is important in interpreting findings.

Comparison of Methods to Detect TBI

Several studies have now shown that DTI methods can uncover white matter abnormalities not visible on conventional clinical scans (Huisman et al., Reference Huisman, Schwamm, Schaefer, Koroshetz, Shetty-Alva, Ozsunar and Sorensen2004; Kim et al., Reference Kim, Branch, Kim and Lipton2013; Skoglund et al., Reference Skoglund, Nilsson, Ljungberg, Jönsson and Rydenhag2008), suggesting that there are likely to be more significant structural changes following TBI than previously assumed. Furthermore, unlike conventional scans, diffusion methods such as tractography can visualize long-coursing connections that can potentially be affected by focal injury produced by the injury force. This is demonstrated in Figure 3, which illustrates different imaging modalities in a patient with severe TBI who suffered extensive frontal lobe damage as a result of surface and deep contusions. These are visualized on CT scan conducted on the day of injury (Figure 3a). MR studies performed more than 4 years later reflect the chronic pathological changes. The axial FLAIR MRI sequence clearly shows a white matter signal abnormality identified as a hyperintensity in the left frontal lobe (white arrow, Figure 3d). The white matter damage can also be seen in the T1 images, but not as distinctly as in the FLAIR image. There has also been loss of gray matter and the development of focal degeneration, referred to as encephalomalacia, in both the frontal and temporal polar areas (Figure 3b, 3c). Furthermore, as observed in the mid-sagittal T1 MRI, frontal white matter damage affected interhemispheric pathways across the anterior corpus callosum and reduced its size (Figure 3e, f). Returning to the FLAIR image, even though the left frontal lobe hyperintense white matter signal is so prominent and divergent from its contralateral counterpart, it would be a mistake to assume that this white matter damage and the surrounding focal encephalomalacia produces just a focal effect. This becomes evident when viewing the DTI tractography map (taken from a healthy control subject; Figure 3g), which shows six white matter pathways that would normally course through the site of focal frontal parenchyma damage sustained by the TBI patient and would conceivably be influenced by the injury. Thus, focal white matter damage has the potential to influence major fiber tracts, some of which may disrupt functioning distal to where the focal white matter damage occurred. Although not shown in the figure, it is important to note that damage to white matter tracts due to the force of the initial injury may independently contribute to loss in connectivity.

Fig. 3

(a) Day-of-injury computed tomogram (CT) showing extensive hemorrhagic lesions (white scattered lesions) especially notable in the inferior left frontal region (red arrow). (b) The sagittal T1-weighted MRI obtained four years post-injury shows extensive left anterior inferior frontal (top red arrow with white border) and temporal pole (bottom red arrow with dark border) wasting, referred to as encephalomalacia. The green line depicts the orientation of the oblique coronal cut as shown in (c). The particularly dark signal (white arrow) underlying the focal encephalomalacia represents hemosiderin. (d) In contrast, the MR signal in the white matter regions of the frontal and temporal lobes as depicted in the T1 images is abnormal, but not as distinct as the hyperintense signal in the axial fluid attenuated inversion recovery (FLAIR) sequence. T1 mid-sagittal MRI showing reduced size of the anterior aspect of the corpus callosum (e) in comparison to an age-matched control (f). (g) Pathways in the normal brain that would normally course through what is now damaged parenchyma and thus would be damaged in the TBI patient. These include, in addition to the corpus callosum, the anterior cingulum (green), occipitofrontal fasciculus (brown), uncinate fasciculus (teal), occipitotemporal fasciculus (red), and arcuate fasciculus (purple).

Tracts Most Often Impacted by TBI

Some of the tracts most vulnerable to head injury include long coursing and commissural fibers. Midline structures such as the corpus callosum may be more susceptible to rotational and shearing forces than other brain regions (Gentry et al., Reference Gentry, Thompson and Godersky1988; Reinarz et al., Reference Reinarz, Coffman, Smoker and Godersky1988; Tate & Bigler, Reference Tate and Bigler2000). Consistent with this, a landmark port-mortem study reported damage to axons in the corpus callosum and fornix following TBI of varying severity, suggesting particular vulnerability in these regions to axonal injury (Blumbergs et al., Reference Blumbergs, Scott, Vis, Wainwright, Simpson and McLean1995). DTI studies have largely supported this postmortem finding by providing in vivo evidence for the vulnerability of these midline regions (Kinnunen et al., Reference Kinnunen, Greenwood, Powell, Leech, Hawkins, Bonnelle and Sharp2011; Kraus et al., Reference Kraus, Susmaras, Caughlin, Walker, Sweeney and Little2007). Other major tracts that have consistently shown abnormalities in TBI include the internal capsule (Kraus et al., Reference Kraus, Susmaras, Caughlin, Walker, Sweeney and Little2007), superior and inferior longitudinal fasciculi (Kinnunen et al., Reference Kinnunen, Greenwood, Powell, Leech, Hawkins, Bonnelle and Sharp2011), and cerebellar fibers (Mac Donald et al., Reference Mac Donald, Johnson, Cooper, Nelson, Werner, Shimony and Brody2011; Warden et al., Reference Warden, French, Shupenko, Fargus, Riedy, Erickson and Moore2009). With increasing injury severity, involvement of the brainstem is highly likely, leading to transient loss of consciousness and coma. According to one influential model of TBI, injury to the brain occurs in a centripetal manner (Ommaya & Gennarelli, Reference Ommaya and Gennarelli1974), with loss of consciousness resulting from greater disruptions between deep structures (e.g., brainstem) and cortical and subcortical structures.

