Article Text

Download PDFPDF

Interpreting CT perfusion in stroke
  1. Stevan Christopher Wing1,
  2. Hugh S Markus2
  1. 1 Department of Neurology, Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK
  2. 2 Department of Clinical Neurosciences, University of Cambridge, Cambridge, UK
  1. Correspondence to Dr Stevan Christopher Wing, Department of Neurology, Cambridge University Hospitals, NHS Foundation Trust, Cambridge CB2 0QQ, UK; stevan.wing{at}


CT perfusion images can be rapidly obtained on all modern CT scanners and easily incorporated into an acute stroke imaging protocol. Here we discuss the technique of CT perfusion imaging, how to interpret the data and how it can contribute to the diagnosis of acute stroke and selection of patients for treatment. Many patients with acute stroke are excluded from reperfusion therapy if the onset time is not known or if they present outside of traditional treatment time windows. There is a growing body of evidence supporting the use of perfusion imaging in these patients to identify patterns of brain perfusion that are favourable for recanalisation therapy.

  • stroke
  • cerebral blood flow
  • cerebrovascular disease
  • fibrinolysis
  • neuroradiology

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Traditionally non-contrast CT alone is used in the initial assessment of patients presenting with a clinical diagnosis of acute ischaemic stroke, primarily to exclude haemorrhage and alternative pathology such as tumours. Increasingly CT angiography is used as part of the initial imaging protocol to identify large vessel occlusion that may be amenable to endovascular therapy. Conventionally, for those patients whose imaging is appropriately supportive, selection for reperfusion therapy is based on clinical assessment and the time since symptom onset. However, while time can predict the extent of salvageable tissue, we have known since the 1990s that the rate at which tissue dies following cerebral artery occlusion varies widely between individuals, largely due to the differing extent of collateral supply.1 Increasing evidence suggests that we should take the extent of ‘ischaemic penumbra’ (defined as the extent of potentially salvageable tissue) into account when deciding in whom to attempt recanalisation.2 This has been highlighted by the recent publication of the DEFUSE-III (Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke 3)3 and DAWN (DWI or CTP Assessment with Clinical Mismatch in the Triage of Wake-Up and Late Presenting Strokes Undergoing Neurointervention with Trevo)4 trials, which showed that, in patients with potentially salvageable tissue identified on CT perfusion or MRI, thrombectomy was effective even up to 24 hours after stroke onset.

CT perfusion allows various aspects of cerebral perfusion to be determined. Its primary role in acute stroke is in determining whether brain tissue is hypoperfused and therefore at risk of infarction, or whether it is already irreversibly damaged. It can also make major contributions to diagnosis.

The DAWN and DIFFUSE-III trials provided considerable evidence that CT perfusion can influence decision-making in acute stroke; it is therefore likely to become part of the standard acute stroke imaging protocol in many centres, particularly in the extended time window. Here we discuss the principles behind CT perfusion imaging, how to interpret the data and how the technique can help in clinical practice.

What is CT perfusion imaging for acute ischaemic stroke?

The aim of performing perfusion imaging in the context of cerebral ischaemia is to distinguish likely infarcted and unsalvageable areas of brain (the ischaemic core) from areas of potentially salvageable (penumbral) tissue. It is this viable penumbra that is the target for reperfusion therapy. Those with a large established core of infarct are unlikely to benefit from reperfusion and are at a greater risk of complications such as haemorrhagic transformation should it be pursued. MRI can provide estimates of penumbral tissue through the combination of diffusion and perfusion imaging, but in practice this is less available and harder to implement in routine stroke care. In contrast, CT perfusion can be performed in a few minutes on all modern CT scanners.

During CT perfusion acquisition, the brain is repeatedly scanned during the intravenous infusion of iodinated contrast media. As the contrast flows through the region, the relative increase, peak and then decrease in radiodensity, measured in Hounsfield units, allows an attenuation–time curve to be derived (figure 1). These curves are calculated for an arterial input function and a venous outflow function, allowing various measures of perfusion to be calculated for each voxel. These measures of perfusion include cerebral blood flow, cerebral blood volume, time to peak, time to drain, mean transit time and tissue permeability (figure 1). Each of these is typically displayed as a map of the brain with a colour scale representing the values (figure 2C–H). Postprocessing methods vary between scanner manufactures, acquisition methods and software packages and so other measures may be produced, such as the time to maximum (Tmax) of the residual function (in deconvolution-based methods) and delay time (delay time to residual function peak taking into account the arterial transport function).

