Historical note and nomenclature

A brief history of stroke neuroimaging

Brain parenchymal imaging. Advances in imaging of the brain parenchyma have contributed immensely to the diagnosis of ischemic stroke (26). In 1971, CT was invented by engineer Godfrey N Hounsfield and revolutionized brain parenchymal imaging. This improvement was dramatic, considering that prior to the advent of CT imaging, the primary ways to image brain parenchyma were with pneumoencephalography and catheter angiography. Magnetic resonance imaging was developed shortly after CT with Peter Lauterbur, who shared the Nobel Prize in 2003 with Peter Mansfield, by demonstrating the first images produced by magnetic resonance in 1973 (25; 26). MRI now represents a means of obtaining far superior resolution of parenchymal anatomy than CT without the effects of ionizing radiation. Of significant importance to acute stroke imaging was the introduction of diffusion-weighted MRI to identify areas of acute ischemia in 1990 (35), as well as the development and refinement of perfusion imaging with CT and MR.

Brain vascular imaging. Advances in vascular imaging began with catheter angiography of the carotid arteries first performed in 1927 by Egas Moniz. Moniz and others considerably developed cerebral angiography in the 1930s. In the 1960s, the transfemoral approach essentially replaced direct puncture of the carotid and brachial arteries. The development and improvement of MR angiography in the 1980s and 90s and the advent of helical CT angiography in 1991 represent major advances in vascular imaging, which now make rapid and noninvasive contrast-enhanced angiography possible. Despite these advances in MR and CT angiography, catheter angiography remains the reference standard for vascular imaging today because of its superior temporal and spatial resolution. Because of its favorable complication profile, the radial approach now represents a significant contribution to neurointerventional field (13; 46).

Brain imaging in the endovascular reperfusion era. 2014 and 2015 saw the publication of multiple positive trials for endovascular mechanical thrombectomy in treating acute ischemic stroke due to large vessel occlusion. By proving the effectiveness of this powerful therapy, these trials revolutionized the field of acute ischemic stroke, as did the NINDS trial of IV-tPA before them. Furthermore, these trials underscore the importance of early vascular imaging with CTA or MRA to identify large vessel occlusion in patients suspected to have acute ischemic stroke. Advanced perfusion imaging in high-volume centers with appropriate interpretation can complement these forms of lumenography for multiple reasons. First, perfusion imaging can establish candidacy for endovascular therapy in the late time window or for wakeup strokes, as shown in both the DAWN and DEFUSE 3 trials (03; 37). Second, perfusion imaging proves very sensitive in identifying more distal occlusions amenable to effective and safe endovascular reperfusion (31; 39). These distal occlusions are more conspicuous on perfusion imaging than on vascular imaging with MRA or CTA alone (08; 07).

The future of large vessel occlusion imaging. Thrombectomy for acute large vessel stroke is one of the most effective therapies in all of medicine and is, of course, critically dependent on time. In the endovascular reperfusion era, it has become imperative to rapidly identify patients with acute stroke and large vessel occlusion. Multiple innovative strategies for expediting identification and treatment of large vessel occlusion are now available and under investigation. First, early identification of large vessel occlusion now represents a promising potential application of artificial intelligence. Indeed, software now exists that uses artificial intelligence to identify large vessel occlusion from CTA, allowing for near real-time alerts for an advanced stroke team. This technology can dramatically impact the field of acute stroke, especially for transferred patients (34). Second, in patients with clinical examination signs of large vessel stroke, significant time savings have been demonstrated in taking these patients directly to the angiography suite for evaluation and possible thrombectomy (42). Third and finally, vascular imaging in the prehospital setting may also serve to hasten the identification of emergent large vessel occlusion in patients able to be evaluated with vascular or perfusion imaging in a mobile stroke unit (15). As described above, identifying distal occlusions is more important than ever as our field advances because this is a new frontier for endovascular therapy (see clinical vignette) and for intravenous thrombolysis. In evaluating patients with acute stroke, identifying distal occlusions is consistently more difficult than identifying proximal occlusions because the clinical and imaging findings are more subtle in distal occlusions. Therefore, the application of future large vessel occlusion detection must focus not only on speed but also on sensitivity of detecting these distal occlusions.

