Accuracy and reliability of PBV ASPECTS,CBV ASPECTS and NCCT ASPECTS in acute ischaemic stroke: a matched-pair analysis

Abstract

Background and purpose

To investigate the reliability and accuracy of Alberta Stroke Program Early Computed Tomography Scores (ASPECTS) derived from flatpanel detector computed tomography pooled blood volume maps compared to non-contrast computed tomography and multidetector computed tomography perfusion cerebral blood volume maps.

Methods

ASPECTS from pooled blood volume maps were evaluated retrospectively by two experienced readers for 37 consecutive patients with acute middle cerebral artery (MCA) M1 occlusion who underwent flatpanel detector computed tomography perfusion imaging before mechanical thrombectomy between November 2016 and February 2019. For comparison with ASPECTS from non-contrast computed tomography and cerebral blood volume maps, a matched-pair analysis according to pre-stroke modified Rankin scale, age, stroke severity, site of occlusion, time from stroke onset to imaging and final modified thrombolysis in cerebral infarction (mTICI) was performed in a separate group of patients who underwent multimodal computed tomography prior to mechanical thrombectomy between June 2015 and February 2019. Follow-up ASPECTS were derived from either non-contrast computed tomography or from magnetic resonance imaging (in seven patients) one day after mechanical thrombectomy.

Results

Interrater agreement was best for non-contrast computed tomography ASPECTS (w-kappa = 0.74, vs. w-kappa = 0.63 for cerebral blood volume ASPECTS and w-kappa = 0.53 for pooled blood volume ASPECTS). Also, accuracy, defined as correlation between acute and follow-up ASPECTS, was best for non-contrast computed tomography ASPECTS (Spearman ρ = 0.86 (0.65–0.97), P < 0.001), while it was lower and comparable for pooled blood volume ASPECTS (ρ = 0.58 (0.32–0.79), P < 0.001) and cerebral blood volume ASPECTS (ρ = 0.52 (0.17–0.80), P = 0.001). It was noteworthy that cases of relevant infarct overestimation by two or more ASPECTS regions (compared to follow-up imaging) were observed for both acute pooled blood volume and cerebral blood volume ASPECTS but occurred more often for acute pooled blood volume ASPECTS (25% vs. 5%, P = 0.02).

Conclusion

Non-contrast computed tomography ASPECTS outperformed both pooled blood volume ASPECTS and cerebral blood volume ASPECTS in accuracy and reliability. Importantly, relevant infarct overestimation was observed more often in pooled blood volume ASPECTS than cerebral blood volume ASPECTS, limiting its present clinical applicability for acute stroke imaging.

Keywords

Stroke FD-CT perfusion dyna-CT

Introduction

Mechanical thrombectomy (MT) is the treatment of choice in patients with acute ischaemic stroke due to large vessel occlusion (LVO).¹ As high recanalisation rates of 80% and higher are achieved with the use of modern aspiration catheters and stent retrievers,² time from stroke onset to recanalisation remains one of the most important factors determining patient outcome in acute stroke therapy.³ C-arm-mounted flatpanel detector (FD) computed tomography (CT) imaging offers the possibility to rule out intracranial haemorrhage and prove LVO^4,5 immediately in the angiographic suite. Thereby, depending on the clinical setting, the time from hospital admission to the start of the interventional procedure can be shortened.^6,7

Still, to select patients who potentially benefit from MT, a reliable estimation of the infarct core by acute stroke imaging is crucial for further treatment decisions. To address this question within the setting of C-arm FD-CT imaging in the angiographic suite, modern angiographic systems allow for generating whole-brain pooled blood volume (PBV) maps. Previous studies have found good correlation of such FD-CT-derived PBV maps with multidetector (MD) CT-derived cerebral blood volume (CBV) maps^4,5,8,9 and PBV reductions were found to correlate well with final infarct volume before.⁶

While FD-CT perfusion represents a promising tool for acute stroke imaging, the validity of infarct core assessment by FD-CT PBV maps remains a topic of debate and critical overestimations of ischaemic cores were reported previously.¹⁰ In this retrospective study we therefore investigated: (a) the reliability and (b) the accuracy of Alberta Stroke Program early computed tomography scores (ASPECTS) derived from FD-CT perfusion parenchymal blood volume (PBV) maps compared to non-contrast computed tomography (NCCT) and CT perfusion CBV maps in predicting final infarct size and clinical outcome 3 months after stroke within the setting of a matched-pair analysis.

