Computed Tomography–Based Imaging Scores in Basilar Artery Occlusions: A Comparison of Predictive Abilities for Functional Outcomes

Abstract

BACKGROUND:

Posterior circulation large-vessel occlusion strokes have significant morbidity and mortality, but patient selection for acute interventions remains understudied. Multiple computed tomography (CT)–based scores exist, including the CT perfusion–based critical area perfusion score (CAPS), CT angiography–based basilar artery treatment management and posterior circulation CT angiography scores, and CT angiography source image or noncontrast CT–based posterior circulation Alberta Stroke Program Early CT Score, but their predictive values for long-term outcomes after thrombectomy have not been directly compared.

METHODS:

We conducted a retrospective multicenter cohort study of patients with basilar artery occlusions treated with endovascular thrombectomy. Four CT-based scores were assessed: posterior circulation Alberta Stroke Program Early CT Score, basilar artery treatment management score, posterior circulation CT angiography score, and CAPS. The primary outcome of interest for the study was a favorable functional outcome at 3 months (modified Rankin Scale score, 0–3). We calculated sensitivity, specificity, positive and negative predictive values, and generated receiver operating characteristic curves measuring area under the curve (AUC), which were compared with nonparametric methods.

RESULTS:

Ninety-eight patients were included for analysis, with an average age of 64.9 ± 15.6 years. The median National Institutes of Health Stroke Scale score was 13.5 (interquartile range, 7.0–23.0). AUC values were highest for the CAPS (AUC, 0.72 [95% CI, 0.63–82]) and lowest for the posterior circulation CT angiography score (AUC, 0.57 [95% CI, 0.45–0.68]; P=0.019). There was a trend toward the CAPS outperforming the basilar artery treatment management score (AUC, 0.66 [95% CI, 0.55–0.77]) and posterior circulation Alberta Stroke Program Early CT Score (AUC, 0.63 [95% CI, 0.52–0.75]) though this was not statistically significant (P=0.29 and P=0.23, respectively). CAPS ≤3 demonstrated 100% sensitivity and negative predictive value (95% CI, 92.6–100.0%; 95% CI, 75.7–100.0%, respectively) for good outcome but had low specificity (24.0% [95% CI, 14.3–37.4]).

CONCLUSIONS:

The CT perfusion–based CAPS had the best performance for predicting good functional outcome at 3 months and had high sensitivity and negative predictive values though our finding is limited by the small number of patients who had CAPS >3. The CAPS should be validated in prospective studies.

CLINICAL PERSPECTIVE.

What Is New?

• In a multicenter cohort of patients with basilar artery occlusion treated with endovascular thrombectomy, the computed tomography perfusion–based critical area perfusion score demonstrated the highest overall discriminative ability for predicting 90-day functional outcome compared with computed tomography angiography–based and noncontrast computed tomography–based posterior circulation scores.

• Critical area perfusion score ≤3 showed 100% sensitivity and negative predictive value for good functional outcome (modified Rankin Scale score, 0–3), including in patients with successful recanalization, highlighting its strength as a rule-out imaging marker for unfavorable prognosis.

What Are the Clinical Implications?

• Computed tomography perfusion–based scoring with critical area perfusion score may support more standardized and reliable prognostication in acute basilar artery occlusion, helping clinicians identify patients unlikely to benefit from thrombectomy while reinforcing the need for prospective validation before routine adoption.

Posterior circulation large-vessel occlusions remain one of the most challenging forms of cerebrovascular disease to treat, given their significant morbidity and mortality despite substantial advancements in the treatment of ischemic stroke over the past decade.¹ Initial posterior circulation thrombectomy trials yielded controversial results.^2,3 With the relatively recent emergence of clinical trial evidence demonstrating the efficacy of endovascular thrombectomy in patients with acute basilar artery occlusions (BAOs), it is evident that patient selection for hyperacute treatment in this high-risk population is crucial; however, direct translation of anterior circulation imaging–based patient selection strategies using structural, angiographic, and perfusion features necessitates modified or reformulated approaches to evaluating the extent of disease.