Scientific basis

Physiological basis of acute stroke imaging

Acute ischemic stroke affects approximately 700,000 persons annually and remains the leading cause of adult long-term disability. The first discernible sign of ischemia on MRI is restricted diffusion of water on diffusion-weighted imaging caused by cytotoxic edema (35). In acute ischemia, the increased water in the intracellular compartment is restricted in its diffusion because the intracellular environment is packed with nuclei and organelles unlike the extracellular environment. Diffusion-weighted MRI is highly sensitive to ischemia and easily interpreted.

Noncontrasted CT, on the other hand, is less sensitive for diagnosing acute ischemia because cytotoxic edema does not produce profound decreases in brain density. However, cytotoxic edema probably does underlie the subtle early changes seen on CT, including decreased density in gray matter best seen in gray and white junctions, such as cortical ribbons, thalami, and basal ganglia (47). The density of normal gray and white matter is very similar, at 35 Hounsfield units (HU) and 25 HU, respectively. Therefore, to better visualize subtle differences in gray and white matter junctions, adjusting the center level to 30 HU and window width to 30 HU on CT reading software can be helpful. This is sometimes referred to as using narrow windows or stroke windows. A center level of around 30 HU will place the center of the gray scale between gray and white matter, and a window width of around 30 HU will make the gray scale range from 15 HU (black) to 45 HU (white), thereby producing a grayscale window that is well-tuned for differences in the gray/white junction.

Vasogenic edema occurs after more prolonged ischemia and results in increased water in the brain from the intravascular space. On MRI, this is characterized by hyperintense signal on T2-weighted imaging, such as FLAIR and T2 hyperintense shine-through on ADC maps. Because water (0 HU) is less dense than brain parenchyma (25 to 30 HU), vasogenic edema is seen on CT as more conspicuous areas of hypodensity. Evidence of vasogenic edema on FLAIR imaging suggests some chronicity and can be used to estimate time of ischemia onset. An MRI diffusion FLAIR mismatch can now establish thrombolysis candidates with unknown stroke onset time (49).

Brain perfusion imaging, which can be performed reliably on MRI or CT in acute stroke, is based on the equation stating that cerebral blood flow is equal to cerebral blood volume minus mean transit time (CBF = CBV – MTT). Each of these parameters change in predictable ways in response to ischemia and infarction. Years of empiric observation of these parameters on perfusion imaging have led to iterative refinement of the thresholds for ischemic penumbra and infarct core. Brain with a Tmax of longer than 6 seconds has now been widely adopted as a brain at risk in the setting of acute stroke. Work shows that several thresholds for relative CBV (0.30 to 0.34) and CBF (0.32 to 0.34) provide an accurate estimate of core infarct (33). Perfusion imaging to distinguish core infarct from salvageable brain is now a proven approach to establishing candidacy of both thrombectomy and thrombolysis in patients presenting in a late time window and wake-up stroke patients (03; 37; 27).

Indications

Which patients should be imaged?

The major indication for acute stroke neuroimaging workup is acute onset neurologic deficit as reported retrospectively or as seen on examination by an experienced clinician. Although some have used an NIH stroke scale cutoff, eg, NIHSS greater than 6 as a threshold for vascular imaging, this cutoff will still miss 13% of large vessel occlusions (48). As described below, the risks of intravenous contrast are much lower than previously thought and can be mitigated with modern approaches. Finally, as stated above in our key points, consistency of the imaging strategy within a given institution is critical to facilitate reliable technical execution, rapid evaluation, and accurate interpretation of results. Therefore, at our institutions acute stroke imaging, including head and neck vessel angiography and perfusion imaging, is standardly performed for all acute stroke patient evaluations.

Although acute imaging to identify candidates for reperfusion therapy is optimally performed within 6 hours from the last known normal time, many patients can still greatly benefit from endovascular reperfusion when they present in a delayed manner, as has been shown in both the DAWN and DEFUSE 3 trials. More recently, the EXTEND and WAKE UP trials also showed benefit from thrombolysis in many patients who present beyond the conventional thrombolysis window (49; 27). Therefore, it is imperative that even patients presenting in a delayed manner still be evaluated with a rapid and complete acute stroke imaging protocol.

