Computed tomography perfusion (CTP) has been established as a critical imaging technique for assessing the penumbra during interventions such as intravenous thrombolysis (IVT) and mechanical thrombectomy in an extended time window, i.e., ≥4h, 5h, or 6 h from symptom onset (1–4). CTP provides information not only about the presence of penumbra but also about specific types of brain ischemia. Such information could be helpful for stroke management, assisting not only in decision-making about acute treatments, such as intravenous thrombolysis (IVT), but also in overall patient care. At the time of admission, it is essential to know if a patient has experienced a stroke because such information is critical for patient management, which include decision-making about admission to the stroke unit, conducting dysphagia screening, managing blood pressure and temperature, among others. CTP might serve as an additional tool that can help in deciding whether a patient has a stroke or a stroke mimic.

The Czech Republic has one of the highest thrombolytic rates (with over 20% of ischemic stroke patients receive IVT) and one of the shortest door-to-needle times (DNT) for intravenous thrombolysis (with the median national door-to-needle times being approximately 20 min) (5). Although there are highly developed and effective stroke services and networks, many hospitals do not use CT perfusion imaging. The argument is that indications for mechanical thrombectomy are expanding, and patients in the later time window and with pervasive early ischemic changes are recommended for recanalization treatment (6, 7). However, the opinion from the centers that are using CTP is that CTP is changing clinical management not only because it provides penumbra imaging but also because it is increasing physicians' confidence in stroke diagnoses in general.

Therefore, this study aimed to explore the yield of CTP beyond the established and guidelines-supported recommendations. The primary goal of this study was to assess the accuracy of baseline CTP compared to a clinical stroke diagnosis in 0–6 h and 6–24 h time windows. To remove any subjectivity associated with a stroke diagnosis, we also compared CTP findings with radiological stroke diagnoses.

We conducted a cohort study that included patients with a suspected acute ischemic stroke within both the hyperacute phase (0–6 h) and the extended time window (6–24 h) admitted to the Neurology Department, with an acute neurological deficit at the time of admission between November 2018 and November 2019.

Microsoft Excel version 2010 and SPSS statistics software were utilized for performing analyses. The results were expressed as means, medians, and percentages. The sensitivity, specificity, and positive and negative predictive values were calculated separately for clinical and radiological stroke diagnoses.

From November 2018 to November 2019, 585 patients with acute neurological deficits underwent multimodal CT imaging. A total of 72 cases (12%) were excluded due to the CTP images were of poor quality or had severe motion artifacts. The study also excluded another six patients due to secondary transport to a comprehensive stroke center, and another seven patients were excluded due to incomplete data. The remaining 500 (85%) patients were included in the analysis: 232 (46%) patients in the 4 to 5-h thrombolytic window and 274 (55%) patients who had CTP in the 6-h window from the onset of symptoms. The average age was 71 ± 16 years, and 267 (53%) patients were female individuals. A flow diagram representing the excluded patients is shown in Figure 1. A total of 125 patients (25%) received intravenous thrombolysis, with a median door-to-needle time of 20 min. Table 1 shows all the baseline characteristics of the included patients.

Out of 500 patients, 185 (37%) had a positive CT perfusion was found in 185 patients (37%). A typical CTP pattern observed in a stroke patient is shown in Figure 2. All of these patients (100%) were dismissed with a final clinical diagnosis of stroke. Among the remaining 315 patients (63%) who showed negative initial brain perfusion, 179 (57%) had a clinical diagnosis of stroke, while the remaining 135 (43%) patients were identified as having conditions that mimin stroke. The overall sensitivity, specificity, and positive and negative predictive values of baseline CT for the clinical diagnosis of stroke were 51% [confidence interval (CI) 46–56], 100% (CI 97–100), 100% (CI 98–100), and 43% (CI 41–46), respectively. In the hyperacute phase (0–6 h from symptom onset), the sensitivity, specificity, and positive and negative predictive values of baseline CT for the clinical diagnosis of stroke were 53% (CI 46–60), 100% (CI 95–100), 100 (CI 97–100), and 41% (CI 38–45), respectively. The diagnoses of stroke mimics and their frequencies are shown in Table 2. In addition, the patient characteristics with the CTP-negative results, distinguishing between those with stroke mimics and strokes, are presented in Table 3, and a comparison of the patients in the hyperacute phase and extended time window is shown in Table 4. CTP-negative patients with stroke mimics were younger (66 ± 17 years vs. 73 ± 13 years) and had lower NIHSS scores [median 0, interquartile range (IQR) 0–2 vs. median 4, IQR 2–6] compared to CTP-negative patients with strokes. A bar chart demonstrating the distribution of NIHSS scores in patients admitted under the suspicion of stroke is shown in Figure 3.

A clinical (discharge) diagnosis of stroke in CTP-negative patients was established based on their clinical presentation and patient history in accordance with follow-up neuroimaging findings (N = 51), symptomatic vessel occlusion (N = 6), and significant symptomatic stenosis (N = 7). A total of 18 patients underwent intravenous thrombolysis, resulting in the resolution of symptoms and no infarction evident on the follow-up CT scan.