Materials and methods

Patient data

This retrospective study was approved by the institutional review board. We included data from 54 consecutive patients with acute occlusion of the middle cerebral artery (MCA) M1 segment who underwent FD-CT perfusion imaging prior to MT between November 2016 and February 2019 at our institution. Data from 17/54 patients (31%) had to be excluded from further analysis due to severe artefacts from motion and beam hardening (n = 12), missing contrast agent filling (n = 4) and/or incomplete acquisition of FD-CT perfusion (n = 1). To compare PBV ASPECTS with NCCT and CBV ASPECTS, a matched-pair analysis was conducted with patients who underwent multimodal MD-CT imaging prior to MT between June 2015 and February 2019 at our institution (n = 186). Patients were matched according to pre-stroke modified Rankin scale (mRS), age, stroke severity (measured by the National Institutes of Health stroke scale; NIHSS), site of occlusion, time from stroke onset (or time from last seen well) to imaging and final modified thrombolysis in cerebral infarction (mTICI). Favourable clinical outcome 3 months after stroke was defined as mRS of 3 or less; mRS was evaluated by an independent neurologist blinded to this study and was available in 34/37 patients in the FD-CT cohort and in 37/37 patients in the MD-CT study cohort. In two patients (one per cohort) premorbid mRS was higher than 3 and these patients were not considered in the final outcome analysis.

Imaging

MD-CT imaging

Multimodal MD-CT imaging was carried out on a 64-slice CT scanner (Somatom Definition AS; Siemens Healthineers, Erlangen, Germany) and consisted of NCCT, CT angiography and CT perfusion. NCCTs were acquired at 120 kV with automated adjustment of the tube current in the Xcare technique (Siemens, Erlangen, Germany). Images were reconstructed with a kernel of J40s at 4 mm slice thickness. CT perfusion imaging was obtained at 180 mAs, 80 kV, total acquisition time of 60 s with a z-coverage of 8 cm. In total, 480 slices were reconstructed (30 full temporal volumes with each 16 slices at a slice thickness of 5 mm) using a kernel of H20f. Acquisition was started 10 seconds after intravenous administration of iodine contrast agent (Xenetix 350; Guerbet, Sulzbach, Germany) at 6 ml/s followed by a 20 ml saline flush. CBV maps were derived from the four-dimensional (4D) dynamic dataset using Syngo volume perfusion CT Neuro on a Syngo CT Workplace, version VA10A (Siemens Healthineers, Erlangen, Germany).

FD imaging

FD-CT imaging was performed on a biplane FD angiographic system (Axiom-Artis Q; Siemens Healthineers, Erlangen, Germany). As described previously,⁵ the limited speed of the gantry movement compared to conventional MD-CT does not allow dynamic measurements of the contrast bolus passage. Instead image acquisition consisted of two separate 200° rotational runs with a first, native, initial rotation (mask run) followed by a second, contrast-enhanced rotation (fill run). To ensure that the fill run is acquired during a steady state phase of the contrast medium in the brain parenchyma, the C-arm returns to the (original) start position after acquisition of the mask run and standard two dimensional (2D) digital subtraction angiography (DSA) acquisitions are initiated at a rate of two images per second. The fill run is then started manually when opacification of the venous sinus is observed by the operator (‘bolus watching’ approach). The acquisition time of the mask and fill run was 6.6 seconds whereby 2D projections were acquired at a rate of 60 frames per second (leading to a total 397 projections per run, 0.5° rotation per frame) at 70 kV; 616 × 480 matrix size; projection on a 30 × 40 cm flat panel size. Post-processing of FD perfusion was carried out using Syngo DynaPBV Neuro on a Syngo X Workplace, version VD11C (Siemens Healthineers, Erlangen, Germany). First, the mask run and fill run were reconstructed at 4 mm slice thickness using both ‘smooth’ and ‘very smooth’ image reconstruction characteristics. As previously described, images are then subtracted from another and air and bone are excluded from further analysis by automated segmentation. By histogram analysis of the vessel tree, a steady state arterial input value was calculated and applied as a scaling factor to the image volume. Finally, a filter was applied to reduce pixel noise.⁵ Overall, post-processing time for FD-CT perfusion data was sufficiently fast and comparable to MD-CT.

