ical examination demonstrated hemosiderin (a blood by-product)-laden macrophages and lymphocytes in the white matter (WM). Obviously, this finding suggests perturbation of brain microvasculature and WM injury that was well below the detection of the “normal” CT. Such microscopic lesions are undoubtedly the basis of many neurobehavioral sequelae associated with brain injury when imaging is “nor-mal.” This is further supported by the work of Gorrie et al.
(2001) who examined 32 children at postmortem who suc-cumbed to road accidents. With direct visual inspection, 17 of these TBI cases demonstrated no macroscopic abnormal-ities of the type that would be detected by CT imaging.
However, when viewed at ×100 magnification, all cases readily demonstrated microscopic injury.
readily identified, quantified (either as volumes or surface areas), and compared to a normative sample (Bigler 1999).
The table in Appendix 5–1 summarizes regions that have been shown to exhibit atrophy in response to trauma.There is extensive clinical literature on the use of MR in the acute and subacute diagnosis and management of TBI (Atlas 2001; Gean 1994; Orrison 2000), but as indicated above, with regard to neuropsychiatric morbidity abnormalities identified in the chronic stage typically have better correla-tion with outcome than the acute or sub-acute findings (Henry-Feugas et al. 2000; Jorge et al. 2004; van der Naalt et al. 1999; Vasa et al. 2004; Wilson et al. 1988). Accord-ingly, the primary focus of the remainder of this chapter is MR imaging performed more than 45 days postinjury so that the more stable and chronic lesions can be related to neurobehavioral deficits, particularly those resulting in neuropsychiatric sequelae.
Indications
There is a multitude of reasons for performing MR imaging in the TBI patient, but typically the reasons center on mon-itoring the status of the patient, often during the subacute and more chronic phases of recovery. For example, because of its capacity for exquisite anatomic detail and detection of water, MR is suitable for monitoring edema, midline shift,
and the changing status of a hemorrhage and for evaluating lesions that may underlie posttraumatic epilepsy. It is also helpful in the clinical correlation of the patient’s acute status, as depicted in Figure 5–7, and the structural imaging. The patient shown in this figure had normal CT reading on admission but was in a coma (GCS score of 5). MR imaging performed later on the DOI was also read as “normal”; how-ever, the MR scan performed 4 days later clearly demon-strated the beginnings of significant degenerative changes, including areas of shearing that were not definitively observed on the DOI CT or MR scan. Another reason for MR imaging is to monitor changes over time, which is important because the degeneration often takes months to reach an endpoint. Blatter et al. (1997) demonstrated that the time that elapses between injury and brain volume stabi-lization equivalent to that expected with normal aging may be more than 3 years, although most pathological changes occur within the first 6 months. Thus, acute and subacute MR imaging is performed to assess potentially medically treatable abnormalities associated with brain trauma, track degenerative changes that occur with time, and relate imag-ing findimag-ings to neurobehavioral sequelae.
As indicated in the section Computed Tomography Imaging, often all early and subacute neuroimaging is done with CT, particularly with patients on life support, due to the incompatibility of life-support equipment with F I G U R E 5 – 6 . Ventriculomegaly in traumatic brain injury (TBI).
Hydrocephalus ex vacuo is a common sequela of brain injury and is often proportional to the severity of injury. The top row shows a frontal view based on three-dimensional magnetic resonance renderings of the brain, with the visible ventricular system depicted in black. The bottom row represents the lateral view: the image on the left is from a noninjured control subject, the image in the middle is from a moderately injured TBI patient, and the image on the right is from a subject with severe brain injury. It is important to note that it is the entire ventricular system that typically dilates, indicating the diffuse nature of impact brain injury. By taking the volume of the ventricular system, as shown in black, and dividing it by the volume of the brain, a ventricle to brain ratio (VBR) can be calculated. Increasing VBR is a sign of increasing cerebral atrophy. Typically, increased VBR is associated with worse outcome (see Figure 5–11 and Ariza et al. 2004).
F I G U R E 5 – 7 . Comparison of similar sagittal magnetic resonance (MR) images to demonstrate injury and subsequent atrophy to the corpus callosum at different stages postinjury.
The midsagittal day-of-injury MR scan (A, top left) was taken on admission to the hospital after the patient sustained a severe TBI.