Figure 1

Attenuation–time curve for a given region of interest. The time to peak (TTP) is the time taken for the maximum peak of attenuation and time to drain (TTD) the return to baseline. Cerebral blood flow (CBF) corresponds to the slope of the curve and cerebral blood volume (CBV) the area underneath it. The mean transit time (MTT) is the average time taken for contrast to flow through a brain region. HU, Hounsfield units.

Figure 2

A 55-year-old right-handed woman presented 40 min after a sudden onset of right-sided hemiparesis and aphasia while at the gym. Her initial non-contrast CT scan (A) showed a hyperdensity in the left middle cerebral artery (arrow) but was otherwise normal. Her CT angiogram (B) showed occlusion of the M1 segment of the left middle cerebral artery (white arrow) with good collateral circulation. Her CT perfusion showed delay in time to peak (E), mean transit time (F) and time to drain (G) with reduced cerebral blood flow in the left middle cerebral artery territory (C). There appears to be no change on the map of permeability (H). On reviewing the cerebral blood volume map (D) there are largely normal values with slight increase in volume more superficially, although there is a small area of matched deficit in the left corona radiata indicating core infarct (arrow). This is consistent with a large volume of ischaemic penumbra and a favourable profile for reperfusion therapy.

  • Cerebral blood volume is defined as the total volume of flowing blood in a given volume in the brain; low flow is considered a marker of already infarcted tissue.

  • Cerebral blood flow is a measure of rate, and is defined as the volume of blood flowing through a given volume of brain per unit time. Low cerebral blood flow values are associated with hypoperfused tissue, which could be salvageable; however, more severe cerebral blood flow reductions are also considered a marker of infarction.

  • The time to peak is the time from the start of injection until the maximum peak of contrast enhancement.

  • The mean transit time is the average time taken for contrast to flow through a brain region.

  • The Tmax represents the time from the start of the scan until the maximum intensity of contrast material arrives at each voxel.

  • The time to drain is the time from maximum enhancement to a defined low cut-off.

The ischaemic ‘core’ represents likely irreversibly damaged tissue and is identified by markedly reduced cerebral blood flow and reduced cerebral blood volume, with a marked delay in time to peak and mean transit time (see figure 3). The ischaemic penumbra—which in most cases surrounds the infarct core—also has prolonged mean transit time or Tmax but in contrast has only moderately reduced cerebral blood flow; importantly, it also has near normal or even increased cerebral blood volume due to autoregulatory vasodilatation. In this state, termed luxury perfusion, there is usually reduced cerebral metabolism despite the relative hyperaemia.

Figure 3

A 73-year-old right-handed man with a history of hypertension and myocardial infarction presented 80 min after experiencing a sudden onset of left arm and leg weakness while gardening. On examination, he had left facial weakness and dense left hemiparesis with neglect. The non-contrast CT (A) shows early ischaemic change in the right hemisphere with a hyperdensity in the M1 segment of the right middle cerebral artery (arrow). On CT angiography (B) the proximal M1 occlusion is confirmed (arrow). Paucity of vessels in this hemisphere is due to the occlusion and lack of collateral vessels. There is delay in the time to peak (C) with reduced cerebral blood flow (D) in the corresponding regions. The deficit in cerebral blood flow is matched by a reduction in cerebral blood volume (E) representing a large core infarct. This scan indicates no penumbral tissue and reperfusion is unlikely to confer any benefit.