Contraindications

Contraindications, caution, and precautions when performing acute stroke imaging

Impaired renal function has historically been a reason for caution in administering contrast agents given the risks of nephrogenic systemic fibrosis with gadolinium MR contrast and contrast-induced nephropathy with iodinated CT contrast. However, it should be noted that the risk of renal injury with iodinated contrast is far less than previously thought (29). Evidence suggests the risk of contrast-induced nephropathy with present-day low-osmolality intravenous iodinated contrast is only significant if eGFR is less than 30. Patients with GFR less than 30 and with signs and symptoms of large vessel occlusion on an expert’s clinical evaluation should still be seriously considered for vascular imaging with CTA, accepting some risk of contrast-induced nephropathy in exchange for identifying large vessel occlusion. This recommendation matches current guidelines (40). Alternatively, at centers where MRI can be conducted without delaying evaluation of the acute stroke patient, high-resolution time of flight MRA can be used, as it does not require intravenous gadolinium contrast.

A second reason for caution with intravenous iodinated CT contrast is documentation of a contrast allergy. However, this should not represent an absolute contraindication to iodinated contrast in evaluating the acute ischemic stroke patient. Rather, we recommend establishing a pretreatment protocol per institutional guidelines in patients with documented iodine allergy who may have a large vessel occlusion causing acute stroke. For example, it is reasonable to use diphenhydramine and steroids prior to CTA or CTP in patients with documented allergies rather than forgoing the imaging evaluation. Again, MRA without contrast can be performed at institutions where it does not delay evaluation.

Goals of acute stroke imaging

The primary goals of stroke neuroimaging are to discover factors that directly change the management of the acute stroke patient.

Candidates for thrombolytic therapy. The most widespread use of acute imaging when making the intravenous thrombolytic decision in acute stroke is whether or not hemorrhage is present, ie, any intracranial hemorrhage is a contraindication for thrombolytic therapy. As such, the only imaging necessary to complement expert clinical evaluation for thrombolytic administration in the conventional time window is a negative head CT. Other imaging findings need not delay administration of thrombolytics in patients who are deemed good candidates on experienced clinical evaluation. However, imaging evaluation of the acute stroke patient should not stop at ruling out hemorrhage. The EXTEND and WAKE UP trials have shown benefit of thrombolysis in many patients who present beyond the conventional thrombolysis window (49; 27), which is yet another reason that we strongly recommend either a CT perfusion or MRI be in the standard algorithm for code stroke imaging at any comprehensive stroke center. Even in the early time window within 4.5 hours from last known well time, there are other findings on acute stroke imaging with MRI or perfusion imaging that can aid the thrombolysis decision in challenging clinical scenarios. These findings are useful correlates to the acute clinical evaluation of a stroke patient when evaluated in real time in conjunction with the clinical picture of a potential stroke patient.

Endovascular candidates: vascular imaging. Approximately 20% of acute ischemic stroke patients will present with a large vessel occlusion. Among these patients, it is critical to diagnose persistent flow-limiting vascular lesion. In those presenting beyond 6 hours since last known well time, endovascular intervention is still beneficial for selected patients. Several methods for selecting endovascular candidates in the delayed time window can be used. These include perfusion imaging (12; 03; 37), the status of collateral vessels (17), or simply an ASPECTS score on CT (09).

Distal occlusions amenable to reperfusion therapy. Of six positive thrombectomy clinical trials for M1 or internal carotid artery terminus occlusions in early-presenting patients, four used advanced perfusion imaging to establish candidacy for thrombectomy. However, two did not employ perfusion studies and still demonstrated benefit of endovascular thrombectomy. Nevertheless, there are three principle reasons that the authors of this review strongly recommend advanced perfusion imaging of the acute stroke patient even in the early time window. First, perfusion imaging is more sensitive than vascular imaging in isolation for the identification of intermediate or distal occlusions, which are subtle on CTA and MRA but very conspicuous on perfusion imaging. Patients who undergo endovascular revascularization of these distal lesions can benefit significantly (08; 31; 39). Second, with the publication of the EXTEND trial, perfusion imaging is an important tool in identifying late-presenting patients who can benefit from thrombolysis. Third, reliable technical execution and interpretation of perfusion imaging requires regular performance of such imaging rather than its use only in “special cases.”