Follow-up imaging

Follow-up imaging was acquired 21 (15–25) hours after MT by NCCT (n = 30) and magnetic resonance imaging (MRI) (n = 7) at a 3T Scanner (Magnetom Verio/TIM Trio/Prisma Fit/Skyra; Siemens Healthineers, Erlangen, Germany), respectively, including T2/fluid-attenuated inversion recovery (FLAIR), diffusion-weighted and susceptibility-weighted imaging.

ASPECT scoring

As previously described,¹⁰ quantitative analysis of PBV maps was not feasible due to the inhomogeneity of PBV maps and large standard deviation (SD) on a voxel level, respectively. The extent of ischaemic cores was therefore assessed qualitatively by ASPECTS and was evaluated separately on FD-CT PBV maps, on MD-CT NCCT and CBV maps, and on the follow-up imaging by two experienced readers (AP, JP, both with more than 5 years of experience in acute stroke imaging) blinded to clinical and other imaging data. For PBV and CBV maps, areas of decreased PBV or CBV (indicated dark blue to purple) were thereby regarded as ischaemic core. In the case of disagreement, consensus rating was reached. Besides, source images were further checked, whether FD-CT perfusion was acquired correctly during the venous phase by noting opacification of the intracranial venous sinus before the start of the acquisition of the fill run.

Statistical analysis

Statistical analysis was performed with R* (the R Project for Statistical Computing, V3.1.2). Group differences were assessed with Fisher’s exact t-test and Pearson’s chi-squared test (for count data) and Welch’s t-test (for normally distributed variables) or the Mann–Whitney U test (for not normally distributed variables). Interrater reliability was assessed with weighted kappa coefficients. Correlation of acute and follow-up ASPECTS was assessed with Spearman correlation and linear regression analysis. Univariate logistic regressions were carried out for neurological outcome at 3 months. For all P values the significance level was set to α = 0.05. Means are given with their SD, medians with their interquartile range (IQR), all confidence intervals (CIs) are quoted as 95% CI.

Results

Baseline patient characteristics

Baseline patient characteristics for both the FD and the MD-CT cohorts are listed in Table 1. There were no differences in age, prestroke mRS, baseline NIHSS, time from symptom onset or, in cases of unknown onset time (n = 16 for both cohorts), time from last seen well to admission and rates of successful recanalisation (86% in the FD-CT cohort vs. 94% in the MD-CT cohort, P = 0.43). There were further no significant differences for the side of MCA M1 occlusion (P = 1.0) or the administration of intravenous thrombolysis (P = 0.82).

Table 1.

Baseline patient characteristics (if applicable median and interquartile range are given).

Age	79 (70–84)	77 (70–84)	0.75
Gender (female/male)	19/18	25/12	0.24
Pre-stroke mRS	1 (0–2)	1 (0–2)	0.73
Baseline NIHSS	17 (9–21)	14 (11–20)	0.36
Side of occlusion (left/right)	17/20	16/21	1.0
Prior external imaging and secondary patient transfer to the stroke center	19 (51%)	29 (78%)	0.02
Time from symptom onset to admission (min)	270 (161–503)	245 (152–581)	0.81
Time from admission to imaging (min)	17 (14–31)	20 (15–28)	0.75
Time from admission to groin puncture (min)	41 (31–47)	67 (54–90)	<0.001
Intravenous lysis (y)	19 (51%)	17 (46%)	0.82
Successful recanalisation^a (y)	32 (86%)	35 (94%)	0.43

^aRecanalisation was defined to be successful when mTICI was 2b or better.

mRS: modified Rankin scale; NIHSS: National Institutes of Health stroke scale.

It was noteworthy that while time from initial hospital admission at our site to imaging was comparable for both cohorts (P = 0.75), time from admission to groin puncture was shorter in patients who directly underwent FD-CT imaging (41 (31–47) minutes) compared to conventional MD-CT imaging (67 (54–90) minutes, P < 0.001). Further on, this significant correlation holds true (P = 0.002) when considering only the subgroup of patients who did not undergo prior imaging at other hospitals and secondary transfer to our institution (18 (49%) in the MD-CT cohort vs. eight (22%) in the FD-CT cohort, P = 0.02), a scenario which could have affected processing time at our institute.