The terms matched and unmatched deficits describe these patterns of perfusion and refer to whether or not the reduction in cerebral blood flow is matched by a reduction in cerebral blood volume (table 1). Extensive validation work from multiple groups has shown that cerebral blood flow less than 30% of normal tissue is highly predictive of core, and either Tmax more than 6 s or delay time more than 3 s is an accurate measure of penumbra.5

Table 1

Perfusion measurements derived from CT perfusion acquisitions and their usual findings for core and penumbra in acute stroke

How to interpret CT perfusion acquisitions

Although this article focuses on CT perfusion, this usually forms part of a multimodal CT assessment that also includes a non-contrast CT and CT angiography. Each of these scans gives valuable information. Adding CT perfusion requires only a little extra time on modern CT scanners. In our unit it takes a total of 4 min to acquire the data and process it using the automated MIStar software (see below) and to return the processed results to

In looking at the perfusion data, our preferred method is, first, to try to confirm whether there is regional hypoperfusion conforming to a vascular territory that would explain the clinical picture and, second, to decide on the extent of both core and penumbra. If there is an area of focal hypoperfusion, the time to peak is always delayed and is a sensitive measure. We then review the cerebral blood flow map. In ischaemia, there should be corresponding reductions in the areas identified on the time to peak map. We look at the intensity of the reduction in cerebral blood flow; severe reduction indicates already infarcted tissue. In a patient with focal hypoperfusion we make comparisons between the cerebral blood flow and the cerebral blood volume maps. Areas with both reduced cerebral blood flow and cerebral blood volume are considered unsalvageable and likely to become infarct (core); however, areas of near normal or increased cerebral blood volume within the areas of reduced cerebral blood flow represent hypoperfused penumbral tissue, an ideal target for reperfusion therapies. However, rigorous analyses have shown that Tmax or delay time provides more accurate measures of penumbral tissue than cerebral blood volume.5 We estimate the relative volumes of these areas: the amount of core, the amount of penumbra and the core:penumbra ratio.

Automated software for analysing CT perfusion data can be valuable in quantifying the volumes of both core and penumbra using established thresholds. We use MIStar (Apollo Medical Imaging Technology) but other programmes include RAPID CTP (iSchemaView). The software can generate parametric maps superimposed onto the non-contrast CT scan of these patterns of perfusion. Additionally, having set the desired criteria, the software can compute and give guidance on whether, in imaging terms, the patient would be suitable for therapy. Such software has the advantage of increasing interobserver reproducibility of CT perfusion and ensuring that validated thresholds are used, but it is important not to rely only on the output from the software but also to check the images to ensure it has not misinterpreted these; for example, if they are poor quality or have con-concomitant pathology such as old infarcts that could confuse the software.

How can CT perfusion help decision-making?

In diagnosis

CT perfusion is more accurate at diagnosing acute ischaemic stroke than non-contrast CT alone, with two pooled analysis reporting high sensitivity (80%–82%) and very high specificity (95%–96%) for stroke.6 7 However, it is important also to be aware of its limitations in diagnosis. CT perfusion has a relatively limited spatial resolution and therefore can miss small strokes, particularly lacunar infarcts, although it is still more sensitive for lacunar infarcts during the hyperacute phase than non-contrast CT. In one study, CT perfusion had a sensitivity of 62% in detecting diffusion-weighted MRI confirmed lacunar infarcts compared with just 19% for non-contrast CT.8 Sensitivity for posterior circulation infarcts is lower than for hemispheric cortical infarcts, but CT perfusion is more sensitive in diagnosing acute posterior circulation stroke than non-contrast CT, with reported sensitivity of 76% and specificity of 93%.6

We frequently find CT perfusion helps in diagnosing a stroke mimic that would otherwise be undetectable on non-contrast CT or CT angiogram. In functional weakness CT perfusion should be normal; however, it is important to remember that a normal CT perfusion scan does not exclude a lacunar infarct. Therefore, in a patient with an aphasia and hemiparesis, a normal CT perfusion study will argue against a stroke cause, while in a patient with an isolated hemiparesis one cannot exclude a stroke on the CT perfusion alone. CT perfusion can be abnormal in several other acute neurological presentations. Perfusion changes associated with migraine are well documented and a commonly reported sign is minor pan-hemispheric delays in time to peak and cerebral blood flow accompanied by normal values for cerebral blood volume; there are several described patterns (see (figure 4) for an example). Seizures may both increase and reduce cerebral perfusion. Postictally the most common changes are a prolonged mean transit time with a fall in both cerebral blood flow and cerebral blood volume; these changes are usually confined to the cortex and do not conform to expected vascular territories.9 Interictally there can be focal hyperperfusion with increases in both cerebral blood flow and cerebral blood volume and conversely, delays in time to peak and mean transit time. Tight carotid stenosis causing haemodynamic compromise can reduce the transit time; occasionally, if there is very severe haemodynamic compromise, cerebral blood flow can be reduced with preserved cerebral blood volume. Therefore, the CT perfusion scan should be interpreted in conjunction with imaging of the cerebral arteries (see figure 5).