White Matter Disruption Observed in Cross-sectional Studies

Alterations in white matter integrity are observed as early as within 24 hours of injury (Marmarou et al., Reference Marmarou, Signoretti, Fatouros, Portella, Aygok and Bullock2006; Newcombe et al., Reference Newcombe, Williams, Nortje, Bradley, Harding, Smielewski and Menon2007) and are still evident several years post-injury (Kennedy et al., Reference Kennedy, Wozniak, Muetzel, Mueller, Chiou, Pantekoek and Lim2009; Kraus et al., Reference Kraus, Susmaras, Caughlin, Walker, Sweeney and Little2007; Nakayama, Reference Nakayama2006; Palacios et al., Reference Palacios, Sala-Llonch, Junque, Roig, Tormos, Bargallo and Vendrell2012; Xu et al., Reference Xu, Rasmussen, Lagopoulos and Håberg2007). Studies of chronic (i.e., >3 months post-injury) moderate to severe TBI reliably reveal decreased anisotropy and increased diffusivity, suggesting loss of axonal integrity and myelin degeneration. As an example, Kraus et al. (Reference Kraus, Susmaras, Caughlin, Walker, Sweeney and Little2007) demonstrated reductions in FA in several fiber tracts including the internal capsule, corpus callosum, and superior longitudinal fasciculus, among other regions, in adults with moderate to severe TBI. Consistent with these results, Palacios et al. (Reference Palacios, Sala-Llonch, Junque, Roig, Tormos, Bargallo and Vendrell2012) found TBI-related FA reductions in the corpus callosum, superior longitudinal fasciculus, and other long and short fibers in participants 2 months to 21 years post-injury. Taken together, the findings suggest that white matter abnormalities following moderate to severe TBI persist in the chronic stage and while there is variability in the tracts that have been shown to be abnormal, some consistent findings also emerge.

Individuals with mild TBI (mTBI) also exhibit white matter abnormalities early in the course of injury recovery (Lo et al., Reference Lo, Shifteh, Gold, Bello and Lipton2009), and in a subset of individuals, these abnormalities persist years post-injury (Fakhran et al., Reference Fakhran, Yaeger and Alhilali2013; Geary et al., Reference Geary, Kraus, Pliskin and Little2010; Lipton et al., Reference Lipton, Gellella, Lo, Gold, Ardekani, Shifteh and Branch2008). As in more severe injury, commissural and long coursing fibers are altered in mTBI, albeit less extensively so (Goetz et al., Reference Goetz, Blamire, Rajagopalan, Cadoux-Hudson, Young and Styles2004; Kraus et al., Reference Kraus, Susmaras, Caughlin, Walker, Sweeney and Little2007). However, there is less anatomical consistency in terms of injured regions in mTBI than in moderate and severe TBI, which may be related to mechanism and type of injury. Nonetheless, a recent meta-analysis of studies of mTBI concluded that the corpus callosum is particularly vulnerable (Aoki et al., Reference Aoki, Inokuchi, Gunshin, Yahagi and Suwa2012).

Sports Concussion

mTBI has been studied extensively in the context of sports concussion. A recent review of DTI studies of sports-related mTBI led to the conclusion that microstructural white matter abnormalities may be present both in the acute and in the sub-acute/chronic phase in mild injury (Gardner et al., Reference Gardner, Kay-Lambkin, Stanwell, Donnelly, Williams, Hiles and Jones2012). Of particular concern is that athletes in sports such as boxing and football are often subjected to repetitive concussions and subconcussive hits that cumulatively impact brain structure and cognitive function. The adverse consequences of repetitive injury have gained greater attention recently due to reports of poor long-term neural and behavioral outcomes in professional football players. The few human studies that have systematically examined this issue suggest that white matter abnormalities increase with the number of repetitive concussions and subconcussive hits (Bazarian et al., Reference Bazarian, Zhu, Blyth, Borrino and Zhong2012; Koerte, Ertl-Wagner, et al., Reference Koerte, Ertl-Wagner, Reiser, Zafonte and Shenton2012; Koerte, Kaufmann, et al., Reference Koerte, Kaufmann, Hartl, Bouix, Pasternak, Kubicki and Shenton2012). Additional studies are now underway to uncover the relationship between repetitive head injury and neurodegenerative disease such as chronic traumatic encephalopathy (McKee et al., Reference McKee, Stein, Nowinski, Stern, Daneshvar, Alvarez and Baugh2013).