Many radiographic prognostic scores exist with the objective of identifying the severity of ischemic injury to identify patients who will benefit from endovascular thrombectomy of the vertebrobasilar system. The posterior circulation Alberta Stroke Prognosis Early Computed Tomography Score (pc-ASPECTS) was among the first and utilizes either noncontrast computed tomography (CT) or CT angiography (CTA) source images, with improved sensitivity for outcome prediction using CTA source images compared with noncontrast CT.⁴ Since 2008, acute imaging for ischemic stroke has further evolved, with continuous improvements in CT technology, with larger detector coverage allowing accelerated acquisitions and improved iterative reconstruction strategies to support radiation dose reductions. CTA-based scores such as the basilar artery treatment management (BATMAN) and posterior circulation CTA (pc-CTA) scores measure collateral status,^5,6 which are associated with a higher chance of good functional outcomes at 3 months.⁷ The critical area perfusion score (CAPS) uses CT perfusion to identify areas of extreme hypoperfusion (ie, time to maximum of the residue function >10 s) as a surrogate marker for critical penumbra that is at risk for progression to ischemic core, which is a negative predictor for favorable outcomes postthrombectomy.⁸ However, time to maximum of the residue function >10 s volume alone is an imperfect surrogate for posterior circulation ischemia.^9,10 CAPS incorporates both perfusion severity and anatomic topographical weighting, rather than relying on a single perfusion threshold, which has been previously shown to outperform single perfusion thresholds in BAOs.¹¹

Despite the existence of multiple imaging scores for prognostication in acute basilar occlusions, their predictive values for outcomes after acute reperfusion therapy have not been directly compared with each other. In this multicenter cohort study, our objective was to evaluate the predictive abilities of 4 CT-based imaging scores for 3-month functional outcome after thrombectomy. Our hypothesis was that a CT perfusion (CTP)–based imaging score would more accurately distinguish those who achieve good functional outcomes compared with other CT imaging modalities.

Methods

The data that support these findings are available upon reasonable request.

Study Design and Setting

We conducted a multicenter retrospective cohort study of patients with BAO who underwent cerebral perfusion and vessel imaging before thrombectomy from January 1, 2015, to December 31, 2019. Institutional review board approval was provided for inclusion in the study by each site, and the study complied with the Health Insurance Portability and Accountability Act and the guidelines of the Declaration of Helsinki. Patient informed consent was waived by our local institutional review boards.

CT Perfusion Image Acquisition

An iodinated intravenous contrast bolus (35–50 mL) was power-injected into an antecubital vein at a rate of 4 to 5 mL/s. Serial brain images were acquired after the contrast injection using either 1- or 2-slab acquisition. Images were processed using RAPID (v4.9; iSchemaView, Menlo Park, CA). Automated arterial input functions were automatically assigned by RAPID and could include either anterior or posterior circulation vessels. All images were checked and deemed to be of sufficient quality before inclusion in the cohort. This initial quality control was performed by the individual sites that contributed patients and imaging data. Exclusion from the current cohort included noninterpretable CT or magnetic resonance perfusion imaging due to excessive patient motion, incomplete coverage of the posterior fossa, or failed contrast bolus.⁸ No independent confirmation or adjudication was performed.

Participants

We included consecutive adult patients with acute posterior circulation ischemic stroke due to BAO diagnosed on either CT or magnetic resonance angiography and a National Institutes of Health Stroke Scale (NIHSS) score of >2, and thrombectomy was performed within 36 hours from symptom onset. Initial exclusion criteria included significant prestroke functional dependence (ie, modified Rankin Scale score ≥2), concurrent anterior circulation large-vessel occlusion, or poor-quality perfusion images. We excluded 5 patients with incomplete vessel imaging for assessment of collateral-based scores.

