Applied Radiology

Dr. Tong is an Assistant Professor in the Department of Radiology, Loma Linda University School of Medicine, Loma Linda, CA.

Trauma continues to be the number one cause of death in pediatric and young adult populations in the United States. ¹ However, with medical advances that have resulted in decreased mortality, ² there is a rise in the morbidity that follows after surviving traumatic brain injury (TBI). This increased morbidity not only affects the victim, but also the victim's family and, ultimately, society. This creates dilemmas in determining aggressiveness of management--a problem that is compounded by the facts that the clinical course can be complex and prognosis is often difficult to predict.

Head injury can entail a variety of initial injuries, including: fractures; epidural, subdural, and subarachnoid hemorrhage; contusions; and diffuse axonal injury. Secondary injuries, such as infarction, can develop as a result of mass effect/herniation or vascular injury. Computed tomography (CT) is the most important first-line tool to detect lesions that require emergent neurosurgical intervention, such as depressed skull fractures, large hematomas, or massive brain edema. However, other significant trauma-related pathology, such as diffuse axonal injury (DAI) and infarction, are better evaluated with magnetic resonance imaging (MRI), which is emerging as an important second-line tool in many institutions. In particular, DAI is usually underdiagnosed in an initial evaluation, but may be the most devastating injury in a surviving patient. Since it is often invisible on conventional CT or MRI, DAI has been appropriately called a "stealth pathology." ³ It is also evident that TBI is not a static event, but is in fact a progressive injury that can be complex at the cellular level. Newer imaging methods now permit detection of subtle pathology that may have serious prognostic implications.

The emerging technologies that can aid in the evaluation of TBI involve various modalities including positron emission tomography (PET), CT, and MRI. For example, subtle microhemorrhages are now visible with MR susceptibility-weighted imaging (SWI). Chemical markers of injury are detected on MR spectroscopy, even when the brain appears normal on conventional images. Nonhemorrhagic shearing injuries and embolic infarcts may be seen only with MR diffusion-weighted imaging (DWI). Disturbances in blood flow can be measured with PET, CT perfusion, or MR perfusion. Finally, metabolic and cognitive dysfunction can be detected with PET and functional MRI, respectively. Imaging is no longer limited to acute triage. The combination of technologies has expanded the role of the diagnostic radiologist to allow a series of functions in the course of a patient's care: determining initial severity of injury, monitoring acute progression of injury, evaluating secondary injury, predicting clinical outcome, and monitoring recovery. In the next sections, this article will review several of the newer MR techniques that have improved the evaluation of TBI.

Diffusion-weighted imaging

MR DWI is available on most clinical medium- and high-field MR scanners. Diffusion-weighted imaging is based on the ability to detect water mobility (diffusion) at a cellular level. Water within different tissues has differing mobility or diffusion. Mobility is faster along certain pathways (eg, fiber tracts) and can be altered by the gradient direction and strength of an applied magnetic field, thereby resulting in the perceived unevenness ("anisotropy") of water diffusion in normal brain tissue. Acute cellular injury results in decreased diffusion as well as decreased anisotropy. The principle of DWI can be used in two methods: conventional DWI and diffusion-tensor imaging (DTI). Both have been applied to TBI evaluation, although manifestations and utility are different. For more detailed descriptions of the principles of DWI, the reader is encouraged to review more comprehensive texts. ^4-6

Certain types of pathology can result in restricted diffusion, the classic of which is acute infarction. The simplistic explanation is that cell death results in cellular swelling ("cytotoxic edema") and impaired water mobility­­both intracellularly as well as extracellularly. However, the complex molecular mechanisms behind this phenomenon continue to be the source of intensive debate and research. It is also evident that this phenomenon is no longer exclusive to infarction, as more lesions with diffusion restriction are continually being reported in the literature, including abscesses, hemorrhage, some cellular neoplasms, Creutzfeld-Jakob disease, and some acutely demyelinating lesions, to name a few.