Extensive baseline infarction. Identification of large, completed infarcts is important for multiple reasons. First, large infarcts predict higher risk of reperfusion-associated hemorrhage in a patient who is a candidate for intravenous thrombolysis and endovascular therapy. Although large infarcts benefit less from endovascular therapy, evidence suggests a benefit of endovascular therapy in patients with moderate to large completed infarcts prior to intervention (54; 20; 43). Finally, large infarct volume predicts risk of malignant middle cerebral artery syndrome and should prompt early consideration of decompressive hemicraniectomy. Data on infarct size are, thus, helpful for making medical decisions, counseling patients and families regarding likely clinical outcomes, and involving neurosurgical colleagues early when appropriate.

Stroke etiology. An essential function of neuroimaging beyond guiding acute interventions is to provide data that would otherwise change our clinical management of the patient. Specifically, stroke imaging combined with appropriate clinical acumen informs the workup of stroke etiology and, therefore, secondary prevention.

Stroke imaging algorithms

A 2013 joint statement from the American Society of Neuroradiology, the American College of Radiology, and the Society of Neurointerventional Surgery recommended an algorithmic approach for the imaging workup for patients with acute stroke or transient ischemic attack symptoms that is still applicable today (53). If the patient is a candidate for intravenous thrombolysis, the recommendation is to proceed to noncontrast CT. MR diffusion and GRE sequences may be obtained instead of noncontrast CT at centers where this does not cause a delay in treatment. If the patient is being considered for endovascular reperfusion therapy, any of three strategies are deemed equivalent. The first strategy is to perform noncontrast CT followed immediately by digital subtraction angiography for vascular assessment. The second strategy is to perform noncontrast CT and CTA for vascular assessment with or without CT perfusion imaging. The third strategy is to perform MRI and MRA with or without MR perfusion imaging at centers that offer MRI 24 hours per day without delaying treatment. For the reasons stated previously, these authors strongly recommend the routine use of perfusion imaging to complement CTA or MRA.

In acute stroke patients with anterior circulation stroke or transient ischemic attack who are not eligible for acute reperfusion therapy, the carotid arteries should be evaluated after the brain parenchyma has been imaged. This can be done with carotid Doppler ultrasonography, CTA, or MRA of the neck, which are considered equivalent and sometimes complementary strategies. Discrepancy across these imaging modalities can be settled by catheter angiography, the reference standard for vascular imaging of the carotid arteries.

Accomplishing the goals of stroke imaging

Establishing absence or presence of intracranial hemorrhage: thrombolysis candidacy. Intracranial hemorrhage is considered an absolute contraindication for intravenous thrombolytics. On CT, intracranial hemorrhage (60 to 100 HU) is significantly contrasted against a background of brain with gray and white matter densities (25 to 35 HU) and, therefore, does not require aggressive windowing to identify. Detecting CT signs of early ischemia may require narrow stroke windows or use of CT angiography source images in patients with unclear presentation, but absence of these signs should not delay administration of intravenous thrombolytics when ischemic stroke is otherwise suspected. On MRI, the classic signature of hyperacute intraparenchymal hemorrhage is seen on GRE as a rim of hypointense deoxygenated blood surrounding an oxygenated isointense core. The GRE sequence is at least as sensitive as noncontrast CT for the purpose of identifying intraparenchymal hemorrhage (23). FLAIR imaging is at least as sensitive for identifying subarachnoid hemorrhage as CT because CSF mixed with blood in the subarachnoid space appears hyperintense on FLAIR (36).

If hemorrhage has been ruled out with a noncontrast CT or GRE, intravenous thrombolysis does not need to be delayed by additional imaging findings in patients presenting in the conventional time window for thrombolysis. However, some ancillary findings may complement clinical findings when making this decision in ambiguous or challenging clinical situations. A wedge-shaped perfusion defect in the correct area of clinical impairment on perfusion imaging or restricted diffusion on MRI can help to rule in ischemic stroke and prompt more definitive use of thrombolytics. Either CT perfusion or MR imaging is necessary to establish candidacy for delayed thrombolysis (49; 27).

