Imaging for the Diagnosis and Management of Traumatic Brain Injury

An epidural hematoma (EDH) typically occurs at the site of traumatic impact or “coup” site and is almost always associated with an overlying skull fracture [15]. Injury to a meningeal artery/vein, diploic vein, or dural venous sinus results in a classically lentiform-shaped collection of blood that strips the dura away from the inner table of the skull. EDHs are most common in the temporal or temporoparietal regions, often due to laceration of the middle meningeal artery or one of its branches (FIG. 1) [16]. A venous EDH can occur in the posterior fossa due to laceration of the transverse/sigmoid sinus, or in the middle cranial fossa due to injury of the sphenoparietal sinus (FIG. 2) [17].

On imaging, EDHs typically appear lentiform or biconvex in shape and do not cross cranial sutures because the periosteal layer of the dura adheres tightly to the suture. However, unlike subdural hematomas, EDHs can cross the midline where the periosteal layer of the dura forms the outer wall of the superior sagittal sinus, which may be displaced from the inner table of the skull (FIG. 3). Also, unlike subdural hematomas, EDHs can extend above and below the tentorium cerebelli, which is a dural reflection separating the cerebellum from the occipital lobes.

The decision to surgically evacuate an acute EDH is based on a number of findings, including the patient’s GCS score, pupil exam, comorbidities, age, and CT findings. On imaging, findings that suggest a poorer prognosis and prompt early surgical evacuation include significant hematoma thickness (>15 mm), large hematoma volume (>30 cm³), significant midline shift (>5 mm), compression of the basilar cisterns, and mixed density of the hematoma, which often indicates active bleeding (Table 1) [16, 18‐20]. The hematoma volume can be calculated by the ABC/2 method described by Kothari et al [21]. where the CT slice demonstrating the largest area of hemorrhage is selected, and A reflects the diameter of the hemorrhage on that slice, B is the measurement taken perpendicular to A, and C is the approximate number of 10-mm slices with hemorrhage.

A subdural hematoma (SDH) can occur either at the coup or contrecoup site, although the latter is more common. Injury to superficial bridging veins results in bleeding between the meningeal layer of the dura and arachnoid, and blood may continue to accumulate in this space as bridging veins are progressively stretched and injured. SDHs commonly occur over the cerebral convexities, along the tentorium cerebelli, and along the falx cerebri, in descending order of frequency [22].

On imaging, SDHs appear crescentic in shape and do not cross the midline; they can, however, cross cranial suture lines (unlike EDHs). Small SDHs can be extremely subtle, so careful adjustment of window settings and scrutiny of the falx cerebri and tentorium cerebelli for subtle asymmetric high density are important (FIG. 4). Most SDHs in the setting of acute TBI are homogeneously hyperintense on noncontrast CT scans. However, isodense hematomas can occur if the patient is severely anemic or CSF from an arachnoid tear is admixed with hemorrhage. Heterogeneous SDHs can occur in the context of active bleeding, coagulopathy, or acute-on-chronic hemorrhage (FIG. 5).

Surgical management of SDHs, like EDHs, depends on a number of factors, including the patient’s GCS score, pupil exam, comorbidities, age, and CT findings. Neurologic deterioration over time is another important factor in surgical decision-making [23]. As with EDHs, CT findings that appear to predict a poor prognosis include clot thickness, midline shift, and basilar cistern compression. While specific threshold values are difficult to determine, and vary among different studies, the Surgical Management of Traumatic Brain Injury Guidelines issued by a consortium of neurosurgeons following a review of the available scientific literature recommends surgical evacuation of acute SDHs that are >10 mm thick or those causing a midline shift >5 mm visible on the CT scan (Table 2) [23].

Subarachnoid hemorrhage (SAH) occurs in ~40% of patients with moderate to severe head injury and is commonly associated with other types of intracranial hemorrhage [24]. SAH may result from direct laceration of the small cortical vessels traversing the subarachnoid space, redistribution of intraventricular hemorrhage exiting the fourth ventricular outflow foramen, or direct extension from cortical contusion/hematoma. Patients with traumatic SAH have a significantly worse outcome than those without SAH: 41% of patients without traumatic SAH achieved a level of good recovery compared with only 15% of patients with SAH, according to the results of a large study conducted by the European Brain Injury Consortium [25].

