Magnetic Resonance Imaging or Computed Tomography Before Treatment in Acute Ischemic Stroke: Effect on Workflow and Functional Outcome

Since the benefits of recanalization are highly time-dependent in acute ischemic stroke,^1–4 minimizing the workflow duration is crucial. Computed tomography (CT) is the most widely used diagnostic tool because of its availability and rapid acquisition time. However, magnetic resonance imaging (MRI) has advantages over CT. Diffusion-weighted imaging (DWI) best detects acute ischemia,⁵ stroke-like events^6–8 and is the reference for infarct core extent measurement, which is crucial in selecting patients for endovascular treatment in the later time window.^9,10 In patients with unknown onset time, MRI identifies strokes that occurred approximately within the previous 4.5 hours and could benefit from thrombolysis.¹¹ MRI drawbacks theoretically include screening time for contraindications, limited access, additional time needed to clear the MRI scan and place the patient on the table, and longer scan duration.

The recent trials showing the efficacy of endovascular therapy in acute ischemic stroke used primarily CT-based approaches.¹² Consequently, detailed data on MRI-based workflow are limited.^13–15 A recent study¹⁴ showed that the workflow times of 128 MRI-selected patients before endovascular therapy were similar to those reported in 3 large CT-based trials.^16–18 However, this monocentric study had no CT-selected subgroup for comparison, excluded patients with large core and did not record the precise duration of the MRI step.

In the multicenter THRACE trial (Thrombectomie des Artères Cérébrales),¹⁹ enrolling centers were free to use CT or MRI before randomization, resulting in a large proportion of MRI-selected patients. We compared the workflow and functional outcome in acute ischemic stroke patients screened by MRI or CT, with particular emphasis on the imaging step duration.

The data that support the findings of this study are available from the corresponding author on reasonable request.

Among 414 randomized patients in the THRACE trial, 401 patients were included in the present study (2 patients withdrew consent, screening imaging modality was unknown in 8 patients, and imaging time information was not available for 2 MRI-selected patients and 1 CT-selected patient). Of the 401 patients from 25 centers, 299 were in the MRI group (median [IQR] DWI volume: 16 mL [9–52]) and 102 in the CT group (median [IQR] Alberta Stroke Program Early CT Score: 9 [8–10]). Of the 25 centers, 15 used MRI and 5 used CT as the prime imaging modality, defined as a modality used for >80% of the examinations performed. The other 5 centers used MRI or CT without priority being given to either modality and according to availability. The detailed imaging protocol of each center is shown in Table I in the online-only Data Supplement.

Differences between the MRI and CT groups are given in Table 1. In univariable analysis, MRI-selected patients were less frequently diabetic and had less renal insufficiency than CT-selected patients. There was no statistical difference in baseline National Institutes of Health Stroke Scale scores between groups. MRI and CT were similarly performed during off hours (MRI: 50/299 [17%]; CT: 18/102 [18%]; P=0.83), and weekends (MRI: 55/299 [18%]; CT: 14/102 [14%]; P=0.28). None of the delays from stroke onset differed between the CT and MRI groups (Table 2). The delay between clinical examination at hospital arrival and start of imaging did not differ between the MRI and CT groups (MRI: 17 minutes [11–27]; CT: 12 minutes [7–32]; P=0.13). As shown in the Figure, the mean duration of each discrete step in patient workflow was similar between the CT and MRI groups, except for scan duration and imaging-to-intravenous tPA time.