Endovascular thrombectomy candidacy. Randomized clinical trials from 2014 and 2015 have established AHA Level 1 Class A evidence demonstrating the clinical efficacy of endovascular treatment performed with second generation thrombectomy devices (09; 12; 17; 21; 44; 32), all of which show overwhelming clinical benefit from modern endovascular treatment in patients with large vessel occlusion. The primary function of neuroimaging in endovascular therapy is to select appropriate endovascular candidates with persistent proximal flow-limiting vascular occlusive lesions. These studies support the notion that time to revascularization remains one of the most critical factors in any effort to achieve a good clinical outcome. It is critical to perform a parallel evaluation for endovascular therapy while thrombolytic candidacy is being determined. CTA, a widely available modality, is a quick and reliable way of identifying large vessel occlusion in patients presenting with acute ischemic stroke. The risk of contrast-induced nephrotoxicity should not delay performance of a CTA to determine presence of large vessel occlusion if suspected clinically. Perfusion imaging also increases overall sensitivity for intermediate and distal vessel occlusions that are candidates for intravenous or endovascular reperfusion therapy (08; 31).

A secondary role of acute neuroimaging in evaluating endovascular candidates is to characterize penumbral tissue and collateral vessels. There is a particular benefit of reperfusion therapy in patients who have a significant volume of brain at risk on perfusion imaging prior to intervention (24). The goal of CTP/MR DWI-PWI-based endovascular patient selection is to detect infarct core volume and select patients with a mismatch of this core to salvageable “at-risk” tissue. Because mismatch ratios are highly variable and depend on patient-specific physiology, perfusion imaging allows for selection of patients with delayed presentations and so-called “wake-up” strokes (22; 03; 37).

When MR or CT perfusion imaging is used, Tmax of greater than 6 seconds is a reasonable threshold for brain at risk (53). Some published clinical trials have used absolute minimum mismatch ratios and core infarct volume cut-offs as part of their inclusion criteria (12; 44). Optimal use of these criteria requires availability of real-time postprocessing of perfusion imaging data at the time of acquisition (02). Without a standardized advanced workflow with perfusion imaging, a “clinical penumbra” may be the best surrogate marker to select patients for mechanical thrombectomy. This requires a clinical concern of large vessel occlusion by an experienced provider and an estimate of small core based on available imaging.

However, we advise extreme caution against using perfusion imaging alone to exclude patients presenting early (inside 6 hours) from thrombectomy because a “ghost core” may be present that overestimates core infarct, especially in young patients presenting early (28).

Similar to perfusion imaging, collateral vessel imaging can aid in identifying patients for whom endovascular therapy will not likely benefit (01; 17; 38) and improve sensitivity of detecting distal occlusions (30).

Extensive completed infarction. Large, completed infarcts may affect management decisions regarding intravenous thrombolysis treatment, endovascular reperfusion, and consideration of early decompressive hemicraniectomy for malignant middle cerebral artery syndrome. Regarding the decision to give intravenous thrombolysis, a posthoc analysis in the seminal ECASS trial demonstrated that patients with completed infarcts of greater than one third of the middle cerebral artery distribution are at higher risk of hemorrhage when given intravenous thrombolysis (18). Therefore, a completed infarct of greater than one third of the middle cerebral artery is considered a relative but not absolute contraindication of intravenous thrombolysis administration. Large infarcts that present this increased risk are best measured by an ASPECTS score of less than 7 on noncontrast CT, lesions of greater than 70 to 100 ml of DWI positivity on MRI, and mean cerebral blood volume of less than 1.8 or Tmax delay of greater than 8 seconds for greater than 100 ml of brain tissue on perfusion imaging (04; 19; 10). Importantly, to achieve 100% specificity in identifying patients who will have poor outcome despite intravenous or intra-arterial reperfusion therapy, a threshold of 103 ml of DWI positivity was required (45). More recently, the benefit of thrombectomy in moderate to large completed infarct on presenting imaging was demonstrated in three randomized prospective trials (54; 20; 43). It is important to note that these excluded patients with large, well-demarcated ischemic strokes resulting in midline shift or herniation.

Concerning malignant middle cerebral artery syndrome, in a pooled analysis of the DECIMAL, DESTINY, and HAMLET trials, patients younger than 60 years of age with DWI lesions greater than 145 ml or CT evidence of greater than 50% of the middle cerebral artery territory had a clear survival benefit with decompressive hemicraniectomy compared to conservative management (50). These trials were conducted in the era prior to thrombectomy; therefore, the complete degree to which thrombectomy and hemicraniectomy may interact has yet to be seen.