Acute SAH is readily identified on noncontrast CT scans as linear areas of high attenuation in the cerebral sulci at the convexities, Sylvian fissures, or basilar cisterns. It is important to examine areas such as the interpeduncular cistern, which may contain subtle SAH that is easily missed without conscious attention to this area. In ~5% of patients with TBI, SAH may be the only abnormal finding on a noncontrast CT scan [25].

Hydrocephalus is a common complication of traumatic SAH and may develop acutely or in a delayed fashion. Intraventricular hemorrhage or inflammatory arachnoiditis may result in hydrocephalus in the acute setting, while decreased resorption of CSF by arachnoid villi is responsible for chronic hydrocephalus.

Cortical contusions in closed head injury result when the brain is bruised by the irregular inner surfaces of the skull at the time of impact. They can occur at the coup site when a depressed skull fracture or transient calvarial deformity from a blow to the stationary head grazes the underlying cortex. More commonly, however, they occur at the contrecoup location when the moving head collides against a stationary object. These contrecoup contusions are frequently of greater severity than the injuries at the coup site. The precise mechanism of contrecoup contusions and the explanation for their greater severity are subjects of considerable debate. A recent explanation proposes that when a moving skull collides against a stationary object, such as when a person falls and strikes the back of his head against the ground, the buoyant brain is displaced by CSF in the opposite direction from impact, causing injury opposite the blow [26].

Contrecoup injuries are especially severe when the blow occurs to the occiput, allowing the rough, irregular surfaces of the anterior and central skull base to bruise the undersurface of the brain, resulting in the characteristic hemorrhagic contusions seen in the inferior frontal and temporal lobes (FIG. 6). Contusions appear as small, focal areas of petechial hemorrhage peripherally located in the brain, but they may be very subtle on initial noncontrast CT scans with minimal hemorrhage and/or edema. However, over time, approximately half of contusions evolve and grow larger in size (FIG. 7) [27]. For this reason, serial CT imaging and close clinical monitoring of TBI patients is paramount. Gradient-echo (GRE) sequences on MRI, which are particularly sensitive to inhomogeneities in the magnetic field caused by paramagnetic hemorrhage, are far more sensitive than CT in detecting small hemorrhagic contusions.

With greater severity of head trauma, microhemorrhages associated with a contusion may coalesce into an intracerebral hematoma (ICH), reflecting that contusions and ICH represent a spectrum of injury rather than discrete, independent entities. ICHs may also develop in delayed fashion in a part of the brain that was previously seen to be radiographically normal, an entity known as the “delayed traumatic intracerebral hematoma.” Similar to contusions, ICHs represent a dynamic process of injury. According to one study, only 56% of ICH >3 cm in diameter developed within 6 h of injury, and only 84% of ICH reached maximal size by 12 h, [28] underscoring again the importance of intensive clinical monitoring and serial imaging.

The specific indications for craniotomy and surgical evacuation in patients with contusion or ICH have not been established. Generally, a combination of clinical and radiologic factors is felt to be important, including hemorrhage location and volume, extent of mass effect on CT (cisternal effacement or midline shift), GCS score, ICP, and neurological deterioration [29‐32]. Patients with contusions/ICH and progressive neurological decline, medically refractory increased ICP, or radiographic evidence of mass effect tend to have poor outcome without surgical treatment, although specific surgical criteria have not been established.

Traumatic axonal injury (TAI) is also commonly referred to as diffuse axonal injury or shear injury. The term TAI is preferred by the authors of this review as injuries are not distributed diffusely throughout the whole brain, but occur in characteristic, discrete locations, including the parasagittal white matter near the cerebral cortex, corpus callosum, and brainstem. Mild (grade I) TAI involves the gray–white junction of the lobar white matter, particularly the parasagittal frontal lobes; moderate (grade II) TAI involves the fibers of the corpus callosum, particularly the splenium, in addition to the subcortical white matter; severe (grade III) TAI involves the dorsolateral midbrain, in addition to the subcortical white matter and corpus callosum.

The mechanism of injury is one of cytoplasmic shear-strain of the axonal cytoskeleton due to sustained acceleration/deceleration, such as that which occurs with a high-speed motor vehicle crash or prolonged shaking [33]. Damage to the neurons occurs not only at the time of mechanical injury but in the hours, days, and weeks, even years, following the traumatic event due to a deleterious cascade of biochemical events and Wallerian-type degeneration with progressive neuronal loss. Changes associated with TAI are thought to be responsible for the majority of global cognitive defects seen after TBI, particularly with regard to difficulties with memory and information processing.