Blast Exposure

Another increasingly common cause of mTBI is blast exposure. The consequences of blast-induced brain injury have gained visibility in recent years in large part due to the stateside return of thousands of Iraq and Afghanistan service members exposed to explosive ordnance such as improvised explosive devices and rocket propelled grenades. One important difference between blast-induced mTBI and mTBI from civilian events is that the acceleration/deceleration effects that are typical of sports and motor vehicle trauma are less likely to occur following exposure to primary blast. Naturally, an important question that arises is whether distinct mechanisms (e.g., blast vs. direct head impact) result in different neural sequelae. Few studies have directly examined the impact of injury mechanism on white matter, but studies of primary blast injury show white matter abnormalities that appear to mimic the abnormalities seen in the civilian literature (Hayes et al., Reference Hayes, Morey and Tupler2012; Warden et al., Reference Warden, French, Shupenko, Fargus, Riedy, Erickson and Moore2009). In actuality, the majority of blast injuries involve multiple mechanisms, both blast and acceleration/deceleration forces that arise from secondary and tertiary blast effects. Thus, the bulk of studies examining military blast mTBI have included individuals with both primary blast exposure and additional secondary and tertiary injuries. These studies have generally shown decreases in white matter integrity both in the subacute (Mac Donald et al., Reference Mac Donald, Johnson, Cooper, Nelson, Werner, Shimony and Brody2011) and chronic phase (Davenport et al., Reference Davenport, Lim, Armstrong and Sponheim2012; Hayes et al., Reference Hayes, Miller, Lafleche, Salat and Verfaellie2015; Jorge et al., Reference Jorge, Acion, White, Tordesillas-Gutierrez, Pierson, Crespo-Facorro and Magnotta2012; Mac Donald et al., Reference Mac Donald, Johnson, Cooper, Nelson, Werner, Shimony and Brody2011; Morey et al., Reference Morey, Haswell, Selgrade, Massoglia, Liu, Weiner and McCarthy2013), similar to the civilian mTBI literature (but see Bazarian et al., Reference Bazarian, Donnelly, Peterson, Warner, Zhu and Zhong2013; Levin et al., Reference Levin, Wilde, Troyanskaya, Petersen, Scheibel, Newsome and Li2010; Watts et al., Reference Watts, Thomas, Filippi, Nickerson and Freeman2014). However, findings of greater neurologic deficits in individuals exposed to combat-related mTBI than civilian mTBI (Ruff et al., Reference Ruff, Riechers, Wang, Piero and Ruff2012) provide some evidence for increased noxious effects due to blast and suggest the need for further studies comparing blast-induced mTBI and civilian TBI. Additionally, the high co-occurrence of PTSD in this population and the presence of white matter abnormalities associated with PTSD further complicate interpretation of structural brain connectivity (Davenport et al., Reference Davenport, Lim and Sponheim2015). Nonetheless, evidence from a recent post-mortem study of veterans suggests that blast exposure is associated with axonal pathology at the histological level (Ryu et al., Reference Ryu, Horkayne-Szakaly, Xu, Pletnikova, Leri, Eberhart and Koliatsos2014).

Longitudinal Studies and Recovery after Injury

Given the dynamic nature of traumatic axonal injury, DTI studies hold much promise for charting the progression and possible recovery of white matter injury over time. Interpretation of longitudinal findings is challenging, however, given that the role of primary and secondary injury processes will vary at different times post-injury and that there is as of yet no comprehensive model of TBI-associated diffusion changes over time. Changes in DTI metrics can be interpreted with reference to animal models of TBI (e.g., Mac Donald et al., Reference Mac Donald, Dikranian, Song, Bayly, Holtzman and Brody2007), but the time frame of pathophysiological responses in humans may not directly map onto that seen in animals. Additionally, given the known limitations in characterizing white matter in complex regions where fibers intersect, interpretation of longitudinal changes in such regions can be especially difficult. For instance, an FA increase over time that is consistent with axonal recovery or regeneration could also be observed when one of the crossing fiber populations degenerates, and the net result would be increased orientational coherence (e.g., Sidaros et al., Reference Sidaros, Engberg, Sidaros, Liptrot, Herning, Petersen and Rostrup2008). Conversely, a lack of change in DTI metrics over time need not reflect absence of pathophysiological change. Evaluating white matter integrity in concussed football players 1–6 days postconcussion and again 6 months later, Henry et al. (Reference Henry, Tremblay, Tremblay, Lee, Brun, Lepore and Lassonde2011) found similarly increased FA in the corpus callosum and corticospinal tract at both stages. Nonetheless, they speculated that distinct processes might be involved, with findings in the acute stage reflecting intracellular edema, and those in the chronic stage reflecting a loss of crossing fibers.