Figure 4

In this case, a 34-year-old right-handed woman presented with a 30 min evolution of word finding difficulties and a sudden onset of right facial droop in association with a frontal headache on a background of migraine with aura. On examination, she had decreased visual acuity in the right temporal visual field. Non-contrast CT and CT angiogram were normal. Her time to peak map shows some minor delays (A) but there is a striking delay in the time to drain in the left hemisphere (B). The cerebral blood flow map shows diffuse reductions in flow across the left hemisphere in an unusual distribution for a vascular process (C). The cerebral blood volume map shows matched reductions with some scattered areas of normal perfusion (D). Her symptoms resolved spontaneously and completely a short while after presentation. We concluded the perfusion changes most likely represented migraine.

Figure 5

A 51-year-old right-handed man had developed a left middle cerebral artery ischaemic stroke 6 months before that had left him aphasic with a right hemiparesis; he presented 3 hours after experiencing worsening strength of his right leg in the context of a chest infection. His non-contrast CT scan showed old infarcts in the left centrum semiovale, left superior frontal and right parietal lobes, moderate small vessel disease and bilateral frontal lobe atrophy. His time to peak map shows bilateral and symmetrical delays in the frontal lobes (A) with reduced cerebral blood flow (B). His cerebral blood volume is relatively preserved (C). His CT angiogram (D) showed near occlusions of both internal carotid arteries (arrows). He was fluid resuscitated, given antibiotics and his symptoms resolved an hour after presentation. This case shows an unmatched deficit due to tight haemodynamically significant internal carotid artery stenosis in the absence of infarction and the CT perfusion findings are similar to that of penumbral tissue. However, this is a chronic state and the frontal lobe atrophy is consistent with chronically poor perfusion in this region. This situation is termed cerebral misery perfusion where the cerebral autoregulatory capacity in these brain regions has been exhausted and blood flow is insufficient to meet metabolic demands. Relatively minor changes in blood pressure or perfusion can produce transient symptoms resembling stroke or can, in severe cases, cause infarction in watershed areas.


CT perfusion predicts final infarct size

The volumes of CT perfusion-defined core infarcts correlate well with diffusion-weighted MRI lesion volumes at presentation, and also predict infarct size on MRI after 24 hours (see figure 6).6 In those who do not achieve reperfusion, CT perfusion predicts the resultant infarct volume from progression of hypoperfused tissue (core and penumbra volumes combined) and predicts the infarct volume in those where the at-risk tissue is saved (core volume only). Established thresholds for defining core and penumbra have been derived from comparing CT perfusion to diffusion-weighted MRI; however, these comparisons are largely from patients who have either not received therapy or were treated with thrombolysis.10 There is recent evidence to suggest that with earlier times to reperfusion, particularly in the context of thrombectomy, where recanalisation may occur earlier and be more complete, these thresholds may overestimate the infarct size and underestimate penumbra.10

Figure 6

The 55-year-old woman in figure 2 was referred for thrombectomy. Angiography confirmed a complete left middle cerebral artery occlusion (arrow) with good collateral circulation (A). She underwent thrombectomy with good angiographic results (B) and a complete recovery of her symptoms on the angiography table. Her diffusion-weighted MRI at 24 hours showed lesions in the areas identified on CT perfusion to be core infarct. She left the hospital without symptoms.