Unfortunately, TAI is notoriously difficult to diagnose by CT. Only 10% of TAI patients demonstrate the classic CT findings of petechial hemorrhages at the gray–white junction of the cerebral hemispheres, corpus callosum, or brainstem. Indeed, the results of most admission CT exams are normal because >80% of TAIs are nonhemorrhagic [33, 34]. MRI is far more sensitive than CT as it can detect both hemorrhagic and nonhemorrhagic lesions (FIG. 8). Acute hemorrhagic lesions demonstrate focal susceptibility and signal loss due to the paramagnetic effects of deoxyhemoglobin on gradient-recalled echo (GRE) and susceptibility-weighted images (SWI). These susceptibility effects are further magnified at 3 Tesla (3T) compared to 1.5T, with nearly twice as many hemorrhagic lesions detectable at 3T [35]. Nonhemorrhagic lesions are best appreciated on fluid attenuation inversion recovery imaging, which makes subcortical and periventricular lesions particularly conspicuous (as compared with conventional T2-weighted imaging). Diffusion-weighted imaging (DWI) is also sensitive for TAI and can detect lesions that are not seen on either fluid attenuation inversion recovery or GRE sequences (FIG. 9) [36]. TAI lesions typically appear hyperintense with reduced apparent diffusion coefficient values in the acute setting, likely reflecting cytotoxic edema [37]. Unfortunately, there are no good treatment options for TAI patients at this time, and the value of improving the accuracy of imaging diagnosis lies chiefly in the ability to better prognosticate clinical outcome, as will be discussed in further detail later in this review.

There is an ongoing debate in the literature as to whether cerebral edema or cerebral hyperemia with increased blood volume is the underlying cause of cerebral swelling. Both mechanisms may be at play. A generally held premise is that cerebral hyperemia due to dysautoregulation with vascular engorgement and increased cerebral blood volume is the principal mechanism of cerebral swelling, although recent data suggest that blood volume may actually be decreased following trauma and that cerebral edema may be the major fluid component of brain swelling [38, 39]. There are two types of cerebral edema: vasogenic, due to disruption of the blood–brain barrier allowing accumulation of extracellular water, and cytotoxic, due to the failure of cell membrane pumps, resulting in intracellular water leakage. Of these two types of edema, cytotoxic edema may be primarily responsible for cerebral swelling, as demonstrated by the increased water content and reduced apparent diffusion coefficient values in one MRI study [40].

On imaging, cerebral swelling due to hyperemia is represented as the loss of sulci (sulcal effacement), compression of the basilar cisterns, and flattening of the ventricular margins (FIG. 10). Gray and white matter attenuation and differentiation will remain intact. However, cerebral swelling due to vasogenic edema will be revealed as areas of low attenuation, while swelling due to cytotoxic edema will appear as a loss of gray–white differentiation.

Children are particularly susceptible to diffuse cerebral swelling following TBI, with the incidence of diffuse swelling being approximately twofold higher in children than in adults [41]. Children and young adults are more prone to post-traumatic dysautoregulation, which leads to vasodilatation, hyperemia, and cerebral swelling. When swelling is severe, ICP increases and CPP falls, resulting in infarction and cerebral damage.

Medical strategies to combat the deleterious effects of cerebral swelling and lower ICP include head elevation, moderate hyperventilation, hyperosmolar therapy with mannitol, and the appropriate use of sedatives and analgesics (barbiturates) to prevent pain or agitation from exacerbating elevated ICP. However, when ICP is refractory to maximal medical therapy, decompressive craniectomy has been used in an attempt to lower ICP. Studies on the efficacy of decompressive craniectomy generally suggest that it can lower ICP and may improve functional outcome [42‐44]. Decompressive craniectomy also results in radiologic improvement, with better visualization of basilar cisterns and decreased midline shift, both of which correlate with improved outcome [45]. However, the specific clinical criteria and context for decompressive surgery have yet to be established.

Cerebral herniation is a result of unmitigated increased ICP. These patients typically undergo decompressive craniectomy with mass lesion evacuation.