Variables

The 4 imaging scores assessed in this study are summarized in Table S1. Imaging scores were assessed by a stroke fellow (R.L.) and a senior neurology resident (P.S.). Readers were blinded to clinical outcomes and to the other reader’s assessments. Disagreements were resolved by review with a senior neuroradiologist with over 10 years of experience (J.J.H.) to maximize the accuracy of each score assessed. The pc-ASPECTS score was assessed using CTA source images if available or noncontrast CT if CTA source images were unavailable. Dichotomization of each imaging score was reported according to their original publications according to their respective associations with outcomes.^4–6,8 For the pc-CTA score, we performed an exploratory analysis of dichotomization at various thresholds and found that the cutoff of <3 versus ≥3 was associated with higher predictive abilities for good functional outcomes at 3 months.

Outcomes

The primary outcome of interest for the study is a favorable functional outcome at 3 months, defined as a modified Rankin Scale score of 0 to 3, ascertained via phone interview or in-person visit as part of routine clinical care. Other outcome measures of interest include (1) symptomatic intracerebral hemorrhage, defined as new intraparenchymal or subarachnoid hemorrhage with an associated NIHSS score decline of ≥4 and (2) successful reperfusion, defined as modified Thrombolysis in Cerebral Infarction score of 2b-3.

Statistical Methods

Categorical variables are presented as counts (percentages). Continuous variables were visually inspected for normality based on histograms; normally distributed variables are reported as means±SDs, and nonnormal variables are reported as medians with interquartile ranges. We generated receiver operating characteristic (ROC) curves for each score for the primary outcome. We computed area under the curve (AUC) values and 95% CIs using stratified bootstrap replicates. To ensure consistent directionality between the imaging scores and outcomes, we presented the axis of the plots for the ordinal outcome scale for the BATMAN and pc-ASPECTS scores. ROC curves were constructed for each model using the same patient cohort. AUCs were compared using paired, nonparametric tests for correlated ROC curves, equivalent to pairwise DeLong tests, to account for within-subject correlation. Reported P values represent the null hypothesis that the difference in AUCs between models equals zero.

Next, we performed a sensitivity analysis of 3-month clinical outcomes according to the 4 dichotomized imaging scores when the population was isolated to a successful recanalization cohort (defined as Thrombolysis in Cerebral Infarction of 2b or greater). This is to avoid the potential confounding effect of failed thrombectomy. Last, we calculated the sensitivities, specificities, positive predictive values, and negative predictive values from 2 × 2 contingency tables for the 4 dichotomized imaging scores, in the entire cohort, as well as in the successful recanalization-only cohort. To ensure that directionality is consistent for ease of interpretation, test positive was defined as the favorable imaging category for all 4 scores (ie, CAPS ≤3, BATMAN score, 8–10, pc-CTA score <3, and pc-ASPECTS score, 8–10). 95% CI were computed using Wilson binomial intervals. All statistical analyses were conducted using SPSS, version 29.0 (IBM, Armonk, NY).

Results

The original cohort had 103 patients. Details surrounding enrollment are outlined in Figure 1. We excluded 5 patients because they did not have vessel imaging for assessment of the 2 collateral-based scores. Baseline demographics and clinical characteristics of the 98 included patients are summarized in Table 1. Additional details surrounding baseline demographics and characterizations according to the stratified CAPS have also been previously summarized in the original paper by Cereda et al,⁸ published in Annals of Neurology 2022 (Table 1). The included patients were predominantly men (62/98; 63.3%), with an average age of 64.9 ± 15.6 years. Hypertension was the most common comorbidity (73/98; 74.5%), followed by hyperlipidemia (44/98; 44.9%), diabetes (25/98; 25.5%), and coronary artery disease (22/98; 22.4%). Approximately 20% (19/98) of patients had atrial fibrillation, and 21/98 (21.4%) were current smokers at the time of their stroke. There were 40/98 (40.8%) patients on antiplatelets or anticoagulants. The median NIHSS score was 13.5 (7.0–23.0) at presentation (Table 1). The median time from last seen normal to presentation was 4.0 (1.8–9.1) hours, and about 1 of 5 of patients were wake-up strokes. In total, 32/98 (32.7%) patients received intravenous thrombolysis.

Figure 1.