Some movement artifact diminished the quality of the image but was interpreted as within normal limits. However, within 1 week postinjury (B), signal intensity changes are clearly visible in the corpus callosum both anteriorly (black arrow) as well as posteriorly.
At 4 years postinjury (C), corpus callosum atrophy is clearly evident and is generalized including all aspects (compare the original size of the corpus callosum in A with that observed in C). Generalized atrophy is also noted by the dark signal, especially seen in the frontoparietal aspects of the midsagittal view of C, indicating increased cerebrospinal fluid (CSF) in the space of the interhemispheric fissure, a sign of reduced brain volume (note that brain parenchyma in A and B is light gray, but a dark signal covers the midsagittal surface in C because of increased CSF in these regions secondary to atrophy). Also, as clearly visible (white arrow in C), a major shear lesion is evident where most of this segment of the corpus callosum has been transected. For better clarification of this lesion involving the corpus callosum, the injured corpus callosum has been enlarged and highlighted in D. When viewing A (the day-of-injury scan) in retrospect, there is some signal change noted in the region that eventually shows the shear lesion. The colorized images in E, F, and G are all from diffusion-tensor imaging sequences in which tractography involving the projections of the corpus callosum in a noninjured subject is displayed (Lazar et al. 2003). The images are color-coded on the basis of their projection (i.e., red shows frontal projection). In E, the diffusion scan on the left is depicted in the axial plane, which shows the projections across the corpus callosum from this perspective. The scan to the right in E is from the injured patient. In F, the colorized projections are shown in the midsagittal view. Accordingly, by comparing the view of the location of the lesion in D with the view in F, one can see that this injury would result in disrupted projections in primarily the midfrontal region. G shows the tractography plots mapped through the corona radiata. The vertical line in E is the approximate location of these maps that depict the hemispheric projections of callosal white matter fiber tracks.
Source. Diffusion-tensor imaging tractography color images courtesy of Mariana Lazar, Ph.D., and Andrew Alexander, Ph.D., Uni-versity of Wisconsin, Madison.
the MR imaging environment. It is helpful to compare baseline CT images with follow-up MR images, as dem-onstrated in Figures 5–3, 5–10, and 5–12.
Typical Lesions Identified by Magnetic Resonance Imaging
More details concerning the neuropathology of TBI are presented in Chapter 2, Neuropathology. For the purposes
of this discussion, just a brief overview of the neuropathol-ogy observed in MR imaging of the brain in TBI is offered, but the reader should be aware that a multitude of pathol-ogies exist that can be detected by MR imaging (Atlas 2001;
Gean 1994; Orrison 2000). The typical lesions described below are the ones most commonly observed to relate to significant neuropsychiatric sequelae (Bigler 2001b) and most commonly occur because of the greater likelihood of frontotemporal damage (see Figure 5–8). Table 5–2 is F I G U R E 5 – 8 . Voxel-based morphometry (VBM) in traumatic brain injury (TBI).
VBM provides a method to simultaneously compare––voxel-by-voxel––where the major differences occur in subjects with TBI com-pared with age-matched control subjects without damage. In this figure, by using three-dimensional (3D) magnetic resonance (MR) imaging, the diffuseness of frontotemporal involvement can be more fully appreciated (shown in red) when TBI subjects who had sustained frontotemporal contusions are compared with control subjects by using VBM techniques; the differences (i.e., regions of reduced voxel density of either gray or white matter) are plotted on a standard 3D surface plot of the brain. VBM was applied to MR imaging performed on 6 subjects (mean age = 16; standard deviation = 5.1) with moderate-to-severe TBI (all had Glasgow Coma Scale scores at or below 8) compared with 18 control subjects (3 control subjects within 2 years per TBI patient). Young subjects were selected to minimize any long-term age effect that could potentially relate to volume reduction. The VBM findings (A) distinctly demonstrate extensive frontotemporal differences in the TBI subjects, particularly in the ventral frontal region, more so in gray matter than white. Given the ventral basis of the changes seen in this illustration, the basal forebrain (slanted white arrow, control subject, sagittal view, lower right)––including the region involving the anterior commissure (AC), a thin white matter band critical for white matter interhemispheric connections, as shown in B––was also quantified and compared with the control subjects. Quantitatively, the basal forebrain region demonstrated over a 15% reduction in volume in the TBI subjects, who also were found to have significantly reduced AC widths of 2.00 mm (SD = 0.44) compared with control subjects, in whom the mean width was 3.18 mm (SD = 0.40). In the TBI subject presented in B, the AC width was 1 mm compared with an age-matched control subject whose AC width was 3.5 mm.