Interrater reliability

Interrater reliability was best for NCCT ASPECTS with κ = 0.74 (0.59–0.85), while it was substantial for CBV ASPECTS with κ = 0.63 (0.47–0.74). For PBV ASPECTS, interrater reliability was dependent on the image reconstruction characteristics with κ = 0.53 (0.38–0.65) for ‘smooth’ image reconstruction characteristics and κ = 0.61 (0.49–0.71) for ‘very smooth’ image reconstruction characteristics (see Table 2). Interrater reliability did not improve when considering only the FD perfusions, which were acquired within the venous phase (n = 32/37 (86%)) with κ = 0.56 (0.41–0.67) for ‘smooth’ image characteristics and κ = 0.60 (0.48–0.70) for ‘very smooth’ image characteristics.

Table 2.

Interrater reliability of ASPECTS (weighted κ coefficients) and correlation of acute to follow-up ASPECTS in dependence of the modality (and for FD-CT PBV map-derived ASPECTS in dependence of the image reconstruction characteristics ‘smooth’ or ‘very smooth’).

	Interrater reliability	Correlation to follow-up ASPECTS
	Interrater reliability	Spearman correlation		Linear regression
Modality	κ (CI)	ρ (CI)	P value	r²
NCCT	0.74 (0.59–0.85)	0.80 (0.61–0.94)	<0.001	0.63
CBV	0.63 (0.47–0.74)	0.52 (0.17–0.80)	<0.001	0.28
PBV (‘smooth’)	0.53 (0.39–0.64)	0.58 (0.32–0.79)	0.001	0.33
PBV (‘very smooth’)	0.61 (0.49–0.71)	0.63 (0.40–0.83)	<0.001	0.42

ASPECTS: Alberta Stroke Program early computed tomography scores; CBV: cerebral blood volume; CI: confidence interval; FD-CT: flatpanel detector computed tomography; NCCT: non-contrast computed tomography; PBV: pooled blood volume.

Correlation with follow-up ASPECTS

To assess the accuracy of ASPECTS derived from NCCT and from CBV and PBV maps to predict final infarct size, consensus scores were compared to follow-up ASPECTS 24 hours after MT. First, on a group level, a tendency towards lower ASPECTS on follow-up imaging was noted in the FD cohort compared to the MD cohort (median (IQR) of 6 (3–8) vs. 8 (5–8), P = 0.05).

With ρ = 0.80 (0.61–0.94), P < 0.001 in the Spearman analysis and R² = 0.63 in the linear regression (see Table 2), NCCT ASPECTS was found to correlate best with follow-up ASPECTS. Still, a tendency to underestimate the final infarct extent by ASPECTS derived from acute NCCT (P = 0.05, intercept of the linear regression –2.9 ± 1.3, see Figure 1) was noted.

Figure 1.

Scatter plots for acute Alberta Stroke Program early computed tomography scores (ASPECTS) versus follow-up ASPECTS for (a) pooled blood volume (PBV)-derived acute ASPECTS (black circles: for ‘smooth’ image characteristics, red triangles: for ‘very smooth’ image characteristics); (b) cerebral blood volume (CBV)-derived acute ASPECTS; and (c) non-contrast computed tomography (NCCT)-derived acute ASPECTS. Linear regression analysis showed best correlation between acute NCCT ASPECTS and follow-up ASPECTS (R²=0.63) compared to CBV-derived acute ASPECTS (R²=0.28) and PBV-derived acute ASPECTS (R²=0.33 (‘smooth’ image characteristics) and R²=0.42 (‘very smooth’ image characteristics)). Moreover, linear regression reveals a systematic overestimation of the final infarct extent on PBV maps (intercept (standard deviation; SD): 2.1 ± 0.9 for ‘smooth’ image characteristics and 1.7 ± 0.8 for ‘very smooth’ image characteristics) compared to CBV map-derived ASPECTS (intercept (SD): 0.9 ± 1.6), while a tendency to underestimate final infarct extent is noticed for acute NCCT ASPECTS (intercept (SD): –2.9 ± 1.3).

