Flat-panel Detector Perfusion Imaging and Conventional Multidetector Perfusion Imaging in Patients with Acute Ischemic Stroke

In recent decades, flat-panel detector technology has become integral to (neuro)interventional angiography suites [1‐5]. Applications of flat-panel detector computed tomography (FDCT) include initial diagnostic imaging in the one-stop management workflow [6, 7], assessment of acute periprocedural complications in the angiography suite [3, 4] and detection of residual occlusions and their anatomic location during and shortly after neurovascular procedures. Recently introduced multiphase FDCT perfusion (FDCTP) acquisition protocols are capable of providing time-resolved whole brain dynamic perfusion imaging, including cerebral blood flow (CBF) and time to maximum (Tmax) maps [5, 8]. In acute ischemic stroke (AIS), perfusion imaging aids in the detection of occluded vessels (especially in peripheral occlusions) and influences therapeutic decision making [9‐11]. Perfusion maps help to visualize the volume of critically hypoperfused, irreversibly lost tissue regardless of subsequent reperfusion status (“infarct core”) and the hypoperfused, potentially salvageable ischemic penumbra (“tissue at risk”) [11‐15].

Unlike conventional multidetector computed tomography (MDCT), FDCT can be acquired directly in the angiography suite. Thus, it provides additional and potentially crucial information for pre-interventional and peri-interventional decision making [1, 2, 16] and increases the likelihood of improved clinical outcomes [2, 16]. The MDCT may be considered the reference standard but given the increasing availability and use of FDCT, it is essential to establish comparability between imaging results obtained with the two techniques. The differences in acquisition techniques, which are primarily related to temporal resolution, such as volume acquisition time and time between consecutive acquisitions should be taken into account [5, 17]. This comparability is crucial to provide a reliable basis for optimal therapeutic decisions.

However, to date, studies evaluating the comparability of new-generation FDCTP and conventional MDCTP are still scarce [5, 17, 18]. Therefore, the aim of this study was to assess the comparability of new-generation FDCTP and conventional MDCTP.

This study adhered to the principles outlined in the Declaration of Helsinki and received approval by the local ethics committee. The requirement for active informed patient consent was waived due to the retrospective nature of this study.

Consecutive patients with AIS, treated between June 2019 and March 2021 were selected from the neuroradiologic database. Patients who fulfilled the following additional criteria were included in this study: (a) they had undergone mechanical thrombectomy for anterior circulation stroke, (b) received a baseline conventional MDCTP and a pre-interventional FDCTP of diagnostic quality in the angiography suite, (c) had no change of occlusion between baseline CT angiography and the first digital subtraction angiography run (e.g., M1 occlusion on both images) and (d) a time interval < 120 min between the conventional MDCTP and the FDCTP. Of the 19 patients identified, 16 were “mothership” patients, and 3 “drip and ship” patients. Clinical data were extracted from the institutional stroke database or from the clinical information system.

Of 32 patients for whom pre-interventional standard cross-sectional perfusion imaging and pre-interventional FDCTP imaging were available, 8 were excluded because they underwent magnetic resonance perfusion imaging as baseline imaging and 5 because of insufficient imaging quality, leaving a final cohort of 19 patients (Supplementary Fig. 1). Relevant baseline characteristics are shown in Table 1.

The overall agreement for the volumes between the raters and different imaging techniques was assessed using a mixed effects model. The calculated agreement ranged from 0.88 (95% CI 0.80–0.95) to 1.0 (95% CI 0.99–1.0), indicating reliable ratings.

A strong correlation and robust agreement was observed between the volume of hypoperfused tissue as determined by the manually segmented maximum visible extent on MDCTP Tmax and FDCTP Tmax (r = 0.85, 95% CI 0.65–0.94, p < 0.001; ICC = 0.85, 95% CI 0.69–0.94; Figs. 1 and 2, Table 2). Similar results were obtained when comparing the volume of hypoperfused tissue on MDCTP TTP with FDCTP TTP (r = 0.91, 95% CI 0.78–0.97, p < 0.001; ICC = 0.90, 95% CI 0.78–0.96; Figs. 1 and 2, Table 2). Furthermore, in both cases, a significant proportion of the variance was shared by the two variables (MDCTP visible Tmax versus FDCTP visible Tmax: r² = 0.73; MDCTP visible TTP versus FDCTP visible TTP: r² = 0.83). The comparison between the results of the other variables showed lower correlations, weaker agreements, and less shared variance (Table 2).