The blue arrow identifies the location of the AC, and the conjoined white arrows show where shear injuries occurred in the TBI subjects, leaving regions of cavitation in the basal ganglia and internal capsule. In the sagittal view, the control subject’s AC (B, lower right) is clearly visible (vertical white arrow), whereas the AC is almost not discernible in the sagittal view of the TBI patient (lower left). Note also the thinness of the corpus callosum in the TBI patient, another reflection of generalized injury.
offered as a guide to integrating MR imaging findings using standard imaging sequences (i.e., T1, T2, fluid-attenuated inversion recovery [FLAIR], gradient recalled echo [GRE]) in detecting abnormalities associated with TBI. The image sequences depicted in Table 5–2 based on one patient with severe TBI 1 year postinjury demonstrates how different image sequences identify structural pathol-ogy. Tong et al. (2004) and Goetz et al. (2004) have clearly demonstrated how certain clinical sequences may simply be insensitive in detecting structural pathology and rein-force the recommendation to use multiple sequences to increase the likelihood of detecting clinically significant abnormalities caused by brain injury. The key in integrat-ing scans is to look for changes in symmetry or differences in signal intensity in comparison to normal tissue. By using Table 5–2, where normal appearance is summarized, detec-tion of pathology can often be readily made. However, it must be emphasized that the information offered in Table 5–2 can change with certain scan parameters; therefore, these findings are not absolutes.
The traditional T1 image is most useful for establish-ing the presence of focal atrophy. The combination of T1 and T2 imaging is best in establishing ventricular and cerebrospinal fluid (CSF) changes. The GRE sequence is often excellent in detecting hemosiderin changes, whereas the FLAIR and proton density (PD) sequences
may be more sensitive to general WM pathology, as may different types of DW imaging. Because there is so much that can be done clinically with MR imaging, it is best that the clinician work closely with the neuroradiologist in attempting to identify clinically useful protocols for imaging patients with TBI.
Shear Injury
The CC is a structure in which shearing due to TBI fre-quently occurs (Johnson et al. 1994; Levin et al. 2000). In the patient shown in Figure 5–7, there is literally a tear in the anterior aspect of the CC. When shearing occurs out-side of the CC, it is most frequently observed at the junc-tion of WM and gray matter, particularly in the frontal and temporal regions. Because the tensile forces that are sufficient to shear axons are also sufficient to shear blood vessels, sites where axonal shearing is suspected are often also sites where hemosiderin deposits are detected.
Detection of such abnormalities is also dependent on the image sequence, as shown in Figure 5–8.
Contusion
Contusion most commonly occurs where bony ridges (i.e., the sphenoid) or protuberances (i.e., crista galli) are located. Acutely, these lesions may also be associated
T A B L E 5 – 2 . Appearance of magnetic resonance (MR) images based on the type of image sequence
Typical MR imaging sequences for detecting TBI abnormalities
Additional MR imaging sequences for evaluating TBI
T1 T2 FLAIR T2 GRE PD DW imaging
See Figure 5–5 See Figure 5–13
Appearance of cerebrospinal fluida
Low High Dark Medium gray Isointense Dark
Appearance of edemaa
Low High Bright Light gray High Bright
General appearance of an
abnormalitya
Low to black High Bright unless CSF or hemosiderin
Bright or dark depending
High Bright
Hemosiderin Darker Darker Darker Darkest Dark Darkest
Air Signal loss Signal loss Signal loss Signal loss Signal loss Signal loss
Note. CSF=cerebrospinal fluid; DW=diffusion-weighted; FLAIR=fluid-attenuated inversion recovery; GRE=gradient recalled echo; PD=proton density.
aCompared with normal adult brain parenchyma.
with focal edema. Acute contusions may resolve, leaving no detectable abnormality on MR imaging. This cir-cumstance represents another case in which it is impor-tant to have the DOI information, because an acute con-tusion most likely results in damaged parenchyma, regardless of the MR imaging findings. As with shear injuries, sites of contusion often reveal hemosiderin deposits (Figure 5–9).