In the Spearman correlation analysis, ASPECTS derived from CBV maps and from PBV maps were found to perform comparably with ρ = 0.52 (0.17–0.80), P = 0.001 for CBV maps-derived ASPECTS and ρ = 0.58 (0.32–0.79), P < 0.001 for PBV maps-derived ASPECTS (for ‘smooth image’ characteristics). The accuracy of PBV ASPECTS improved further when derived from PBV maps with ‘very smooth’ image characteristics (ρ = 0.63 (0.40–0.83), P < 0.001). However, linear regression revealed a systematic error of PBV-derived ASPECTS with significant overestimation of the final infarct extent (intercept of the linear regression: 2.1 ± 0.9 (‘smooth’ image characteristics) and 1.7 ± 0.8 (‘very smooth’ image characteristics) vs. 0.9 ± 1.6 for CBV-derived ASPECTS, see Figure 1). In particular, we observed relevant infarct overestimation by at least two ASPECTS regions in 10 of 37 patients (27%) for PBV maps-derived ASPECTS (see exemplarily Figure 2). Importantly, such potentially misleading, and hence relevant, overestimation of final infarct extent was observed significantly less often in CBV map-derived ASPECTS (n = 2/37 (5%), P = 0.02). It was noteworthy that such relevant overestimations were not observed for NCCT-derived ASPECTS. For FD-CT PBV-derived ASPECTS, correlation did not improve, when considering only PBV maps derived from perfusion data or, respectively, fill runs which were acquired in the venous (steady state) phase: ‘smooth’ image characteristics: ρ = 0.54 (0.25–0.80), P = 0.001; ‘very smooth’ image characteristics: 0.62 (0.35–0.87), P < 0.001) and, in particular, relevant infarct overestimation was still observed in 25% of these cases (eight of 32).

Figure 2.

Examples for relevant infarct overestimation (defined as overestimation by two or more Alberta Stroke Program early computed tomography scores (ASPECTS) regions when compared to follow-up imaging) for flatpanel detector computed tomography (FD-CT)-derived pooled blood volume (PBV) ASPECTS (a) and multidetector computed tomography (MD-CT)-derived cerebral blood volume (CBV) ASPECTS (c) when compared to the follow-up non-contrast computed tomography (NCCT) one day after mechanical thrombectomy (MT) (b) and (d).

Correlation between acute and follow-up ASPECTS was not found to depend on prior contrast agent administration. So, Spearman correlation between acute and follow-up imaging was ρ = 0.61 (0.29–0.84), P < 0.001 for PBV ASPECTS (for ‘smooth image’ characteristics), ρ = 0.50 (0.0–0.9), P = 0.03 for CBV ASPECTS and ρ = 0.77 (0.34–1.0), P < 0.001 for NCCT ASPECTS for the subgroups of patients who underwent prior external imaging and secondary transfer to our hospital.

Clinical outcome analysis

A favourable clinical outcome was observed in 48% (16/33) of patients who underwent FD-CT imaging and in 58% (21/36) of patients who underwent conventional NCCT and CT perfusion prior to MT (P = 0.42). In the logistic regression, NCCT-derived ASPECTS (odds ratio (OR) 2.30 (1.40–4.72), P = 0.007) and PBV-derived ASPECTS (‘smooth’ image characteristics: OR 1.48 (1.31–2.08), P = 0.010; ‘very smooth’ image characteristics: OR 1.55 (1.17–2.23), P = 0.006) were predictors of a favourable clinical outcome, while CBV-derived ASPECTS missed significance in our cohort (OR 1.29 (0.90–1.95), P = 0.19). Besides, NIHSS on admission was a predictor of favourable clinical outcome (OR 0.89 (0.81–0.96), P = 0.008). While time from admission to groin puncture was shorter for patients who underwent FD imaging compared to MD-CT (see above), neither time from symptom onset to admission, nor time from symptom onset to groin puncture (both P = 0.59) were predictors of clinical outcome. There was a tendency for successful MT as a predictor of favourable clinical outcome in the univariate analysis (OR 8.31 (1.31–161), P = 0.06) in our study.