Notwithstanding these difficulties, several patterns of findings are emerging from recent longitudinal studies of TBI. Studies of mTBI that compare the acute or semi-acute stage to follow-up in the chronic stage have reported changes in DTI metrics consistent with continued evolution of the injury. For instance, evaluating veterans with blast-induced mTBI over a 6- to 12-month interval, Mac Donald et al. (Reference Mac Donald, Johnson, Cooper, Nelson, Werner, Shimony and Brody2011) observed normalization of some DTI metrics (mean and radial diffusivity) but new abnormalities in others (axial diffusivity), suggesting that cellular edema and inflammation normalize, but axonal injury may persist. Other studies have provided evidence that DTI metrics reflecting axonal injury may partially normalize, suggesting a process of recovery that is incomplete (e.g., Arfanakis et al., Reference Arfanakis, Haughton, Carew, Rogers, Dempsey and Meyerand2002; Grossman et al., Reference Grossman, Jensen, Babb, Chen, Tabesh, Fieremans and Grossman2013; Mayer et al., Reference Mayer, Ling, Mannell, Gasparovic, Phillips, Doezema and Yeo2010; but see Ling et al., Reference Ling, Pena, Yeo, Merideth, Klimaj, Gasparovic and Mayer2012). Taken together, these studies leave open the possibility of continued axonal recovery over more extended time periods, although it is notable that persistent white matter changes have been observed even years after mTBI (e.g., Inglese et al., Reference Inglese, Makani, Johnson, Cohen, Silver, Gonen and Grossman2005).

There is also evidence for partial recovery of white matter injury in moderate/severe TBI (Kumar et al., Reference Kumar, Saksena, Husain, Srivastava, Rathore, Agarwal and Gupta2010; Pal et al., Reference Pal, Gupta, Agarwal, Yadav, Ojha, Awasthi and Narayana2012; Skoglund et al., Reference Skoglund, Nilsson, Ljungberg, Jönsson and Rydenhag2008). For instance, examining a group of patients with frontal injury apparent on clinical neuroimaging, Pal and colleagues (Reference Pal, Gupta, Agarwal, Yadav, Ojha, Awasthi and Narayana2012) found significant increase in FA values in fronto-posterior and fronto-limbic tracts between 1 week and 6 months post-injury, although those values remained lower than in controls. Focusing specifically on the corpus callosum, Kumar et al. (Reference Kumar, Saksena, Husain, Srivastava, Rathore, Agarwal and Gupta2010) similarly reported an improvement in DTI metrics, albeit it not to the level of controls. Interestingly, the latter changes were apparent between the acute and 6-month follow-up scan, but there was no evidence for continued improvement at 2-year follow-up. To date, there is insufficient data to draw conclusions about the temporal window of recovery, but two group studies have indicated continued recovery up to a year (Sidaros et al., Reference Sidaros, Engberg, Sidaros, Liptrot, Herning, Petersen and Rostrup2008) and 4 years after injury (Farbota et al., Reference Farbota, Bendlin, Alexander, Rowley, Dempsey and Johnson2012). Perhaps most striking in this regard is a case study of a patient with severe TBI who emerged from a minimally conscious state 19 years post-injury and was subsequently scanned on two occasions 18 months apart (Voss et al., Reference Voss, Uluç, Dyke, Watts, Kobylarz, McCandliss and Vallabhajosula2006). During that time, normalization of FA was seen in several regions, which was thought to represent axonal regrowth.

A striking counterpoint to the findings of partial recovery is evidence of continued deterioration of white matter in moderate/severe TBI. Greenberg et al. (Reference Greenberg, Mikulis, Ng, DeSouza and Green2008) scanned patients with moderate to severe injury 4.5 months post-injury and then again on average 2 years later, and found increased deep frontal and temporal white matter impairment. Similarly, Wang et al. (Reference Wang, Bakhadirov, Abdi, Devous, Marquez de la Plata, Moore and Diaz-Arrastia2011) found, in a group of patients with mostly severe injury, that fiber tracts that showed structural abnormality acutely had further deteriorated on follow-up a year later. Moreover, additional regions that appeared normal initially also showed impaired DTI metrics. These findings suggest that white matter degeneration may continue for years beyond the initial injury (Bigler, Reference Bigler2013b). Such changes cannot be accounted for by the acute/sub-acute pathological effects of TBI, but likely reflect long-term degenerative mechanisms. Continued myelin degeneration (Meythaler et al., Reference Meythaler, Peduzzi, Eleftheriou and Novack2001), chronic neuroinflammation (Bigler, Reference Bigler2013a, Reference Bigler2013b; Johnson et al., Reference Johnson, Stewart, Begbie, Trojanowski, Smith and Stewart2013) and Wallerian degeneration as a result of axonal disconnection (Bendlin et al., Reference Bendlin, Ries, Lazar, Alexander, Dempsey, Rowley and Johnson2008) have all been postulated as possible mechanisms. It has also been postulated that disuse-related negative plastic changes may further contribute to a downward negative spiral (Tomaszczyk et al., Reference Tomaszczyk, Green, Frasca, Colella, Turner, Christensen and Green2014).