White Matter Signal Abnormalities
Due to the susceptibility of WM to injury in TBI, small, subtle, but nonetheless detectable WM abnormalities may show up as either WM hyperintensities and/or depo-sition of hemosiderin, as already mentioned in the section Shear. These areas of WM damage often correspond to areas where petechial hemorrhages have been noted on DOI CT imaging (see Table 5–2 and Figure 5–9). A sim-ple WM-hyperintensity rating method, easily used by the clinician, is offered in the section Clinical Rating of Scans and Relationship to Neurobehavioral Changes at the end of this chapter.
Focal Atrophy
A variety of trauma factors may coalesce to produce focal atrophy in particular regions of the brain, most commonly in the frontal and/or temporal lobes. This situation is demon-strated in Figure 5–10. A simple clinical rating method for establishing frontal and temporal lobe atrophy is offered in the section Clinical Rating of Scans and Relationship to Neurobehavioral Changes at the end of this chapter. This rating method can be quickly applied by the clinician; the presence of atrophy established by this method is associated with deficits in memory and executive function.
Quantitative Magnetic Resonance Neuroimaging
A most fortuitous circumstance exists at the gross structural level of brain parenchyma—it is comprised of two general tissue types, namely gray matter and WM. Gray matter, composed mostly of cell bodies and dendritic trees (where F I G U R E 5 – 9 . Cortical contusion as seen in the
acute stage (A) and chronic stage (B).
The contusion developed around the sphenoid bone and was caused by the brain parenchyma’s grating against the sphenoid ridges, shearing blood vessels as well as macerating tissue. The density changes in this posterior region of the frontal lobe on the day-of-injury (DOI) computed tomography (CT) scan show a mixture of blood and edema, which also extends into the peri-Sylvian region of the brain. The chronic lesion resulting from this focal injury is seen in B through the use of magnetic reso-nance (MR) imaging. The lesion shows an area of greater cere-brospinal fluid collection, which means loss of parenchymal integrity and atrophy as well as hemosiderin, the dark ring around the lesion site representing old, degraded blood by-products. Of interest is the fact that the temporal lobe contusion aspect of the lesion seen on the DOI scan does not clearly image on the MR scan. When such lesions “resolve,” the clinician should not assume that surrounding tissue is not affected––the DOI CT scan suggests that it is likely that the temporal lobe is more generally damaged at this level but that damage is not de-tected by the MR imaging done during the chronic phase.
F I G U R E 5 – 1 0 . Demonstration of frontal contu-sions and intraparenchymal hemorrhaging that re-sulted in focal bifrontal atrophy after a high-speed motor vehicle–pedestrian accident.
The illustration also demonstrates the progression of pathology in brain injury from the day-of-injury computed tomography (CT) scan (A), to the CT scan at 4 months postinjury (B), to 2 years postinjury, as shown in magnetic resonance (MR) imaging findings (C). Note the ventricular expansion and the better def-inition and more extensive pathology identified by MR imaging during the chronic phases (2 years postinjury).
A
B
C
synapses are located)—the neuropil—and WM, composed mainly of myelinated axons, yield different signal character-istics on MR imaging. These dissimilar signal intensities permit their isolation, and therefore gray matter and WM can be “segmented” from one another (Laidlaw et al. 2000).
Likewise, because CSF spaces are fluid filled, they too have different signal characteristics from brain parenchyma, as does bone. Once these different tissue-CSF compartments are segmented, accurate estimates of the volume of any region of interest can be made because the slice thickness of the scan and the distance between slices are known (Bigler and Tate 2001). Because contemporary MR imaging has resolution to approximately 1 mm, fine structural analysis can be achieved of any region that can be visualized with gross inspection of the brain. As already mentioned in the section Magnetic Resonance Imaging, numerous areas have been quantitatively analyzed and shown to degenerate in response to brain trauma (see Appendix 5–1 for a partial list-ing). In fact, inspection of this table demonstrates the non-specific susceptibility of the brain to traumatic injury and, as discussed below, typically the generalized nature of TBI is in proportion to the severity of the injury. Even mild TBI may show qualitative and quantitative changes (Hofman et al.
2001; McGowan et al. 2000).