Discussion

While FD-CT imaging has been proved to allow for the exclusion of intracranial haemorrhage and diagnosis of LVO in acute ischaemic stroke before,^4,5 we were able to demonstrate in this study that ASPECTS derived from FD-CT PBV maps predicts clinical outcome 3 months after MT and thereby even outperformed MD-CT CBV-derived ASPECTS on a group level.

Nevertheless, NCCT-derived ASPECTS still outperformed both PBV and CBV-derived ASPECTS in the prediction of clinical outcome and revealed superior reliability and accuracy. In particular, the occurrence of relevant infarct overestimation on an individual patient level could prevent FD-CT PBV maps from wide clinical application in acute stroke care. Here, relevant overestimation of final infarct extent (by at least two ASPECTS regions) was present in 27% for PBV-derived ASPECTS, which is in line with a previous study by Ava et al. reporting relevant infarct overestimation on FD-CT-derived PBV maps in 25%.¹⁰

Likewise, even though to a lesser extent, infarct overestimation occurred on MD-CT-derived CBV maps as well, while it was not observed for NCCT-derived ASPECTS. This finding is corroborated by previous studies,^11–14 in which the occurrence of infarct overestimation was already noted for the, nevertheless widely established, MD-CT perfusion mismatch assessment.¹⁵ As the factors contributing to MD-CT CBF and CBV reduction in cerebral ischaemia are complex,¹¹ a variety of reasons for the misclassification of infarct core by CT perfusion have been discussed. The time dependency of infarct growth leading to different infarct volumes in cases of early reperfusion can be decisive,¹³ as well as states of so-called misery perfusion, in which an elevated oxygen extraction fraction may compensate for reductions in CBF and CBV.¹⁰ Variations in vascular anatomy and auto-dysregulation of perfusion pressure, upstream flow restriction and, finally, technical reasons including misplacement and motion of the patient during the CT perfusion scan can play a role as well.¹²

Previous studies have reported good correlation of FD-CT PBV maps with MD-CT-derived CBV maps in animal studies and in humans^5,8 and correlation with final infarct volume has been demonstrated before.¹⁶ Remarkably, relevant infarct overestimation was more frequent for PBV-derived ASPECTS than for CBV-derived ASPECTS in our study. Image quality of FD-CT perfusion may be an important factor that contributes to the disadvantage of FD-CT, limiting accuracy of perfusion-derived infarct assessment due to the low signal to noise ratio and high SD.^11,17 Moreover, patient motion may have affected PBV map quality in our study. We thereby emphasise the finding that in our study 12 of 54 patients (22%) who underwent FD-CT perfusion imaging had to be excluded from the analysis due to strong beam hardening and motion artefacts on PBV maps. Besides infarct overestimation, this presents a second major drawback of the method which could potentially limit the applicability of FD-CT perfusion imaging at the current state of the technology. Stroke patients are often agitated and, in contrast to the above-mentioned studies,^5,8,16 imaging was not carried out under general anaesthesia in our study. Nevertheless, we point out that motion artefacts alone do not explain the above discussed occurrence of infarct overestimation. We remind readers that interrater reliability could substantially be improved by changing the image post-processing algorithm from ‘smooth’ to ‘very smooth’ for the evaluable PBV maps, and hence observer-dependent visual delineation of PBV reduction alone does not account for the observed differences. Rather, interactions between CBV and CBF may contribute to the overestimation of PBV-derived ASPECTS. Kamran and Byrne found that a reduction in PBV does not directly translate to a reduction in CBV, but instead both CBV weighting (∼40%) and CBF weighting (∼60%) contribute to the PBV parameter.¹⁸ While oligemic tissue is now described to exhibit increased CBV at mild reductions in CBF (resulting ideally in constant PBV values), penumbral tissue may already exhibit stronger CBF reductions (still above the typical thresholds for ischaemic core) combined with beginning decreases in CBV.^19,20 Depending on the contribution of these two parameters to PBV, their interaction may already lead to visually severe reductions within penumbral tissue on PBV maps. The interplay of these two parameters may hence impede visual differentiation between infarct core and penumbral tissue, possibly explaining the observed tendency to overestimate the infarct core on PBV maps even compared to CBV maps. As an additional complicating factor, the contribution of CBV and CBF to PBV depends on the acquisition parameters, as namely contrast wash-in or wash-out during the acquisition of the fill run results in variable CBF weighting.¹⁸ Even the ‘bolus-watching’ approach to ensure acquisition during a steady state phase may overlook the differences in individual patient haemodynamics,¹⁸ and we observed relevant infarct overestimation also in the subgroup of patients fulfilling the steady state condition (opacification of the venous sinus). These findings imply that, although noise levels on PBV maps may be reduced in the future, the interplay of CBV and CBF could nevertheless limit the applicability of PBV maps to identify mismatch reliably. We point out that this limitation only applies to the visual delineation of infarcted tissue on PBV maps and not to threshold-based approaches, which may be realised when substantial improvements in image quality and signal-to-noise ratio are achieved. So far, the additional information gained from FD-CT perfusion imaging may not allow for precise treatment decision making in acute ischaemic stroke. It is noteworthy that as a side effect of our matched-pair analysis, we found a reduction in time from initial admission to groin puncture in patients who underwent FD-CT imaging prior to MT compared to patients who underwent primarily conventional MD-CT imaging. Although the study design does not allow for the generalisation of this finding, it is corroborated by previous studies⁷ and still motivates for further research on the technique to identify ischaemic tissue in acute ischaemic stroke by FD-CT imaging.