In patients who had negative initial CT perfusion with no follow-up imaging (n = 65), a definitive discharge diagnosis of stroke was determined based on the presence of symptomatic vessel stenosis or occlusion (n = 20), or evidence of infarction on the initial non-contrast CT scan (n = 5). In 46 out of 65 patients with negative initial CTP and no follow-up imaging, a stroke diagnosis was confirmed after the exclusion of stroke mimics. Most of these patients presented with lacunar syndrome.

Follow-up imaging (brain CT/MRI) was performed in 314 cases (63%). In 180 out of 314 cases (57%), new hypodensity during the CT scan was confirmed, and in 10 (5%) cases, the infarction was localized outside the area of the brain covered by baseline CTP. In CTP-positive patients, a radiological diagnosis of stroke was confirmed in 129 patients (81%), while in CTP-negative patients, 51 patients (33%) had infarction on follow-up imaging. CTP findings, follow-up imaging, and definite diagnosis distribution in all patients are shown in Figure 4. The overall sensitivity, specificity, and positive and negative predictive values of baseline CTP for the radiological diagnosis of stroke were 74% (CI 68–79), 94% (CI 88–97), 97% (CI 93–98), and 61% (CI 65–66), respectively. For patients with CTP within the 6 h of symptom onset, the sensitivity, specificity, positive and negative predictive values of baseline CTP for the radiological diagnosis of stroke were 80% (CI 72–86), 95 % (CI 86–99), 97% (CI 92–99), and 69% (CI 61–76), respectively.

In our study, all patients with an acute neurological deficit underwent CTP and 25% of them received IVT with a door-to-needle time of 20 min. Therefore, in the first place, our study documents the feasibility of conducting CTP as part of routine clinical practice. Such a finding has important clinical implications because some previous studies have demonstrated that CTP prolongs the initiation of IVT and thus should not be used (10). Our study documents that despite adding a few minutes for CTP imaging (usually 4–5 min), typically conducted in our center alongside unenhanced CT and CTA, CTP, in principle, does not pose a barrier to achieving very short DNT. The main reason is that patients are directly transported to undergo a CT scan from an ambulance, and IVT is initiated in a CT scanner, as we have published previously (7, 11).

The primary aim of our study was to compare the accuracy of baseline CTP to a clinical diagnosis of stroke. Our first major finding was that once CTP was positive, regardless of the time window, all patients had a stroke, and none were diagnosed with stroke mimics. Such information is beneficial because it eliminates any diagnostic doubts that physicians might have. Most likely, the value of stroke confirmation will be increased even more in those hospitals and countries where patients are not managed by stroke experts. Patients with positive CTP results should receive complete stroke care, which includes not only IVT but also admission to a stroke unit, management of blood pressure, and screening for dysphagia, among others, as per guidelines.

We were unable to confirm the results of some previous studies that have documented that hypoperfusion (and even hyperperfusion) on CTP imaging can also be found in patients with conditions other than stroke, such as status epilepticus or post-paroxysmal deficit (11–14). In our cohort, 23 patients had epileptic seizures, and all had negative CTP findings. The different results can be explained by the fact that in previous studies, no automatic software was used, no thresholds were defined, and the number of patients was low. As our study did not find any false positives, our results are applicable to clinical practice as long as certified postprocessing tools such as RAPID are used to evaluate CTP.

Our second major finding was that when CTP was negative, 57% had a final clinical diagnosis of stroke. When a radiological definition of stroke was used, 26% of CTP-negative patients suffered a stroke. False negatives were due to lacunar supratentorial infarction (24 patients) or vertebrobasilar stroke (66 patients). In addition, 10 infarcts were located in the brain area outside the coverage of baseline CTP. Our findings reflect the known limitations of brain perfusion imaging (15, 16). The calculated specificity of CTP against a radiological diagnosis of stroke (94%) is comparable to that reported in the most extensive published meta-analysis (17). We observed a numerically lower sensitivity (74%) compared to the radiological diagnosis, as reported in the previous largest and most recent analyses (80%) (17). If CTP covers the entire brain instead of a width of 8 cm, as in our study, the accuracy of CTP will slightly improve because it allows for the detection of infarctions in any part of the brain.

The low sensitivity and negative predictive value of CTP in a stroke can be effectively addressed by using MRI diffusion-weighted images (DWIs) in the acute phase of a stroke. In recent years, there has been a noticeable shift toward replacing CT with short 6–8-min MRI protocols. MRI demonstrates remarkably high sensitivity over 97% even in the initial minutes of a stroke (18–21). One of the most significant advantages of using MRI in clinical practice is its ability to reduce the incidence of thrombolysis stroke mimics by 50% (19). However, this approach is not without its drawbacks. The main disadvantages include the inferiority of MR angiography in vessel occlusion detection, typically distal, and the overestimation of the stenosis degree (20). These drawbacks should be carefully considered when making treatment decisions. Other disadvantages include a door-to-imaging time delay (a few minutes), the need for a second acute imaging modality for diagnostic or therapeutic reasons, higher costs, and MRI contraindications (e.g., recent vascular stents and pacemakers) (19, 21).