Subfalcine herniation or midline shift occurs when the cingulate gyrus herniates under the falx cerebri; uncal herniation results when the medial temporal lobe herniates through the tentorial incisura and compresses the ipsilateral suprasellar cistern; descending transtentorial herniation occurs with downward herniation of both temporal lobes through the tentorial incisura, compressing the basilar cisterns; upward transtentorial herniation occurs in the opposite direction, with the cerebellum extending through the tentorial incisura and effacing the quadrigeminal cistern; tonsillar herniation results when the cerebellar tonsils herniate into the foramen magnum.

In the absence of CT imaging, any patient with a blown pupil is clinically presumed to have cerebral herniation. With uncal herniation, the displaced medial temporal lobe compresses the ipsilateral oculomotor nerve, resulting in loss of the normal pupillary reflex. Other important sequelae of cerebral herniation include cerebral infarction, which will be discussed in the following section, and the Duret hemorrhage, which typically occurs in the ventral and paramedian midbrain/pons following rapid downward herniation (FIG. 11). It is thought that either laceration of the pontine perforator branches from the basilar artery or venous thrombosis with infarct is the underlying mechanism of injury in the Duret hemorrhage.

Cerebral ischemia and infarction occurs in ~2% of patients undergoing head CT for TBI and can occur from a variety of causes [46]. Most commonly, ischemia occurs due to mechanical compression of a blood vessel from cerebral herniation across the falx cerebri and/or tentorium. Infarctions due to mechanical shift are typically seen in the anterior cerebral artery or posterior cerebral artery distribution following acute subfalcine or uncal herniation, respectively (FIG. 12). The callosomarginal branch of the anterior cerebral artery can be compressed against the free edge of the falx and cause infarction with subfalcine herniation. In uncal herniation, the displaced medial temporal lobe can compress the PCA, resulting in infarction, or constrict the anterior choroidal artery, leading to infarction of the posterior limb of the internal capsule.

Other potential causes of ischemia and infarction in patients with TBI include vasospasm, which has been angiographically documented in patients following head trauma.47 Also, extra-axial hematomas that exert a significant mass effect on the adjacent cortex may compress cortical veins and result in venous infarction. Finally, direct vascular injury, such as dissection, occlusion, or pseudoaneurysm from a skull base fracture, may also result in ischemia.

Individual predictors of poor outcome on conventional CT include compressed or absent basilar cisterns, the presence of subarachnoid hemorrhage, midline shift, and intracranial hemorrhagic lesions (e.g., subdural or epidural hematomas, shear injury, and contusions). Compressed or absent basilar cisterns indicate a threefold higher risk of increased ICP and are associated with a two- to threefold increase in mortality [48]. Traumatic subarachnoid hemorrhage is associated with a twofold increase in mortality, and hemorrhage in the basilar cisterns has a 70% positive predictive value for poor outcome [48]. Midline shift indicates increased ICP and is similarly associated with a poor clinical outcome, although this association is somewhat complicated by the fact that midline shift is caused by intracranial hemorrhage that also negatively impacts outcome. Finally, intracranial hemorrhage has an approximately 80% positive predictive value for poor functional outcome, with prognosis worsening as the hematoma volume increases in size [48]. Of note, mortality is higher in patients with acute subdural hematomas compared with those with epidural hematomas.

Marshall et al., [49] in his publication on the Trauma Coma Data Bank, aggregated key findings on CT and proposed an imaging-based classification system of head injury that predicts mortality (Table 3). This system defines 4 types of diffuse head injury and takes the presence of a mass lesion and evidence of increased ICP (status of the basilar cisterns, presence of midline shift) into account. The Marshall Classification is widely recognized and used, but a more recent study published by Maas et al. [50]. from Rotterdam further refined imaging parameters to develop a CT score that better predicts clinical outcome. The Rotterdam Score evaluates the status of the basilar cisterns, presence of midline shift, presence of epidural hematoma, and presence of intraventricular or subarachnoid hemorrhage to derive a score that relates to 6-month mortality (Table 4).

While noncontrast CT is invaluable for the diagnosis, management and prognosis of patients with TBI, it is limited to displaying only anatomic information. Physiologic derangements in cerebral perfusion, blood flow, and oxygenation related to the deleterious cascade of secondary traumatic brain injuries that significantly impacts functional outcome are not captured on conventional noncontrast CT scans. Perfusion CT is a new imaging tool that utilizes dynamic scanning during intravenous contrast injection to illustrate physiologic parameters of cerebral blood volume (CBV), cerebral blood flow (CBF), and mean transit time (time for blood to perfuse a region of tissue).