Global Atrophy Associated With TBI
Moderate-to-severe TBI, defined by a GCS score of 12 or lower, has been shown to be associated with nonspecific volume loss of brain parenchyma (see Appendix 5–1).
Because the CSF housed within the ventricle is under pres-sure, any loss of brain volume results in a passive expansion of the ventricular system (i.e., hydrocephalus ex vacuo) (see Figure 5–6). A straightforward method to demonstrate this quantitatively comes through the use of the ventricle to brain ratio (VBR). This ratio is the total volume of the ven-tricles (lateral, III, and IV) divided by the total brain vol-ume. Because there are inherent differences in head and body sizes (as well as types), the comparison of different patients with a single measure requires a correction for head-size differences. This is automatically accounted for by the VBR. VBR, or increasing atrophy, is directly related to the severity of injury, as manifested by duration of unconsciousness or posttraumatic amnesia.
Regardless of the method used to determine injury se-verity, increasing severity of injury results in greater brain volume loss and ventricular dilatation (see Figure 5–6). In-creased VBR in the TBI patient is reflective of global changes but may disproportionately reflect WM volume loss compared to that of gray matter (Adams et al. 2000;
Gale et al. 1995; Garnett et al. 2000; Strich 1956; Thatcher et al. 1997). This is particularly evident when viewing
changes in the CC (see Figure 5–7). Figure 5–6 shows a three-dimensional comparison of the ventricular systems of a noninjured control, a patient with moderate TBI, and a patient with severe injury. It is obvious in viewing these figures that the ventricular dilatation is nonspecific, affect-ing all aspects of the ventricular compartment––a reflec-tion of global atrophy induced by TBI.
Quick Guide to Visualizing Atrophy for the Clinician
Although neuroimaging is rapidly moving toward auto-mated image analysis systems, another decade will likely pass before quantitative information is routinely included in the neuroimaging report. Likewise, the typical clinician is not equipped with the hardware and software for image analysis, so how can he or she visualize atrophy? As implied in the sec-tion Global Atrophy Associated With TBI, visually inspect-ing scans over time often permits the identification of cere-bral atrophy by comparing the size of the ventricle; in particular, the DOI scan may be compared to scans done weeks or months later. Another way to examine atrophy, if sequential MR imaging has been performed, is to view the CC in midsagittal view. The CC is susceptible to atrophic change because it houses the long, coursing, interhemi-spheric WM-fiber pathways and often is directly injured by shearing action or secondary degeneration due to cortical injury, particularly contusions (see Figure 5–7). Because the CC is organized in an anterior-posterior fashion, when greater atrophy is noted regionally, that is often a sign of more atrophy in a particular lobe (i.e., atrophy of the genu associated with frontal atrophy). In contrast, degeneration of the entire length of the CC is most likely a sign of general-ized, nonspecific WM change secondary to trauma. Several studies have shown modest relationships between CC atro-phy and neurobehavioral sequelae, particularly changes in memory (Johnson et al. 1996; Levin et al. 2000). Last, simple rating methods for lobar atrophy and WM changes may be helpful in identifying MR-detected pathology. These meth-ods are more fully discussed in the section Clinical Rating of Scans and Relationships to Neurobehavioral Changes at the end of this chapter, after additional MR pathology findings in TBI are discussed.
Relationship of Magnetic Resonance Imaging Findings to Outcome
There is no simple answer or review that can be offered on the topic of the relationship of MR imaging findings to outcome (Bigler 2000, 2001a, 2001b). There are mul-tiple reasons for this complexity, including the very nature of what it means to be human and have a brain that
controls and regulates all facets of human behavior.
Accordingly, such individual factors as age, sex, educa-tion, individual differences in intellectual and cognitive abilities, health status at the time of injury, and trauma variables, including lesion location, diffuse injury effects, and presence of secondary injury effects (e.g., hypoxemia, edema, and systemic injury), all enter into the equation that predicts outcome from injury. Relating findings from
brain imaging to neuropsychiatric outcome also depends on what outcome measurements are used and when dur-ing the postinjury time period assessments are made.