In the last few years, traumatic lesions associated with acute DAI have also been added to the list. These lesions (Figure 1) appear bright on DWI, with corresponding decreased apparent diffusion coefficients (ADC). ^7-11 Although these lesions may resemble small embolic infarctions on DWI, investigators have shown that the lesion location and frequent association with small shearing hemorrhages is more typical of DAI than embolic etiology. The forces associated with tearing axons could also disrupt the microvascular architecture and result in cell death. Whether we choose to call these lesions "infarcts" or "shearing lesions" currently remains open for debate. An appropriate compromise might be "infarctive shearing lesions" or "shearing infarcts," although it remains to be seen whether these lesions are caused by lack of blood supply versus cellular injury.

Although conventional DWI is the most sensitive means to detect classic infarcts in the TBI setting, such as those due to herniation or arterial dissection, shearing lesions may not be visible on DWI in all cases, even though lesions may be observed on other sequences. Alternatively, DWI lesions can sometimes occur without corresponding abnormalities on other sequences. The inconsistency may be partly due to different degrees of cellular injury or the timing of scans. Diffusion-weighted imaging is most useful in the acute phase, and, like classic infarctions, the diffusion properties of lesions will change with time. However, the time course is not necessarily the same as that of classic infarction. One should also remember that what the findings represent at a cellular or molecular level is still not fully understood. Visible hemorrhages are not necessarily associated with DWI lesions, suggesting that some of these lesions may be caused by a cellular response to cell membrane disruption without loss of blood supply.

Diffusion-tensor imaging is also being investigated as a means to detect more subtle injuries that can reflect dysfunction of white matter tracts. Injury to these tracts disrupt the preferential mobility (anisotropy) of water along the fibers. Reduction of anisot-ropy is thought to reflect the histologic effects of DAI--misalignment of axonal membranes, cellular swelling, and axonal disconnection. The detection of abnormalities is usually based on quantitative analysis of parameters, such as fractional anisotropy (FA). Visual depiction is also possible with color tensor maps that display fiber tracts, although abnormalities may be subtle. Recent investigators have shown that there are significant differences in FA in patients with mild TBI and negative CT scans compared with control subjects. ¹² Further work may lead to the ability to predict specific types of neurological impairment based on the observed fiber tract injury pattern.

Magnetic resonance spectroscopy

Magnetic resonance spectroscopy (MRS) is a powerful technique for the identification of chemical compounds, based on certain elements, such as phosphorous or hydrogen (proton MRS). The latter is available on most medium- and high-field clinical MR imagers as a noninvasive diagnostic tool to measure biochemical metabolites in the human brain. MR spectroscopy is an evolving clinical imaging technique whose utility has rapidly expanded in the past decade and now includes: the routine assessment of neoplasms (including pre-treatment planning or follow-up); presurgical assessment of epilepsy; and the evaluation of dementia, white matter diseases, metabolic disorders, and a host of other neurologic diseases.

MR spectroscopy is able to detect subtle changes in brain metabolites that reflect the state of health of tissue, even when conventional MRI appears normal. Various investigators have quantified normal metabolite ratios and levels in control populations for use as reference standards. Because metabolite profiles change with brain maturation, pediatric populations have values different from those of adults. N-acetyl-aspartate (NAA) is the primary metabolite found in neurons and axons. Normal levels imply neuronal integrity. Tissue injury usually results in decreased levels of NAA, which may reflect neuronal death or dysfunction. Total choline (Cho) includes phosphoryl and glycerophosphoryl choline, which are metabolites found in cell membranes. Choline levels markedly increase in the presence of rapid cell turnover (eg, neoplasm), while lesser degrees of elevation are observed with membrane disruption, such as demyelination or axonal shearing. Total creatine (Cre) includes phosphocreatine and its precursor creatine. Creatine was once thought to be a fairly stable metabolite complex that reflected brain energy metabolism. However, it is now recognized that creatine levels can be altered in certain disorders. As a result, metabolite ratios that incorporate creatine levels (eg, NAA/Cre, Cho/Cre) may be less reliable in these situations. Many investigators now advocate the measurement of metabolite levels rather than ratios, although this is less available in routine clinical practice. Myoinositol (Ins) is an important osmolyte in astrocytes and rises in conditions that have gliosis or osmotic disorders. Lactate (lac) is not normally present except in the neonatal period, and accumulates as a result of anaerobic glycolysis or in the presence of infection or necrosis.