However, the remaining 43% of the patients with negative CTP were diagnosed with stroke mimics, which is quite a substantial number. These patients were discharged with another diagnosis, such as vertigo of non-vascular cause (23%), epilepsy (18%), orthostatic collapse, or others, as described in Table 2. Negative CTP-patients with stroke mimics were younger (68 years vs. 73 years, p < 0.001) and had lower NIHSS scores (0 vs. 4, p < 0.001) compared to negative CTP-patients with a stroke. Patients with stroke mimics also had fewer cardiovascular risk factors, most likely, due to younger age. The prevalence of stroke mimics in our study was 27%, closely aligning with the findings from the meta-analysis of 29 studies documenting a 26% prevalence of stroke mimics (9, 10). In addition, the diagnoses of stroke mimics were similar to those reported in the published data, as shown in Table 2. Therefore, our study documented a similar number and causes of stroke mimics. None of the patients with stroke mimics received IVT. We suggest that this is mainly due to the low NIHSS score and minor non-disabling deficits (median NIHSS 0 IQR 0–2). We found that if CTP is negative, it should raise a red flag in younger patients with minor neurological deficits and fewer cardiovascular risk factors because they might have stroke mimics. In cases where a patients presents with lacunar syndrome or there is suspicion of a stroke in the vertebrobasilar territory, negative CTP should not prevent treatment with intravenous thrombolysis.

One of the significant limitations of our study is the retrospective design. In a prospectively conducted study, CTP could be compared to the clinician's interpretation of the patient in the acute situation and non-contrast CT to assess the real added value of brain perfusion as a part of the standard imaging protocol. The other major limitation is the missing follow-up imaging in 37% of the cases. Such missing cases could limit conclusions about CTP accuracy against radiological diagnosis but not clinical stroke diagnosis. In our clinical practice, follow-up CT is omitted only when stroke or stroke mimic diagnoses are established. In clinically uncertain cases, we always perform an MRI. Therefore, missing follow-up imaging should not limit the validity of our study, although the confirmation of our results by conducting future studies will be reassuring. Another limitation of our study is that we excluded 12% of the patients with poor-quality CTP; such patients were excluded because they could not be assigned to either group (CTP positive or CTP negative). This exclusion of patients may have altered the specificity calculations, although the number of excluded patients is similar to other previously published studies (22–24). As mentioned earlier, our calculated specificity is comparable to the biggest published meta-analyses (17). Finally, although our results support the use of CTP in stroke management, we did not consider the economic cost in our study.

For the generalizability of our results, especially regarding the high proportion of stroke mimics, it might be essential to consider the referral pattern applied in our hospital, as described in the Methods section. Hospitals with different referral patterns may have fewer stroke mimics, mainly if some triage criteria are applied before imaging, and these hospitals might only refer the most apparent stroke candidates for CT scans. Our results related to stroke mimics (negative predictive value) are thus generalizable to unselected patients with acute neurological deficits. Our results related to the 100% positive predictive value of a stroke should be, however, applicable to any hospital regardless of the referral pattern.

The major strength of our study is the consistency in the use of multimodal imaging, which is not limited to any patient subgroup. This is the advantage over previous studies that included only patients with a stroke in anterior circulation (25), patients in a short time window (23), or patients with MRI follow-up imaging alone (24). Another advantage of our study is the relatively large number of patients, which was much less (usually up to 120 cases only) in previous studies (25–28).

There are several new essential findings in our study. First, we documented the feasibility of using CTP in routine clinical practice without negatively impacting stroke logistics, measured as the door-to-needle time for intravenous thrombolysis. Second, we documented the benefit of CTP beyond the guidelines recommended application, which is currently penumbra imaging ≥4, 5, and 6 h from symptom onset. Our study revealed the usefulness of CTP as a diagnostic tool, even in the hyperacute phase of a stroke. Positive CTP results improve physicians' confidence in diagnosing a stroke and rules out stroke mimics. However, negative CTP results should raise concerns about non-stroke diagnoses, especially in younger patients with minor deficits and non-lacunar syndrome. Negative CTP results should not prevent the IVT treatment.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by the Ethics Committee of St. Anne's University Hospital in Brno. The studies were conducted in accordance with the local legislation and institutional requirements. The Ethics Committee/institutional review board waived the requirement of written informed consent for participation from the participants or the participants' legal guardians/next of kin due to the observational and retrospective nature of the study.

MC: Conceptualization, Formal analysis, Writing – original draft, Writing – review & editing. MH: Data curation, Formal analysis, Methodology, Writing – review & editing. JV: Formal analysis, Methodology, Resources, Writing – review & editing. JŠ: Conceptualization, Investigation, Resources, Writing – review & editing. IF: Data curation, Formal analysis, Writing – review & editing. RM: Conceptualization, Funding acquisition, Supervision, Validation, Writing – review & editing.