In general, the variations in perfusion volumes between MDCTP and FDCTP remained consistent regardless of the time interval between scans (Tmax: r = 0.02, 95% CI −0.44–0.47, p = 0.93; TTP: r = 0.28, 95% CI −0.20–0.65, p = 0.25; Supplementary Fig. 2) or patient age (Tmax: r = 0.18, 95% CI −0.29–0.59, p = 0.45; TTP: r = 0.3, 95% CI −0.17–0.67, p = 0.21).

Of the 19 patients 7 received intravenous thrombolysis (36.8%). For the Tmax maps, the variation in perfusion volumes determined by the manually segmented maximum visible extent on MDCTP and FDCTP did not differ between patients who received and those who did not receive intravenous thrombolysis (difference MDCTP-FDCTP: median 3.3 ml, IQR −12.9–23.8 ml, and median −7.1 ml, IQR −24.2–25.7 ml for with and without intravenous alteplase, p = 0.71). For TTP maps, a difference was observed between patients who received intravenous thrombolysis and those who did not (difference MDCTP-FDCTP: median: 8.0 ml, IQR −3.7–21.9 ml, and median −27.0 ml, IQR −34.4–−2.3 ml for with and without intravenous alteplase, p = 0.03); however, these volume differences determined by the manually segmented maximum visible extent between MDCTP and FDCTP, remained consistent regardless of the time interval between the initiation of intravenous thrombolysis and the start of the FDCTP (Tmax: r = −0.17, 95% CI −0.82–0.67, p = 0.72; TTP: r = 0.66, 95% CI −0.19–0.94, p = 0.11).

Bland-Altman analysis for visible Tmax showed a bias of 0.82 ml (95% CI −15.5–17.2 ml) and a range of agreement levels from −65.6–67.2 ml (Supplementary Fig. 3a). Visible MDCTP TTP volumes were lower compared with FDCTP TTP (mean bias −9.11 ml, 95% CI −24.1–5.9 ml; agreement levels −70.1–51.9 ml) (Supplementary Fig. 3b).

The comparison between the volume of hypoperfused tissue as determined by MDCTP Tmax 6 s (not normalized) and the semi-automated computed normalized Tmax maps of FDCTP yielded a strong correlation and good agreement (r = 0.73, 95% CI 0.41–0.89, p < 0.001; ICC = 0.69, 95% CI 0.43–0.87). A significant proportion of the variance in the volume of hypoperfused tissue is shared by MDCTP and FDCTP (MDCTP Tmax 6 s versus FDCTP semi-automated normalized Tmax: r² = 0.53).

The comparison between semi-automated computed normalized Tmax maps of MDCTP and FDCTP yielded improved results (in comparison to MDCTP Tmax 6 s not normalized and the semi-automated computed normalized Tmax maps of FDCTP) with a strong correlation and robust agreement between the volume of hypoperfused tissue (r = 0.78, 95% CI 0.50–0.94, p < 0.001; ICC = 0.77, 95% CI 0.55–0.91; Fig. 3). Furthermore, a significant proportion of the variance was shared by the two variables (MDCTP semi-automated normalized Tmax versus FDCTP semi-automated normalized Tmax: r² = 0.60).

The comparison between semi-automated computed normalized rCBF maps of MDCTP and FDCTP showed low correlation and poor agreement, and minimal shared variance of volumes of infarct core (r = 0.16, 95% CI −0.32–0.57, p = 0.57; ICC = 0.16, 95% CI 0.01–0.84; r² = 0.03).

Bland-Altman analysis for the semi-automated normalized Tmax (MDCTP and FDCTP) showed a bias of −1.45 ml (95% CI −22.6–19.7 ml) and a range of agreement levels from −87.4 to 84.5 ml.