There are various techniques to perform MRS, and the reader should consult more detailed descriptions regarding methodology. ^13-15 There are two commonly used methods to perform MRS: stimulated-echo acquisition mode (STEAM) and point-resolved spectroscopy (PRESS). Certain metabolites, such as myoinositol and glutamate/glutamine (Glx), are best detected with short-echo-time acquisitions. There are also different volumes of MRS that can be acquired: single-voxel, multivoxel slab, and volumetric whole-brain acquisitions. Single-voxel spectroscopy (SVS) allows for the acquisition of a single spectrum from one volume of interest (voxel), often in the range of 2 to 8 cm ³ in size. Two-dimensional chemical shift imaging (2D-CSI), also known as 2D-magnetic resonance spectroscopic imaging (2D-MRSI), allows for the acquisition of up to 64 spectra in one measurement that covers a transverse slice through the brain, usually resulting in voxels that are smaller in size than those acquired with SVS. Whole-brain MRS, also known as three-dimensional magnetic resonance spectroscopic imaging (3D-MRSI) or volumetric MRSI, is more complex and is not routinely available, but it may potentially allow for the evaluation of the entire brain. ¹⁶

Traumatic brain injury has been evaluated using both short- and long-echo SVS techniques, albeit at varying time points after injury. Brain injury is commonly associated with decreased NAA/Cre or NAA/Cho in areas of the brain susceptible to TBI, including frontal white matter, ^17-20 parieto-occipital white matter, ^21-23 corpus callosum, ^24,25 or brainstem. ²⁶ Loss of NAA can occur in both gray and white matter, ²⁷ which is believed to reflect both neuronal and axonal injury. The degree of NAA decrease is associated with the severity of injury. ²⁸ Elevated Cho/Cre is often observed ^{14-16,21-23,27} and is thought to represent membrane disruption after axonal injury. Elevated Ins/Cre ratios ^14,15 suggest glial proliferation.

At our institution, we use MRS in the acute setting to detect metabolite abnormalities that lack corresponding injury on conventional imaging. Our group has developed a semi-automated program to quickly process and compare the results of each 2D-MRSI spectrum, acquired at the level of the corpus callosum, to normal age-matched control ratios. Voxels with ratios that fall outside two standard deviations from the mean are flagged and displayed in color maps overlaid onto the corresponding MR image (Figure 2). This allows a quick overview of abnormal areas. We have been using this program to facilitate the interpretation of clinical studies. It should be noted that this differs from commercially available color maps that are not quantitative, in which metabolites are displayed in color ranges based on the lowest and highest values in the sample, without reference to normal values.

MR spectroscopy can also be used to follow patient recovery. However, the greatest utility of MRS may be in predicting long-term outcome following TBI. Decreases in NAA-derived ratios may predict poor neurologic and neuropsychological outcomes ²¹ (Figure 3). Preliminary data from our institution suggests that early elevation of choline and glutamate/glutamine levels are also strong predictors of poor outcome. ²⁹ Further work is being done to correlate specific areas of MRS abnormalities with patterns of neuropsychological findings.

Susceptibility-weighted imaging

Detection of large amounts of intracranial hemorrhage is of primary importance in the acute management of patients with TBI. However, identification of small hemorrhages and their location now provides useful information regarding mechanism of injury and potential clinical outcome. This is particularly helpful for evaluation of DAI, which is often associated with small hemorrhages in the deep brain.