Clinical prognosis

CT perfusion deficits reflect clinical assessments of severity. The volume of severely hypoperfused tissue on CT perfusion closely correlates with the baseline National Institutes of Health Stroke Scale (NIHSS) score11; the rescue of penumbral tissue (defined by CT perfusion) also correlates highly with improvements in NIHSS.11 Both larger core and penumbral lesions are associated with poorer long-term outcomes and patients who present with a large established core have higher risks of complications from reperfusion therapy. The progression to infarction in at-risk areas on CT perfusion is strongly associated with worse outcomes and salvage with clinical recovery.12

Interestingly, improvements in clinical outcome are explained by the salvage of penumbra as estimated on CT perfusion, and after CT perfusion data is taken into account, time to treatment no longer predicts the outcome.12 This suggests that time is essentially acting as a surrogate marker for the extent of salvageable tissue, and the latter is the key determinant of whether reperfusion will result in clinical benefit.

Selection for therapy

Increasing evidence suggests CT perfusion may help us identify which patients may benefit from intravenous thrombolysis and thrombectomy, even up to 24 hours after stroke onset. In positive trials of mechanical therapy for anterior circulation stroke with large vessel occlusion, there were better outcomes where perfusion imaging was included in the patient selection criteria (SWIFT-PRIME and EXTEND-IA, which used CT perfusion) compared with non-contrast CT alone (REVASCAT and ESCAPE, which used NCCT-ASPECTS).13 Perhaps the biggest impact will be in patients presenting beyond the conventional time windows. Both the DAWN and DEFUSE-III trials showed that the treatment time window can be extended up to 24 hours in certain patients selected for on the basis of perfusion imaging. Both trials were terminated early and showed improved functional outcomes that amounted to the biggest treatment effects seen among acute stroketherapy trials. These two studies show the advantages of stratifying acute stroke patients in terms of their tissue physiology rather than setting hard time constraints.

Current trials are assessing whether CT perfusion may allow patients who wake up with stroke to be selected for intravenous thrombolysis even if the exact time of onset remains unknown. The recent publication of the WAKE-UP trial supports this approach when using MRI markers of penumbra.14 Since CT perfusion accurately estimates unsalvageable tissue, it is possible that CT perfusion may allow patients within conventional time windows to be spared intravenous thrombolysis if there is no penumbra and little to be gained from reperfusion, although this needs more evaluation.


CT perfusion imaging can be performed rapidly at the time of initial imaging and improves diagnostic confidence and accuracy in the patient with hyperacute stroke. It can identify potentially salvageable areas of brain in order to select patients better for reperfusion therapies. Recent clinical trials have shown that CT perfusion can identify those patients who may benefit from reperfusion up to 24 hours, and this is likely to lead to it being used much more widely. Familiarity with the technique is therefore important for a neurologist or stroke physician involved in acute stroke care.

Further reading

  • Albers GW, Marks MP, Kemp S, et al. Thrombectomy for stroke at 6 to 16 hours with selection by perfusion imaging. N Engl J Med 2018;378:708–718.

  • Nogueira RG, Jadhav AP, Haussen DC, et al. Thrombectomy 6 to 24 hours after stroke with a mismatch between deficit and infarct. N Engl J Med 2018;378:11–21.

  • Thomalla G, Simonsen CZ, Boutitie F, et al. MRI-guided thrombolysis for stroke with unknown time of onset. N Engl J Med 2018; DOI: 10.1056/NEJMoa1804355 (accessed 19 May 2018).

  • Textbook of stroke medicine. 2nd edn. Cambridge University Press; 2014.

Key points

  • CT perfusion imaging can be rapidly conducted at the time of initial imaging at the same time as non-contrast CT and can increase the diagnostic accuracy of stroke

  • CT perfusion can differentiate potentially salvageable “penumbra” from likely unsalvageable infarcted core

  • Focally increased time to peak and reduced cerebral blood flow combined with reduced cerebral blood volume indicates irreversibly damaged tissue (“core infarct”); low cerebral blood flow with a near normal or increased cerebral blood volume is characteristic of salvageable penumbra, which is an ideal target for reperfusion

  • Selection of patients for reperfusion therapy on the basis of perfusion imaging has been successful in large clinical stroke trials; initial evidence suggests that patients presenting outside of traditional time windows for intervention can have good outcomes if CT perfusion shows a favourable profile of tissue physiology


We are grateful to Dr Andrew Bivard for reviewing this manuscript.


View Abstract


  • Contributors SCW wrote the manuscript. SCW and HSM revised the manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed by Joshua Klein, Boston, USA.

Other content recommended for you