Finally, we point out that there is a potential bias towards better correlation between NCCT ASPECTS (compared to PBV or CBV ASPECTS) and final infarct extent in our study, as post-treatment ASPECTS was evaluated also on NCCT and not gold standard MRI DWI. Further limitations of our study result from the monocentre, retrospective study design and the potential for selection bias in the groups for the matched-pair analysis. More patients underwent prior external imaging in the MD group compared to the FD group. Thereby, contrast agent extravasation due to blood–brain barrier disruption could mask early ischaemic changes on NCCT. Still, we point out that such an effect would not influence the study results, as already NCCT ASPECTS was found to outperform PBV and CBV ASPECTS. Moreover, compared to infarct overestimation, our study is not designed to investigate the reasons for infarct underestimation. In cases in which infarcts were underestimated compared to follow-up imaging, our data do not allow us to deduce whether final infarct extent was already evident in the acute setting (‘true underestimation’) or developed within the time from imaging to recanalisation and follow-up imaging. Furthermore, the lack of mRS prediction by CBV ASPECTS in our study may be due to the small sample size and the relatively homogenous patient cohort. The cohort size of, in total, 74 patients asks for further validation of our results.

Conclusion

NCCT ASPECTS prior to MT outperformed both FD-CT PBV ASPECTS and MD-CT CBV ASPECTS in accuracy and reliability. Moreover, we observed relevant overestimations of acute infarct core size for both CBV ASPECTS and PBV ASPECTS. Importantly, visual overestimation of the infarct core occurred thereby more often for PBV ASPECTS than for CBV ASPECTS, potentially limiting the applicability of FD-CT PBV imaging for clinical routine and standard stroke protocols.

Footnotes

Conflict of interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Arne Potreck

Johannes AR Pfaff

References

Goyal

Menon

van Zwam

, et al. Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials. Lancet 2016; 387: 1723–1731. DOI: 10.1016/S0140-6736(16)00163-X

Pfaff

Rohde

Engelhorn

, et al. Mechanical thrombectomy using the new Solitaire™ platinum stent-retriever: reperfusion results, complication rates and early neurological outcome. Clin Neuroradiol 2019; 29: 311–319. DOI: 10.1007/s00062-017-0657-x

Saver

Goyal

van der Lugt

, et al. Time to treatment with endovascular thrombectomy and outcomes from ischemic stroke: a meta-analysis. JAMA 2016; 316: 1279–1288. DOI: 10.1001/jama.2016.13647

Struffert

Richter

Engelhorn

, et al. Visualisation of intracerebral haemorrhage with flat-detector CT compared to multislice CT: results in 44 cases. Eur Radiol 2009; 19: 619–625. DOI: 10.1007/s00330-008-1183-7

Struffert

Deuerling-Zheng

Kloska

, et al. Flat detector CT in the evaluation of brain parenchyma, intracranial vasculature, and cerebral blood volume: a pilot study in patients with acute symptoms of cerebral ischemia. AJNR Am J Neuroradiol 2010; 31: 1462–1469. DOI: 10.3174/ajnr.A2083