Stroke etiology. Ischemic stroke may be caused by cardioembolic sources, carotid artery stenoses, vertebrobasilar stenosis, arterial dissection, intracranial atherosclerotic disease, vasculitis, or small artery occlusive disease. All of these entities are managed differently; therefore, identifying the underlying mechanism of stroke is important. The infarct pattern on MRI or CT can be quite useful as a guide to informing the workup for these potential etiologies. For example, in infarcts that extend from the cortex to subcortical structures, a cardioembolic source or large artery atherosclerosis is often found and, therefore, should be deeply investigated (52). An understanding of cerebrovascular anatomy and vascular distribution of ischemia should direct the appropriate workup and treatment strategy for secondary prevention of ischemic stroke.

Stenting or carotid endarterectomy are appropriate and powerful treatments for ipsilateral stroke in a vascular territory served by a stenotic cervical carotid artery (11). Maximal medical therapy is the appropriate first-line therapy for intracranial atherosclerotic disease (14), but other strategies such as intracranial stenting (05) or encephaloduroarteriosynangiosis (16), can be considered after failure of maximal medical therapy. With vertebral or basilar artery stenosis resulting in ischemic stroke, quantification of flow can aid in determining the ongoing stroke risk so treatment is proportional to this risk (06). Establishing small vessel stroke resulting from lipohyalinosis and thrombosis of perforator arteries is important to direct efforts toward treating risk factors for small vessel stroke, such as diabetes, hypertension, and smoking. Additionally, establishing small vessel stroke as an etiology may help direct efforts away from identifying or treating other causes of embolism, such as atrial fibrillation or patent foramen ovale.

Extracranial carotid and vertebral dissections may account for 10% to 25% of strokes in patients from 16 to 45 years of age, and conventional angiography is the gold standard for its evaluation given its superb resolution (51). The classic finding of intimal flap or double lumen is seen rarely, and dissecting pseudoaneurysm is seen more frequently (41). MR/MRA and CT/CTA are less invasive diagnostic measures that attempt to identify the same features of dissection. In addition, T1-weighted, fat-suppressed images may show hyperintense methemoglobin of an intramural hematoma in the false lumen of a dissection. In one study, CT/CTA and MR/MRI were comparable for detecting internal carotid artery dissection, whereas CT/CTA was superior for vertebral dissections (51). The most appropriate study for evaluating dissection will likely be that in which a given institution has the most experience.

Clinical vignette

An 84-year-old female with chronic atrial fibrillation and on warfarin was with her family the day prior to admission and was witnessed to be in her usual state of health 17 hours prior to presentation to the emergency room. On the day of admission, she woke with acute-onset aphasia with poor fluency and semantic and phonemic paraphasic errors.

Her NIHSS score on evaluation in the emergency room was 5 for severe aphasia, level of consciousness questions, and drift in her right upper extremity. Her international normalized ratio was 1.7; therefore, intravenous thrombolysis was not administered. Her pulse was 62, her blood pressure was 172/98, and her oxygen saturation was 96% on room air. She underwent code stroke imaging, which at our comprehensive stroke center includes a noncontrast head CT, CT perfusion with RAPID post-processing, and CTA of the head and neck. Her noncontrast head CT showed no early infarct signs and an ASPECTS score of 10. Her CTA showed a subtle and inconspicuous occlusion of the posterior division of the left middle cerebral artery in the M3 segment. The M3 occlusion was much more conspicuous on CT perfusion. As is often the case, CT perfusion served as an excellent distal occlusion detector. However, it should be noted that the area of Tmax >6 seconds is overestimated in this case related to signal in the orbits and posterior fossa. Given the patient’s aphasia and distal occlusion, the patient was taken for emergent catheter cerebral angiogram, which showed persistent M3 occlusion of the left MCA posterior division. Following one pass of mechanical thrombectomy with a 3 mm by 30 mm stent retriever and 4 max ACE intermediate catheter, TICI 3 reperfusion was achieved, and the thrombus was retrieved. The patient had mild clinical improvement immediately after thrombectomy, with resolution of mild weakness and improvement of her aphasia. On post-procedure day 2, she was improving but had residual aphasia. On post-procedure day 3, she had NIHSS of 0 and was discharged home. At the day 30 visit, she was fully back to her functional baseline. Going forward, she was able to drive, remained an avid reader, was active in her book club, and enjoyed playing bridge.