Of interest, partial recovery and continued degeneration are not mutually exclusive. For instance, over a 10-month interval, Sidaros et al. (Reference Sidaros, Engberg, Sidaros, Liptrot, Herning, Petersen and Rostrup2008) reported an increase in DTI abnormalities in the posterior corpus callosum and cerebral peduncle, consistent with progressive atrophy, but a reduction of DTI abnormalities in the posterior limb of the internal capsule and centrum semiovale. They interpreted this finding as evidence that some axonal recovery or regrowth might be possible even against the backdrop of progressive atrophy. A similar conclusion was reached by Farbota et al. (Reference Farbota, Bendlin, Alexander, Rowley, Dempsey and Johnson2012; see also Bendlin et al., Reference Bendlin, Ries, Lazar, Alexander, Dempsey, Rowley and Johnson2008). In their participants who were followed over 4 years, increases in FA were observed in the superior longitudinal fasciculus and optic radiation, concurrent with progressive FA loss in the corpus callosum. Why certain regions continue to deteriorate while others improve remains to be elucidated, but it has been speculated that both the extent of initial damage and tract-specific characteristics may play a role (Farbota et al., Reference Farbota, Bendlin, Alexander, Rowley, Dempsey and Johnson2012).

Neurocognitive Functions

The cognitive functions typically affected by TBI such as executive functions, memory and attention depend on the connectivity of distributed, yet specific, neural networks. Thus, disruption of the associated white matter connections should have systematic consequences for cognition. Consistent with this notion, studies of TBI demonstrate that particular patterns of white matter abnormalities are associated with distinct profiles of cognitive impairment.

Some of the strongest evidence tying domain-specific cognitive dysfunction to discrete white matter pathways in TBI comes from examination of frontal, frontal-striatal, and frontal-parietal networks in association with executive functions. White matter tracts including the superior longitudinal fasciculus, corona radiata, internal capsule, and cingulum bundle connect frontal with posterior regions and allow for top-down control of activity in association cortices (Catani et al., Reference Catani, Dell’Acqua and Thiebaut de Schotten2013; Miller & D’Esposito, Reference Miller and D’Esposito2005). Disruption of these white matter tracts has been associated with executive dysfunction in TBI. For instance, Kraus et al. (Reference Kraus, Susmaras, Caughlin, Walker, Sweeney and Little2007) found that reduced FA in the anterior corona radiata was associated with lower scores on a composite measure of executive functions. Kinnunen et al. (Reference Kinnunen, Greenwood, Powell, Leech, Hawkins, Bonnelle and Sharp2011) found a relationship between mean diffusivity in the left superior frontal white matter and set shifting on the Trail Making Test. Furthermore, Little et al. (2010) found that thalamic-frontal disconnection was associated with attention and executive function deficits. Although studies examining the range of severity may be ideally suited to observe such correlations, studies limited to moderate/severe (Håberg et al., Reference Håberg, Olsen, Moen, Schirmer-Mikalsen, Visser, Finnanger and Eikenes2015; Palacios et al., Reference Palacios, Sala-Llonch, Junque, Roig, Tormos, Bargallo and Vendrell2012) or mild TBI (Lipton et al., Reference Lipton, Gulko, Zimmerman, Friedman, Kim, Gellella and Branch2009; Niogi et al., Reference Niogi, Mukherjee, Ghajar, Johnson, Kolster, Lee and McCandliss2008) have reported analogous findings. For instance, in patients with acute mTBI, Lipton et al. (Reference Lipton, Gulko, Zimmerman, Friedman, Kim, Gellella and Branch2009) found that FA abnormalities in frontal white matter, including the dorsolateral prefrontal cortex, were associated with omissions on the Continuous Performance Test and the number of failed trials on the Executive Maze Test. These studies support the notion that TBI-associated disconnection of frontal–posterior and frontal–subcortical networks has negative consequences for goal-directed behavior.

TBI-associated white matter abnormalities in specific tracts have also been linked to deficits in memory and learning. The fornix carries signals from the hippocampus to the basal forebrain and diencephalon and is critical for memory performance. Post-mortem pathology studies show that this region is commonly damaged in TBI (Blumbergs et al., Reference Blumbergs, Scott, Vis, Wainwright, Simpson and McLean1995). As expected, disruption of the fornix is associated with poorer memory performance and learning (Kinnunen et al., Reference Kinnunen, Greenwood, Powell, Leech, Hawkins, Bonnelle and Sharp2011; Salmond et al., Reference Salmond, Menon, Chatfield, Williams, Pena, Sahakian and Pickard2006). Similarly, disruption of other fiber tracts that carry information to and from the temporal lobes including the posterior cingulate and uncinate fasciculus have been associated with reduced memory performance in TBI or mTBI (Bendlin et al., Reference Bendlin, Ries, Lazar, Alexander, Dempsey, Rowley and Johnson2008; Geary et al., Reference Geary, Kraus, Pliskin and Little2010; Niogi et al., Reference Niogi, Mukherjee, Ghajar, Johnson, Kolster, Lee and McCandliss2008; Salmond et al., Reference Salmond, Menon, Chatfield, Williams, Pena, Sahakian and Pickard2006).