This study shows that FDCTP and MDCTP provide similar volumetric results when qualitatively assessing hypoperfused tissue on Tmax and TTP maps; however, direct quantitative comparison between infarct core and hypoperfused tissue on FDCTP and MDCTP using thresholding did not yield satisfactory correlations or agreements. For Tmax maps, normalization techniques proved effective in addressing this issue and led to improved results. A possible reason for the limited quantitative comparability between FDCTP and MDCTP is the inherent differences in acquisition techniques. These differences primarily concern temporal resolution, such as the time required to acquire a volume (approximately 0.5 s for MDCT versus 5 s for FDCT), the time between two consecutive acquisitions (approximately 2 s for MDCT versus 6 s for FDCT), and the number of sweeps (30 for MDCTP versus 10 for FPCTP).

On average, the qualitatively calculated volumes (segmented maximum visible extent on Tmax maps) were almost equal for MDCTP and FDCTP (0.82 ml, 95% CI −15.5–17.2 ml). By contrast, the overall volumes of tissue determined on the semi-automated normalized Tmax were smaller with MDCTP than with FDCTP (−1.45 ml, 95% CI −22.6–19.7 ml). This pattern was also observed for TTP maps (−9.11 ml, 95% CI −24.1–5.9 ml). The increased bias observed in TTP maps compared with Tmax maps may be due to measurement errors or possibly, although not significant, some degree of greater sensitivity of TTP maps to time, as volumes tend to expand with increasing duration of vessel occlusion. Because TTP maps are calculated before deconvolution, they are more sensitive to variations such as cardiac function, proximal vessel stenosis, and bolus injection quality [15].

In an exploratory analysis, smaller volumes of hypoperfusion were found on FDCTP in patients pretreated with intravenous thrombolysis, despite no substantial change in occlusion location when comparing MDCTP and FDCTP. This difference was statistically significant on TTP maps. While we cannot exclude that prior administration of intravenous thrombolysis may have had an effect on collateral flow or microcirculatory changes leading to differences in hypoperfused volume, this may also be a chance finding and the analysis is limited by the fact that only 7 of 19 patients received intravenous thrombolysis with alteplase.

Whereas our study found that quantitative results from MDCTP and FDCTP were comparable only for Tmax and not for CBF maps after normalization, previous studies have used the RAPID software (RAPID for ANGIO, iSchemaView) to quantitatively compare hypoperfused areas on MDCTP using rCBF < 30% with FDCTP using rCBF < 45%, as well as MDCTP Tmax > 6 s with FDCTP Tmax > 6 s [5, 26]. These studies demonstrated a strong correlation between these two modalities [5, 26]. In clinical practice, most institutions still rely primarily on qualitative assessment of maps rather than measuring absolute values in different regions of the affected tissue. In addition, existing literature suggests that even in the presence of limited quantitative differences in the infarct core, penumbra, and mismatch estimates between different software packages, their influence on radiologists’ decisions regarding endovascular treatment remains relatively small [27]; however, given the potential benefits it would be advantageous to develop reliable thresholds and automated approaches for comparing MDCTP and FDCTP. Up to now, accepted thresholds for MDCTP identifying infarct core and/or hypoperfused regions include rCBF < 30% and Tmax > 6 s [5, 11, 15], while proposed thresholds for FDCTP involve rCBF < 45% and Tmax > 6 s [5, 26]; however, several studies have reported an overestimation of perfusion deficits on FDCTP [28, 29]. Moreover, Zussman et al. showed that notable discrepancies can occur in quantitative MDCTP results obtained with different vendor software applications, primarily due to vendor differences rather than interoperator or intraoperator variability [30].

The acquired protocol reported here offers the possibility to grade collaterals on multiphase CTA (mCTA) images, which has been shown previously [19]. While multiphase or even single-phase CTA could be considered sufficient to select patients for thrombectomy, evaluating the reliability of a “full” perfusion imaging protocol for FPCTP is important because perfusion imaging after thrombectomy can provide important information, including an assessment of the eloquence of residual small occlusions and can help to assess whether tissue distal to residual vessel occlusions is still viable.