STROBE diagram (Strengthening the Reporting of Observational Studies in Epidemiology) outlining the number of participants from each site. AC indicates anterior circulation; CT, computed tomography; CTA, computed tomography angiography; LVO, large-vessel occlusion; and MR, magnetic resonance.

Table 1.

Baseline Demographics and Clinical Characteristics of Included Patients

The distributions of the frequencies of all 4 imaging scores are displayed in Figure 2. The majority of patients had a favorable profile according to every imaging score except the BATMAN score. A favorable CAPS (≤3) was present in 86 of 98 (87.8%) patients, 57 of 98 (58.2%) patients had a favorable pc-CTA score of <3, and 71 of 98 (72.4%) patients had a favorable pc-ASPECTS of 8 to 10 (reversed to 0–2). For the BATMAN score, only 36/98 (36.7%) had a favorable score of 8 to 10. The medians of all 4 imaging scores are outlined in Table 2. In addition, imaging characteristics are reported stratified by each score’s dichotomization cutoff according to their original publications.^4,5,8

Figure 2.

Distribution of the frequency of scores for each of the 4 imaging scores. A, Critical area perfusion score (CAPS). B, Basilar artery treatment management (BATMAN) score. C, Posterior circulation computed tomography angiography (pc-CTA) score. D, Posterior circulation Alberta Stroke Program Early Computed Tomography Score (pc-ASPECTS) score.

Table 2.

Imaging and Treatment Outcomes According to Dichotomization of Each Imaging Score According to Their Original Publications

We performed a sensitivity analysis where our cohort was restricted to patients who achieved a Thrombolysis in Cerebral Infarction score of 2b or greater after thrombectomy to avoid the confounding effect of a failed thrombectomy. There were 83/98 patients who achieved successful recanalization, of whom 47 of 83 (56.5%) achieved functional independence, and 28 of 83 (33.7%) died (Table 3). The distributions of imaging scores were similar in this subgroup of patients. There were 9 patients who had a CAPS of >3, and none of these patients achieved functional independence at 3 months (Table 3).

Table 3.

90-Day Outcomes According to Dichotomization of Each Imaging Score Only in Patients Who Achieved Recanalization (Defined as TICI Score of 2b or Higher)

Last, we calculated the sensitivity, specificity, positive predictive values, and negative predictive values for the entire cohort and for the successful recanalization subgroup. In the overall cohort, diagnostic performance varied across imaging scores, reflecting differing tradeoffs between sensitivity and specificity. CAPS ≤3 demonstrated perfect sensitivity and negative predictive values for good functional outcome but had low specificity, indicating strong rule-out performance at the expense of false positives. In contrast, a BATMAN score of 8 to 10 showed lower sensitivity but higher specificity and positive predictive value, suggesting better discrimination for identifying patients likely to achieve a good outcome. pc-CTA and pc-ASPECTS demonstrated intermediate performance, with pc-ASPECTS of 8 to 10 showing relatively high sensitivity but limited specificity, and pc-CTA score <3 providing modest sensitivity and specificity. When the analysis was restricted to patients achieving successful reperfusion, overall patterns of diagnostic performance were preserved, with some minor shifts in point estimates given a slightly higher prevalence of the primary outcome in this subgroup (ie, 56.5% versus 49.0%; Table 4).

Table 4.

Diagnostic Accuracy of Dichotomized Imaging Scores for Predicting Good 90-Day Outcome (mRS Score, 0–3) in the Entire Cohort and the Recanalized Cohort (Defined as TICI Score of 2b or Greater)

An ROC was generated using the data from the entire cohort (Figure 3). AUC values were the highest for the CAPS (AUC, 0.72 [95% CI, 0.63–82]) and the lowest for the pc-CTA score (AUC, 0.57 [95% CI, 0.45–0.68]; P=0.019). AUC for BATMAN (AUC, 0.66 [95% CI, 0.55–0.77]) and pc-ASPECTS scores (AUC, 0.63 [95% CI, 0.52–0.75]) were in between though this difference compared with the CAPS was not statistically significant (P=0.29 and P=0.23, respectively). There were only 12 patients with CAPS >3, and all 12 patients had a good functional outcome at 90 days.