Wagner

Kyriakou

du Mesnil de Rochemont

, et al. Does preinterventional flat-panel computer tomography pooled blood volume mapping predict final infarct volume after mechanical thrombectomy in acute cerebral artery occlusion? Cardiovasc Intervent Radiol 2013; 36: 1132–1138. DOI: 10.1007/s00270-013-0574-6

Pfaff

JAR

Schönenberger

Herweh

, et al. Direct transfer to angio-suite versus computed tomography-transit in patients receiving mechanical thrombectomy: a randomized trial. Stroke. Epub ahead of print 10 August 2020. DOI: 10.1161/STROKEAHA.120.029905

Bley

Strother

Pulfer

, et al. C-arm CT measurement of cerebral blood volume in ischemic stroke: an experimental study in canines. AJNR Am J Neuroradiol 2010; 31: 536–540. DOI: 10.3174/ajnr.A1851

Struffert

Deuerling-Zheng

Kloska

, et al. Cerebral blood volume imaging by flat detector computed tomography in comparison to conventional multislice perfusion CT. Eur Radiol 2011; 21: 882–889. DOI: 10.1007/s00330-010-1957-6

10.

Ava

Berkefeld

Lauer

, et al. Predictive value of pooled cerebral blood volume mapping for final infarct volume in patients with major artery occlusions. A retrospective analysis. Clin Neuroradiol 2017; 27: 435–442. DOI: 10.1007/s00062-017-0569-9

11.

Schaefer

Souza

Kamalian

, et al. Limited reliability of computed tomographic perfusion acute infarct volume measurements compared with diffusion-weighted imaging in anterior circulation stroke. Stroke 2015; 46: 419–424. DOI: 10.1161/STROKEAHA.114.007117

12.

Geuskens

Borst

Lucas

, et al. Characteristics of misclassified CT perfusion ischemic core in patients with acute ischemic stroke. PLoS One 2015; 10): e0141571. Published 2015 Nov 4. DOI: 10.1371/journal.pone.0141571

13.

Etherton

Barreto

Schwamm

, et al. Neuroimaging paradigms to identify patients for reperfusion therapy in stroke of unknown onset. Front Neurol 2018; 9: 327. Published 2018 May 15. DOI: 10.3389/fneur.2018.00327

14.

Schnieder

Psychogios

Maier

, et al. The problem of strict image-based inclusion criteria for mechanical thrombectomy – an analysis of stroke patients with an initial low CBV-ASPECTS score. Neuroradiol J 2019; 32: 287–293. DOI: 10.1177/1971400918791700

15.

Albers

Marks

Kemp

, et al. Thrombectomy for stroke at 6 to 16 hours with selection by perfusion imaging. N Engl J Med 2018; 378: 708–718. DOI: 10.1056/NEJMoa1713973

16.

Struffert

Deuerling-Zheng

Engelhorn

, et al. Feasibility of cerebral blood volume mapping by flat panel detector CT in the angiography suite: first experience in patients with acute middle cerebral artery occlusions. AJNR Am J Neuroradiol 2012; 33: 618–625. DOI: 10.3174/ajnr.A2839

17.

González

RG.

Low signal, high noise and large uncertainty make CT perfusion unsuitable for acute ischemic stroke patient selection for endovascular therapy. J Neurointerv Surg 2012; 4: 242–245. DOI: 10.1136/neurintsurg-2012-010404

18.

Kamran

Byrne

JV.

C-Arm flat detector CT parenchymal blood volume thresholds for identification of infarcted parenchyma in the neurointerventional suite. AJNR Am J Neuroradiol 2015; 36: 1748–1755. DOI: 10.3174/ajnr.A4339

19.

Muir

Buchan

von Kummer

, et al. Imaging of acute stroke. Lancet Neurol 2006; 5: 755–768. DOI: 10.1016/S1474-4422(06)70545-2

20.

Di Giuliano

Picchi

Sallustio

, et al. Accuracy of advanced CT imaging in prediction of functional outcome after endovascular treatment in patients with large-vessel occlusion. Neuroradiol J 2019; 32: 62–70. DOI: 10.1177/1971400918805710

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.68 MB