Evidence of normal brain perfusion or hyperemia (high CBV and CBF) on perfusion CT scans in TBI patients has been associated with a favorable clinical outcome, while oligemia (low CBV and CBF) has been associated with unfavorable outcome [51]. These findings on perfusion CT scans have been postulated to reflect the status of cerebrovascular autoregulation: patients with relatively minor head trauma and intact autoregulation maintain or even slightly increase cerebral perfusion, while patients with severe head injury have impaired autoregulation and pressure-passive flow, which often results in oligemia and decreased perfusion. The excitement of using perfusion CT to better understand the physiology of TBI patients is tempered by concerns of increased radiation exposure, particularly in young patients who are often the victims of head injury.

While CT is useful for the diagnosis and prognostication of outcome in patients with hemorrhagic injuries such as subdural and epidural hematomas, patients with subarachnoid hemorrhage and contusions, a subset of TBI, have normal CT scans without hemorrhage yet still fare poorly. TAI patients can suffer significant post-traumatic symptoms and may be profoundly comatose, with cognitive impairment and poor functional outcome despite normal results on conventional CT or MR images. As such, there is considerable interest in developing more sensitive diagnostic tests for TAI and for correlating these tests with long-term functional outcome. Unfortunately, as there are no good treatment options for TAI patients at this time, the value of improved imaging diagnosis currently lies in the potential to better prognosticate clinical outcome.

Diffusion tensor imaging (DTI) is one investigative MRI technique that maps out the microstructural characteristics of the brain based on the intrinsic diffusion properties of neurons. Unlike conventional diffusion weighted imaging, which encodes diffusion in 3 basic directions, DTI assesses diffusion in at least 6 but typically 25–30 directions, yielding a more complete set of diffusivity information that can be used to deduce axonal orientation and create maps of white matter tracts in the brain. TAI interrupts the cytoskeletal network and impairs axoplasmic transport, causing changes in tissue diffusivity that can be visualized and quantified on DTI [52]. DTI has been shown to be more sensitive than conventional 3T MRI in detecting TAI, and results from recent studies suggest that the extent of damage to white matter structures on DTI may correlate with the extent of cognitive impairment and functional outcome following TBI [53, 54].

SWI is another relatively new MRI technique that increases sensitivity for the detection of TAI. SWI utilizes a high-resolution, velocity-compensated, 3-dimensional GRE sequence based on both magnitude and phase data [55, 56]. It is exquisitely sensitive to susceptibility effects due to microhemorrhages, small veins, and changes in iron content, and more sensitive than conventional GRE sequences for the detection of hemorrhagic TAI, revealing four- to sixfold more microhemorrhages than conventional GRE imaging [57, 58]. Initial studies also suggest that lesion burden on SWI correlates with clinical outcome, including the duration of coma and long-term functional disability [59, 60].

Magnetoencephalography is a functional imaging technique that measures the magnetic field generated by neuronal activation. Unlike normal brain tissue that generates α-waves with a frequency of 8–13 Hz, injured neurons produce δ-waves with a low frequency of 1–4 Hz. This magnetic signal can be measured and localized with high spatial (2–3 mm) and temporal (<1 ms) resolution. In TBI patients, magnetoencephalography “slow waves” can be seen in cortical gray matter that is functionally and structurally connected to white matter fiber tracts with axonal shear injury, as demonstrated on DTI, and may be more sensitive than DTI in diagnosing mild TBI [61].

Functional MRI (fMRI) is another functional imaging technique that relies on the relationship between physiologic function, energy consumption, and blood flow to depict brain activity. Most brain fMRI studies rely on a blood oxygen level-dependent signal to demonstrate cerebral function. It is thought that neuronal activity causes an increase in cerebral flood flow and local blood oxygenation with a relative decrease in the amount of deoxyhemoglobin, a product of oxygen consumption. Deoxyhemoglobin is paramagnetic, generating field inhomogeneities that result in signal loss on GRE sequences sensitive to susceptibility effects. Thus, as the relative concentration of deoxyhemoglobin decreases within a region of neuronal activity, an increased signal within this area is seen on GRE images. Blood oxygen level-dependent fMRI studies performed in patients with mild TBI and structurally normal imaging studies have shown differences in brain activation during various tasks as compared with healthy control subjects [62, 63].