Figure 3.

Receiver operating characteristic (ROC) curves with reported area under the curve values for good functional outcomes at 3 months across 4 imaging scores assessing extent of ischemic injury in patients with basilar artery occlusions. BATMAN indicates basilar artery treatment management; CAPS, critical area perfusion score; pc-ASPECTS, posterior circulation Alberta Stroke Program Early Computed Tomography Score; and pc-CTA, posterior circulation computed tomography angiography.

Discussion

We compared 4 different posterior circulation imaging scoring systems (pc-ASPECTS score, BATMAN score, pc-CTA score, and CAPS) using a variety of CT imaging techniques to assess their individual abilities to predict functional outcomes after mechanical thrombectomy in BAOs. Identifying the most accurate scoring system is important as these scores aim to assist physicians with selecting candidates for hyperacute interventions. CAPS ≤3 demonstrated perfect sensitivity and negative predictive value for good functional outcome but had low specificity, suggesting that it is a highly sensitive rule-out tool: the absence of CAPS ≤3 reliably identified patients unlikely to achieve a good outcome in our analysis, but its presence alone is insufficient to predict good functional recovery. CAPS also had the best overall performance according to the AUC from the ROC for predicting 90-day outcome, which was significantly better than the performance of the pc-CTA score but not the BATMAN score or pc-ASPECTS.

CAPS is the only imaging score that is purely CT perfusion–based. CTP produces color maps that are easier to read and interpret, making them less subject to interrater variability compared with noncontrast CT or collateral assessment using CTA.¹² However, selection of arterial input function and venous output function, which can be done manually or by automated software, can lead to variations in the quantitative perfusion values, which can compromise the accuracy of the perfusion results.⁹ Careful arterial input function selection and motion correction are essential to ensure accuracy and reproducibility of the perfusion imaging results.^13,14 CTP may provide an advantage over conventional noncontrast CT scans, in which the ability to judge ischemia in the posterior fossa is limited due to imaging artifacts.¹⁰ The pc-ASPECTS score aims to identify parenchymal hypodensities on CTA source images, but the influence of streak and bone artifact in the posterior fossa limits its accuracy, and the extent of hypodensity (ie, in Hounsfield units) that is associated with irreversible injury is unknown.¹¹ CTA-based scores are limited due to more variability related to the smaller caliber of vessels in the posterior circulation and high prevalence of normal anatomic variants. They are further challenging to interpret due to the additional presence of venous contamination, given the proximity of the venous draining system to the posterior cerebral arteries and posterior communicating arteries. The posterior circulation is unique from the anterior circulation in that it may require different thresholds for imaging criteria, for example, penumbra is defined as TMax >6 seconds in the anterior circulation, but we previously found that a definition of TMax >10 seconds, suggestive of severe hypoperfusion, was more appropriate in the posterior circulation.⁸ Further studies are needed to determine the severity of hypoattenuation or hypoperfusion in the posterior circulation as a determinant of permanent ischemic injury.

Determining which patient will do well with endovascular thrombectomy in BAO remains an important consideration when selecting patients to undergo intervention, given the high burden on the healthcare system due to high morbidity and mortality associated with the disease.¹² A previous post hoc analysis of patients with BAO treated with endovascular thrombectomy from the ATTENTION (Time-to-Maximum Endovascular Treatment for Acute Basilar-Artery Occlusion) and ATTENTION-IA (Intra-Arterial Tenecteplase After Successful Endovascular Recanalization in Patients With Acute Posterior Circulation Arterial Occlusion) trials found that patients selected by CT/CTA compared with CT/CTA/CTP had similar odds of achieving favorable clinical outcomes.¹⁵ However, the specific imaging parameters used to select patients were not reported. It is unclear, therefore, whether posterior circulation–specific parameters for CTP were used (ie, TMax >10 seconds rather than TMax>6 seconds, as previously defined as penumbra in the anterior circulation). Last, CAPS has not been tested in these cohorts to our knowledge, and we think that assigning a score ensures consistency and standardization across various raters/clinicians and may reduce interrater variability. Next, Yoon et al¹⁶ showed that the initial stroke severity as determined by the NIHSS score and the pc-ASPECTS on pretreatment diffusion-weighted imaging were significant independent predictors of good outcome.¹³ Unfortunately, magnetic resonance imaging is not readily available at most centers and may cause significant delays in the hyperacute stroke code situation.¹⁴ On the other hand, given that CT and CTP are more widely available, there are fewer barriers to the wide implementation of CTP-based imaging scores. One of our previous studies found that both CAPS >3 and pc-ASPECTS ≤6 are independent predictors of poor outcome; it is possible that the optimal approach to holistic prognostication may be using multimodal imaging techniques.¹⁵