The corpus callosum is the largest collection of white matter fibers of the brain and, as noted above, it is also one of the most frequently damaged by TBI. Given the vast projections of the corpus callosum and its role in transmitting information across hemispheres, it is not surprising that lesions of the corpus callosum are associated with a diverse array of cognitive deficits. Diffusion abnormalities in this region have been associated with reduced performance in executive functions (Jorge et al., Reference Jorge, Acion, White, Tordesillas-Gutierrez, Pierson, Crespo-Facorro and Magnotta2012; Kinnunen et al., 2011; Kraus et al., Reference Kraus, Susmaras, Caughlin, Walker, Sweeney and Little2007), memory (Palacios et al., Reference Palacios, Sala-Llonch, Junque, Roig, Tormos, Bargallo and Vendrell2012; Salmond et al., Reference Salmond, Menon, Chatfield, Williams, Pena, Sahakian and Pickard2006), and motor performance (Caeyenberghs et al., Reference Caeyenberghs, Leemans, Geurts, Taymans, Vander Linden, Smits-Engelsman and Swinnen2010; Farbota et al., Reference Farbota, Bendlin, Alexander, Rowley, Dempsey and Johnson2012).

At a general level, the extant literature examining the cognitive consequences of white matter disruptions in TBI aligns well with expectations based on understanding of normal structure-function relations. However, it may be an oversimplification to assume that the structure-function relations observed in TBI simply represent a shift to the left on the curve of normal structure-function relations. For example, Niogi and colleagues (Reference Niogi, Mukherjee, Ghajar, Johnson, Kolster, Lee and McCandliss2008) demonstrated that both the control group and the TBI group showed a positive correlation between FA in the uncinate fasciculus and memory performance, and FA in the anterior corona radiata and attentional control. However, in other cases distinct relationships have been observed between DTI metrics and neurocognitive function in TBI and in the undamaged brain (Kinnunen et al., Reference Kinnunen, Greenwood, Powell, Leech, Hawkins, Bonnelle and Sharp2011; Little et al., 2010). Of interest, Kinnunen and colleagues (Reference Kinnunen, Greenwood, Powell, Leech, Hawkins, Bonnelle and Sharp2011) found both patterns for different domains, with similar structure-function associations in TBI and healthy individuals with regard to memory, but distinct associations with regard to executive function. These findings place limits on the extent to which inferences from normal individuals can guide understanding of impairments following TBI.

Postconcussion Syndrome

Aside from neuropsychological performance, another focus of interest has been the link between white matter injury and subjective symptom reports, especially in individuals with mTBI. It is generally accepted that postconcussion symptoms (PCS) are associated with the acute neurologic effects of brain injury, and consistent with this view, studies have found an association between DTI metrics and PCS in the acute stage (e.g., Bazarian et al., Reference Bazarian, Zhong, Blyth, Zhu, Kavcic and Peterson2007). The etiology of persistent PCS, however, is an issue of considerable debate given that many symptoms are non-specific and may also be associated with chronic pain and psychiatric conditions (McCrea et al., Reference McCrea, Iverson, McAllister, Hammeke, Powell, Barr and Kelly2009). In light of this, it has been suggested that a variety of psychological, social, and motivational factors play a critical role in the maintenance of PCS, although some symptoms may be correlated with neuropathology (Bigler & Maxwell, Reference Bigler and Maxwell2012). The availability of sensitive neuroimaging measures of axonal injury has provided an opportunity to directly examine the association between traumatic axonal injury and PCS in the subacute and chronic stage.

Consistent with the notion that PCS in the subacute or chronic stage are associated with neural abnormalities, several studies have shown a correlation between the magnitude of PCS and DTI metrics in various white matter tracts (Levin et al., Reference Levin, Wilde, Troyanskaya, Petersen, Scheibel, Newsome and Li2010; Ling et al., Reference Ling, Pena, Yeo, Merideth, Klimaj, Gasparovic and Mayer2012; Smits et al., Reference Smits, Houston, Dippel, Wielopolski, Vernooij, Koudstaal and van Der Lugt2011; Yeh et al., Reference Yeh, Wang, Oakes, French, Pan, Graner and Riedy2014; but see Petrie et al., Reference Petrie, Cross, Yarnykh, Richards, Martin, Pagulayan and Peskind2014). As an example, Yeh and colleagues (Reference Yuh, Cooper, Mukherjee, Yue, Lingsma, Gordon and Vassar2014) found that postconcussive symptom severity was associated with decreased FA in the cingulum bundle, corpus callosum, corona radiata, internal capsule, and frontal fibers. The diffuse white matter abnormalities observed in this study likely reflect the wide-ranging symptoms that comprise PCS. Other studies have used a group approach to compare individuals with and without PCS; such studies have yielded mixed results, with some studies showing group differences (Bouix et al., Reference Bouix, Pasternak, Rathi, Pelavin, Zafonte and Shenton2013; Hartikainen et al., Reference Hartikainen, Wäljas, Isoviita, Dastidar, Liimatainen, Solbakk and Öhman2010; Messé et al., Reference Messé, Caplain, Pélégrini-Issac, Blancho, Montreuil, Lévy and Benali2012) and others not (Lange et al., Reference Lange, Iverson, Brubacher, Mädler and Heran2012, Reference Lange, Panenka, Shewchuk, Heran, Brubacher, Bioux and Iverson2015; Wäljas et al., Reference Wäljas, Lange, Hakulinen, Huhtala, Dastidar, Hartikainen and Iverson2014). Reasons for these inconsistencies are unclear, although studies comparing dichotomous groups may inherently have lower power to detect white matter associations than do correlational studies.