Nonetheless, taking all these factors into consideration, there is the expected relationship that the greater the residual structural abnormality, the greater the potential for neuropsychiatric morbidity. This relationship can be seen in Figure 5–11, which demonstrates outcome F I G U R E 5 – 1 1 . Box plots demonstrating the relationship between generalized atrophy measured by the ventricle to brain ratio (VBR) and discharge status from in-patient rehabilitation (Rehab) using the Functional Independence Measure (FIM) and the Disability Rating Scale (DRS).
Normal VBR is approximately 1.5. Clearly, presence of increased cerebral atrophy was associated with greater disability. See footnotes 1 and 2 (p. 81) for an explanation of the DRS and FIM.
assessed at the time of discharge from the rehabilitation unit after TBI (the modal patient had a moderate TBI with GCS score of approximately 8) compared to the late MR imaging findings. As can be seen from this figure, increasing cerebral atrophy, meaning increased nonspe-cific effects of the brain injury, was associated with greater disability at the time of discharge from the rehabilitation unit.
As for an even more long-term outcome, research suggests that the more prevalent the structural abnormal-ities, the greater the neuropsychiatric disability (Bigler 2001a; Jorge et al. 2004; Vasa et al. 2004). There is an-other factor that must be mentioned when discussing out-come: the relationship between significant head injury and the aging process. If brain injury results in atrophy and if brain volume loss also occurs with aging, then age effects in the injured brain may start from different base-lines depending on the age of the patient. This combina-tion may result in less-than-optimal aging (i.e., increased cognitive deficits with aging), increasing the likelihood of neurobehavioral sequelae, including affective disorder (Holsinger et al. 2002) and an earlier age of dementia on-set (Guo et al. 2000; Plassman et al. 2000). Because the hippocampus is one of the structures more vulnerable to injury and is the limbic structure most often implicated in degenerative diseases, it seems reasonable that there is likely a connection.
Many of the studies listed in Appendix 5–1 exam-ined the relationship of quantitative imaging to long-term outcome. Because one of the most frequent cog-nitive sequela to be associated with TBI is impaired memory, various quantitative studies (see Appendix 5–
1) have examined temporal lobe structures and mem-ory in TBI patients. In a detailed analysis of the tempo-ral lobe, Bigler et al. (2002a) demonstrated that changes in WM integrity and volume loss of the hippocampus were the sequelae most related to memory deficits after TBI.
Small but Critical Lesions
There are dedicated pathways in the brain, such as the cor-ticospinal pathway, that have little capacity for adaptation, rerouting, or functional reorganization after significant injury. Accordingly, a small but strategically placed lesion in the internal capsule may produce hemiplegia due to direct injury to the corticospinal tract. For example, the child shown in Figure 5–3 with a right internal capsule–
basal ganglia hemorrhagic shear lesion had a dense hemi-plegia, whereas the patient shown in Figure 5–10, who had massive hemorrhagic lesions bifrontally with concomitant focal frontal atrophy, did not have paralysis. Small but dev-astating lesions may also disrupt the integrity of the limbic
system, where a small lesion of the fornix or fornical atro-phy may be responsible for significant memory deficits (Blumbergs et al. 1994; Tate and Bigler 2000). This situa-tion is shown in Figure 5–7, in which it is clearly visible that the fornix progresses through various degenerative stages postinjury. The hippocampus––another relatively small structure and the origin of the majority of WM pathways that make up the fornix—is also particularly vulnerable to injury that also leads to memory impairment (Tate and Bigler 2000). Small temporal lobe lesions, including those of the hippocampus, may be the source of posttraumatic epilepsy (Diaz-Arrastia et al. 2000). It may also be that small, nonspecific lesions detected by MR imaging are the basis of the relationship between head injury and dementia, as even mild injury increases the risk ratio for dementia (Guo et al. 2000; Plassman et al. 2000).
Functional Lesion Likely Larger Than Structural Lesion
Figure 5–12 depicts the structural injuries sustained by a construction worker in a fall. Acute CT imaging demon-strated the presence of hemorrhagic lesions and midline shift that ultimately resulted in focal right frontal and tem-poral atrophy that was quite extensive (shown in red).
However, when the structural MR imaging was integrated with single-photon emission computed tomography (SPECT), the physiological abnormality could be seen to extend far beyond the boundaries of the focal structural lesions observed on the MR scan; the MR-SPECT scan actually shows a left frontal defect with no concomitant structural abnormality (see Umile et al. 2002).