Another study, based in China, aimed to develop a posterior circulation ischemic stroke outcome score, which is based on a combination of clinical and imaging factors.¹⁷ They identified the following factors as independent predictors of poor outcome in posterior circulation strokes: age, modified Rankin Scale score before admission, NIHSS score on admission, ischemic stroke history, infarction distribution, basilar artery, and posterior cerebral artery stenosis or occlusion. While accounting for the involved artery, this scoring system did not consider infarct size or volume, which are important imaging surrogate markers.¹⁷ All of the current imaging scores account for this factor by assigning more points to more regions with evidence of infarct or severe ischemia although quantification is more reliable using CTP-based imaging scores or magnetic resonance imaging. Given that the ischemic core volume is a known predictor for outcome, it may be critical for informing the extent of ischemic injury and, in turn, disability.¹⁸ Further studies are needed to determine whether the use of CAPS improves decision-making and patient outcomes. A National Institutes of Health–funded prospective multicenter study that aims to evaluate the use of CT perfusion, as well as other clinical and imaging factors (ie, CAPS) in their abilities to predict outcomes in BAO, is in progress.

Strengths and Limitations

One strength of our study is that it is a multicenter cohort resulting in a larger and more diverse sample population, thereby enhancing its external generalizability. We also utilized multirater assessments of the imaging scores to assess agreement on prognostic scores though the results are presented in a separate manuscript. A major limitation of our study is the retrospective design and small sample size. There were only 12 patients who had CAPS>3 in this cohort, which is likely underpowered to make definitive conclusions about its predictive abilities. Our sample size also limits the ability to perform adequately powered subgroup analyses. Potential for selection bias was minimized given the systematic and consecutive nature in which patients were identified from multiple centers although we did not have screening log data, so biases in patient selection based on imaging findings likely occurred. It is possible that the study population was skewed toward those with better imaging and outcomes and did not include enough subjects with poor outcomes to detect a meaningful difference. Given these limitations, it will be important to validate these findings in a larger patient group. The analyses performed are exploratory and not confirmatory. In addition, we did not have access to information such as the number of patients who needed suboccipital craniectomy and could not evaluate this clinically relevant outcome. Last, there are other imaging scores that were not assessed in this study, such as the Bern diffusion-weighted imaging or diffusion-weighted imaging Brainstem score.¹⁹ However, we chose the imaging scores in the current study because they are CT-based and, therefore, the most likely to be implemented into routine clinical use as well. Future studies could compare the CAPS to these other proposed scores, specifically magnetic resonance imaging–based imaging scores, which are considered to be the gold standard when evaluating early ischemic burden.

Conclusions

In this study, the previously proposed CAPS was compared with 3 other CT-based scores in their ability to predict patient outcome after endovascular thrombectomy in acute BAO. CAPS had the best performance for predicting good functional outcome at 3 months and had high sensitivity and negative predictive values though our finding is limited by the small number of patients who had CAPS >3. This score should be studied further in future prospective studies evaluating imaging markers in acute BAO to inform patient selection for hyperacute endovascular intervention.

ARTICLE INFORMATION

Sources of Funding

None

Disclosures

Dr Albers consults for and holds equity in iSchemaView. Dr Heit is a consultant and a Medical Advisory Board Member for iSchemaView. The other authors report no conflicts.

Supplemental Material

Table S1

STROBE Checklist