The MR appearance of hemorrhage varies with multiple intrinsic parameters, such as the state of hemoglobin oxygenation and integrity of red blood cells as well as extrinsic parameters such as scanner field strength, receiver bandwith, sequence type, and degree of T1 or T2 weighting. Fortunately, most blood products are paramagnetic (deoxyhemoglobin, methemoglobin, and hemosiderin), making it possible to exploit magnetic susceptibility effects. In acute and early subacute phases, these effects largely occur from deoxyhemoglobin and methemoglobin. As spins encounter heterogeneity in the local magnetic field, they precess at different rates and cause overall signal loss in T2*-weighted (ie, gradient-echo) images. Sensitivity to susceptibility effects of hemorrhage increase as one progresses from fast-spin-echo to routine spin-echo to gradient-echo techniques, from T1 to T2 to T2* weighting, from short to long echo times, and from lower to higher field strengths. ³⁰

Recently, Dr. E. Mark Haacke, Professor of Radiology and Biomedical Engineering at Wayne State University, Detroit, MI, designed a high spatial resolution 3D fast low-angle shot (FLASH) MRI technique that is extremely sensitive to susceptibility changes and is being performed on some conventional scanners. ³¹ The underlying contrast mechanism is associated with the magnetic susceptibility difference between oxygenated and deoxygenated hemoglobin, leading to a phase difference between regions containing deoxygenated blood and surrounding tissues and resulting in signal cancellation. This sequence was originally designed for MR venography, ^31-33 utilizing the paramagnetic property of intravenous deoxyhemoglobin, and was therefore referred to as HRBV (high-resolution blood-oxygen-level­dependent [BOLD] venography).

At our institution, we found the technique was also very sensitive in detecting extravascular blood products, and began using it to evaluate patients with moderate or severe TBI. Because its application can be broader than evaluating venous structures, we have been referring to the technique as susceptibility-weighted imaging. The sequence consists of a strongly susceptibility-weighted, low-bandwidth (78 Hz/pixel) 3D-FLASH sequence (TR/TE = 57/40 ms, FA = 20°) first-order flow compensated in all three orthogonal directions. Thirty-two to 64 partitions of 2 mm are acquired using a rectangular field of view (FOV) (5/8 of 256 mm) and a matrix size of 160 * 512, resulting in a voxel size of 1 * 0.5 *
2 mm ³ . The acquisition time varies from approximately 5 to 10 minutes, depending on the desired coverage. After the data is acquired, additional postprocessing accentuates the signal loss due to the presence of hemorrhage. This primarily involves the creation of a phase mask to enhance the phase differences between paramagnetic substances and surrounding tissue. For further details, the reader is encouraged to refer to the methods described in Reichenbach et al. ³¹

Research at our institution has shown that a significantly greater extent of parenchymal hemorrhage is demonstrated by SWI compared with our conventional gradient-echo imaging sequence. ³⁴ The SWI technique is particularly sensitive to the presence of very small hemorrhages, often demonstrating lesions that are not visible or poorly visualized on other sequences (Figures 4 and 5). The improved lesion visibility was particularly evident in the posterior fossa, for reasons that remain uncertain. This has been very helpful in clinical situations in which the patient remained in an unexplained coma until SWI showed small hemorrhages in the brainstem that were not visible on CT or conventional MRI (Figure 6). The most encouraging preliminary findings were that the extent of hemorrhage was strongly correlated with the initial severity of injury, as measured by the Glasgow Coma Scale (GCS), as well as the subsequent long-term outcome of the patients, as measured by the Glasgow Outcome Score (GOS) at 6 to 12 months after injury. This suggests the SWI technique can play an important role in more accurately diagnosing the degree of injury as well as providing valuable prognostic information that can help determine the aggressiveness of management and rehabilitation. Further work at our institution is under way to refine the correlation of lesion location with specific neuropsychological outcomes.

Conclusion

Although CT remains the first-line imaging procedure in the acute evaluation of head injury, MRI is becoming increasingly important in more detailed analysis of the degree and type of traumatic brain injury, as well as potentially predicting clinical outcome. In particular, the newer MR techniques of diffusion imaging, MRS, and SWI provide unique information that could significantly change the management of TBI. AR