The observation of structural white matter changes associated with PCS reports beyond the acute stage can be interpreted in several ways. On the one hand, it is consistent with the notion that persistent PCS reflect direct neurological effects of mTBI. On the other hand, equally plausible is the possibility that TBI-associated white matter abnormalities play a role in the development or maintenance of mental health conditions including PTSD (Bazarian et al., Reference Bazarian, Donnelly, Peterson, Warner, Zhu and Zhong2013; Costanzo et al., Reference Costanzo, Chou, Leaman, Pham, Keyser, Nathan and Roy2014) and depression (Matthews et al., Reference Matthews, Strigo, Simmons, O’Connell, Reinhardt and Moseley2011), which in turn may be associated with chronic PCS. To date, there are no longitudinal studies examining the link between white matter, PCS, and the development of psychopathology. Such studies may be helpful in further probing the relationship between axonal injury, psychiatric symptoms, and PCS at different time frames following TBI. In this context, consideration of separate clusters of PCS will be important, as such clusters may have a distinct origin and longitudinal course.

Default Mode Network

A strong focus of functional connectivity studies in TBI has been on the DMN. This network is critically involved in internal thought, when external cognitive demands are minimized as during “resting” state scanning, but needs to be inhibited during tasks that require attention to the external environment. Thus, its integrity is critical to cognitive task performance. Because the DMN involves long-coursing parietofrontal and interhemispheric projecting axons, it is vulnerable to TBI. Several studies have demonstrated altered functional connectivity within the DMN, both in moderate/severe and in mild TBI, and such alterations have been linked to the presence of postconcussion symptoms (Messé et al., Reference Messé, Caplain, Pélégrini-Issac, Blancho, Lévy, Aghakhani and Lehéricy2013; Stevens et al., Reference Stevens, Lovejoy, Kim, Oakes, Kureshi and Witt2012; Zhou et al., Reference Zhou, Milham, Lui, Miles, Reaume, Sodickson and Ge2012) and cognitive changes (Bonelle et al., Reference Bonelle, Leech, MKinnunen, Ham, Beckmann, De Boissezon and Sharp2011; Sharp et al., Reference Sharp, Beckmann, Greenwood, Kinnunen, Bonelle, De Boissezon and Leech2011). In patients with moderate to severe TBI, increased functional connectivity between core nodes of the DMN is commonly observed, and especially with the posterior cingulate and precuneus regions (Hillary et al., Reference Hillary, Rajtmajer, Roman, Medaglia, Slocomb-Dluzen, Calhoun and Wylie2014; Palacios et al., Reference Palacios, Sala-Llonch, Junque, Roig, Tormos, Bargallo and Vendrell2013; Sharp et al., Reference Sharp, Beckmann, Greenwood, Kinnunen, Bonelle, De Boissezon and Leech2011; but see Arenivas et al., Reference Arenivas, Diaz-Arrastia, Spence, Cullum, Krishnan, Bosworth and Marquez de la Plata2014; Pandit et al., Reference Pandit, Expert, Lambiotte, Bonelle, Leech, Turkheimer and Sharp2013). However, there is also evidence that TBI is associated with decreased connectivity in some nodes of the DMN and increased connectivity in other nodes (Venkatesan et al., Reference Venkatesan, Dennis and Hillary2015), possibly reflecting the distinct role of different nodes within the network. Similar non-uniformity of network changes has been observed in mild TBI (Johnson et al., Reference Johnson, Zhang, Gay, Horovitz, Hallett, Sebastianelli and Slobounov2012; Stevens et al., Reference Stevens, Lovejoy, Kim, Oakes, Kureshi and Witt2012; Zhou et al., Reference Zhou, Milham, Lui, Miles, Reaume, Sodickson and Ge2012), while other studies in mild TBI have observed generally decreased connectivity (Costanzo et al., Reference Costanzo, Chou, Leaman, Pham, Keyser, Nathan and Roy2014; Mayer et al., Reference Mayer, Mannell, Ling, Gasparovic and Yeo2011). Alterations in the DMN have been linked to structural damage to the cingulum bundle (Costanzo et al., Reference Costanzo, Chou, Leaman, Pham, Keyser, Nathan and Roy2014; Palacios et al., Reference Palacios, Sala-Llonch, Junque, Roig, Tormos, Bargallo and Vendrell2013) and corpus callosum (Sharp et al., Reference Sharp, Beckmann, Greenwood, Kinnunen, Bonelle, De Boissezon and Leech2011).