Numerous studies have conclusively demonstrated the practicality of FDCTP in the angiography suite [5, 8, 28, 29, 31, 32]. Petroulia et al. evaluated the comparability of FDCT and MDCT and found equal diagnostic performance in the supratentorial ventricular system and high specificity and sensitivity for the detection of intracranial hemorrhages, with the best diagnostic performance in detecting intraventricular hemorrhage, followed by intraparenchymal and subarachnoid hemorrhages [33]. High sensitivity and specificity were also demonstrated in the detection of ischemic lesions [33]; however, they found limitations in the diagnostic performance between MDCT and FDCT in all infratentorial structures and in the grey-white matter differentiation of the supratentorial and infratentorial structures [33]. Hoelter et al. compared FDCT to MDCT, including a comparison of FDCTP to MDCTP, and found that FDCTP datasets had equivalent image quality to MDCTP, providing the information needed to estimate infarct core and penumbra [34]. Another study by Niu et al. found that after applying postprocessing methods to enhance image quality, FDCTP maps were not inferior to MDCTP maps [35]. In addition, radiation exposure during FDCTP is comparable to that during MDCTP, with estimated effective doses of 4.52 mSv (without collimation) and 2.88 mSv (with collimation) [36] and both utilized similar volumes of contrast agent [28]. FDCTP is still primarily used in the one-stop management workflow [2]. If proven to provide reliable, reproducible and comparable results to MDCTP, this could pave the way towards wider implementation of FDCTP in clinical practice. The use of FDCTP imaging in the peri-interventional and post-interventional settings of AIS, for instance, has the potential to provide significant benefits to patients. Currently, the management of patients with incomplete reperfusion is challenging due to the scarcity of robust evidence to support treatment decisions [37]; however, if FDCTP performed peri-interventionally and post-interventionally is shown to provide additional reliable and relevant information, it could have a profound impact on the clinical management of these patients. Gaining proficiency in using the FDCT as a conventional CT scanner and interpreting its imaging results will be critical for translating research findings into improved patient care [5].

The distinct advantages of MRI over CT in detecting the infarct core, lacunar infarcts, and posterior circulation infarcts, are due to the high sensitivity and specificity of diffusion-weighted imaging [11]. This has led to the increased use of MRI as a baseline imaging modality. In addition, a comparison of FDCTP with post-interventional magnetic resonance perfusion (MRP) would be useful. There is a growing need to compare MRP with CTP by establishing new thresholds. Campbell et al. demonstrated that quantitative mismatch classification using rCBF and Tmax in MDCTP is similar to the MRI perfusion-diffusion mismatch [38]; however, MRI perfusion is inherently less quantitative than CTP. Also, direct comparison of perfusion parameters between the two modalities is difficult due to the nonproportional relationship between signal and contrast agent concentration, as well as technical and mathematical differences [11, 17].

This study has several limitations. First, the results are based on a retrospective study. Second, the sample size is relatively small. Third, the study included only patients with large vessel occlusion of the anterior circulation. Fourth, the FDCTP was performed by a single neurointerventionalist. Finally, we only compared MDCTP with FDCTP and did not include an evaluation of MRP. To address these limitations, future studies should aim for larger sample sizes and include patients with more distal occlusions, patients with posterior circulation occlusions, and patients with baseline MRI perfusion to gather more robust evidence and increase the generalizability of these findings. In addition, consideration of “smart” acquisition protocols tailored to contrast boluses has the potential to address inherent differences between MDCTP and FDCTP and merits future exploration.

This study showed that flat-panel detector computed tomography perfusion (FDCTP) provides robust volumetric results when qualitatively assessing hypoperfusion on time to maximum (Tmax) and time to peak (TTP) maps, with a high degree of agreement and strong correlation with the results of MDCTP; however, direct quantitative comparisons using thresholding methods did not provide satisfactory correlations. Nevertheless, for Tmax maps the application of normalization techniques proved effective in overcoming this limitation and led to improved results. Further research is needed to investigate the comparability of FDCTP with MDCTP more thoroughly. If proven reliable and consistent, the use of FDCTP in clinical practice could be greatly expanded, opening new avenues for improved patient care.