Aside from alterations in the DMN itself, there is considerable evidence for impaired coordination between the DMN and task-positive networks such as the executive control network (Sours et al., Reference Sours, Zhuo, Janowich, Aarabi, Shanmuganathan and Gullapalli2013; Venkatesan et al., Reference Venkatesan, Dennis and Hillary2015) and salience network (Sours et al., Reference Sours, Zhuo, Janowich, Aarabi, Shanmuganathan and Gullapalli2013). The salience network, in particular, has been shown to play an important role in shifting between internally and externally directed states, and as such, may be critical for balancing network function necessary for efficient cognition. Focusing specifically on interactions between the salience network and DMN during tasks requiring motor control, Sharp and colleagues (Bonelle et al., Reference Bonelle, Ham, Leech, Kinnunen, Mehta, Greenwood and Sharp2012; Jilka et al., Reference Jilka, Scott, Ham, Pickering, Bonelle, Braga and Sharp2014) found that patients with TBI-associated cognitive control problems failed to suppress DMN activity during task performance (Bonelle et al., Reference Bonelle, Ham, Leech, Kinnunen, Mehta, Greenwood and Sharp2012), and that this failure was associated with an inability to transiently upregulate functional connectivity between these networks (Jilka et al., Reference Jilka, Scott, Ham, Pickering, Bonelle, Braga and Sharp2014). Furthermore, the breakdown of this network interaction was related to the extent of structural damage within the salience network, specifically in a white matter tract connecting the right anterior insula with the pre-SMA/dorsal anterior cingulate.

Graph Theory

Recently, whole-brain connectivity patterns have also been analyzed using graph theoretical analyses that characterize networks graphically as a set of interacting nodes (groups of neurons) connected by edges (white matter tracts), and thus more closely resemble the distributed network architecture of the brain. Brain networks are characterized by small-world topology, an architecture in which densely connected local modules are combined with sparse long distance connections that allow for communication across modules. Several studies have demonstrated a reduction of small-world properties in TBI, including decreased network efficiency (Nakamura et al., Reference Nakamura, Hillary and Biswal2009; Pandit et al., Reference Pandit, Expert, Lambiotte, Bonelle, Leech, Turkheimer and Sharp2013), modularity (Messé et al., Reference Messé, Caplain, Pélégrini-Issac, Blancho, Lévy, Aghakhani and Lehéricy2013), overall connectivity (Pandit et al., Reference Pandit, Expert, Lambiotte, Bonelle, Leech, Turkheimer and Sharp2013) and between-network connectivity (Han et al., Reference Han, Mac Donald, Johnson, Barnes, Wierzechowski, Zonies and Brody2014). These alterations are consistent with disruption of long coursing fibers as a result of traumatic axonal injury. However, as noted above, increased local connectivity early in the course of recovery, especially in core networks known as the “rich club”, has also been documented (Hillary et al., Reference Hillary, Rajtmajer, Roman, Medaglia, Slocomb-Dluzen, Calhoun and Wylie2014). Evidence is also emerging suggesting that graph metrics relate to behavioral impairment (Caeyenberghs et al., Reference Caeyenberghs, Leemans, Heitger, Leunissen, Dhollander, Sunaert and Swinnen2012) and may be predictive of response to cognitive training (Arnemann et al., Reference Arnemann, Chen, Novakovic-Agopian, Gratton, Nomura and D’Esposito2015). However, which metrics will be most informative in characterizing alterations in brain structure and function remains to be determined.

Network Analysis of Injury Recovery

Given the dynamic changes during recovery following TBI, the time post-injury when network analysis is performed is critical to understanding the neuropsychological significance of findings. To date, few studies have charted longitudinal changes in network function. Hillary et al. (Reference Hillary, Slocomb, Hills, Fitzpatrick, Medaglia, Wang and Wylie2011) demonstrated in moderate/severe TBI that during a critical period of recovery, between 3 and 6 months post-injury, resting-state connectivity increased in the DMN and decreased in the fronto-parietal control network (Hillary et al., Reference Hillary, Slocomb, Hills, Fitzpatrick, Medaglia, Wang and Wylie2011). Moreover, connectivity between each of these networks and the salience network increased, potentially allowing for more efficient transition between networks. In a more recent study from the same group (Venkatesan et al., Reference Venkatesan, Dennis and Hillary2015), patients with moderate/severe TBI assessed during the subacute period were scanned again on average 2 years post-injury. The authors observed reduced DMN connectivity from the posterior cingulate and increased anti-correlations with task-positive regions, consistent with normalization of network connectivity (see also Nakamura et al., Reference Nakamura, Hillary and Biswal2009). However, at the same time, DMN hyperconnectivity with hippocampal nodes became more pronounced. The significance of this finding is unclear, although the authors suggested a potential compensatory mechanism. However, these findings clearly illustrate the complexity of network changes and challenge the notion of uniform recovery.

Another factor that may contribute to the complex pattern of network dysfunction is the heterogeneous pathology in TBI and the extent to which focal injury may interact with diffuse white matter injury. In this regard, damage to critical hubs that link different networks may have far-reaching consequences that go well beyond the function of any particular network, and may be an important mechanism underlying long term cognitive impairment (Fagerholm et al., Reference Fagerholm, Hellyer, Scott, Leech and Sharp2015).