MRI scan duration was overall longer (MRI: median 13 minutes [IQR, 10–16]; CT: 9 minutes [7–12]; P<0.001), irrespective of the imaging protocol (without perfusion imaging [MRI: 12 minutes (10–15); CT: 7 minutes (2–11); P<0.001] or with perfusion imaging [MRI: 16 minutes (15–19); CT: 10 minutes (9–13); P<0.001]). In the MRI group, 68/299 (23%) scans had perfusion imaging, 83/299 (28%) had cervical MR angiography, and 4/299 (1%) had both. Median MRI scan duration was significantly longer with perfusion imaging (16 minutes [15–19]) or with cervical MR angiography (15 minutes [12–18]) than without contrast injection (11 minutes [9–13]), P<0.001 for both comparisons. In the CT group, 62/102 (61%) patients had CT perfusion in addition to the CTA required to confirm arterial occlusion and 89/102 (87%) had a CTA of the neck. Median CT duration was significantly longer with perfusion imaging (10 minutes [9–13]) than without, that is, simple CT/CTA (7 minutes [2–11]; P<0.001). Similarly, advanced imaging protocols (CT perfusion or MRI) were longer (13 minutes [10–16]; P<0.001) than simple CT/CTA protocols.

Despite a longer scan duration in the MRI group, time from start of imaging acquisition to initiation of intravenous tPA was shorter (MRI: 31 minutes [24–43]; CT: 42 minutes [34–54]; P<0.001), because of a shorter time from end of imaging to initiation of intravenous tPA (MRI: 18 minutes [11–28]; CT: 33 minutes [24–43]; P<0.001; Figure). To disentangle center effect from imaging modality effect on the delay from imaging to initiation of intravenous tPA, we post hoc analyzed the data using a mixed model, with center as a random effect and imaging modality as a fixed variable. Both imaging modality and center effects were significant (P<0.001). Of note, there was no difference in the time from start of imaging to initiation of intravenous tPA between standard and multimodal imaging protocols, using either MRI (30 minutes [22–43] versus 33 [25–42]; P=0.19) or CT (38 minutes [30–60] versus 42 minutes [35–54]; P=0.63).

Favorable outcome was associated with imaging modality in univariable analysis (MRI: 51%, 151 of 294 patients; CT: 38%, 38 of 100 patients; P=0.021). In multivariable analysis with imaging modality forced into the model (n=370/401), favorable outcome was associated with mechanical thrombectomy group (odds ratio [OR], 1.71; 95% CI, 1.08–2.69; P=0.021), younger age (OR, 0.97; 95% CI, 0.95–0.98; P<0.0001), less severe baseline National Institutes of Health Stroke Scale (OR, 0.89; 95% CI, 0.84–0.94; P<0.0001), lower blood glucose (OR, 0.48; 95% CI, 0.25–0.91; P=0.024), proximal portion of the middle cerebral artery occlusion (OR, 3.76; 95% CI, 1.86–7.60; P=0.0002). MRI was not associated with outcome (OR, 0.42; 95% CI, 0.27–2.48; P=0.12). Symptomatic hemorrhagic transformation at 24 hours (MRI: 4 of 299 patients; CT: 3 of 102 patients; P=0.45) and mortality at 3 months (MRI: 11%, 33 of 293 patients; CT: 16%, 16 of 100 patients; P=0.22) did not differ between the 2 groups.

Finally, all the above results were similar after exclusion of the 13 patients (3%) who had both imaging modalities before randomization (9 with CT first then MRI and 4 with MRI first then CT).

This multicenter study provides a comprehensive picture of the acute stroke workflow in a country where MRI has been recommended as the prime brain imaging technique for almost 10 years.²² In the THRACE trial, the choice of imaging modality for patient inclusion was left to the enrolling center and, as a result, 74% of patients were imaged with MRI. Overall, there was a significant difference of a few minutes between CT and MRI imaging duration, but this had no impact on the global workflow as time from stroke onset-to-intravenous tPA and to endovascular therapy did not depend on the choice of imaging modality. Finally, this choice had no impact on functional outcome at 3 months.

There is a prevailing view that stroke MRI scan duration is in the order of 20 minutes and therefore much longer than CT.²³ However, in centers that have been using MRI as the prime diagnostic imaging tool for years, standardized stroke MRI protocols last around 10 minutes.^14,24,25 This duration could be further reduced to 6 minutes with the use of echo planar imaging at 3 Tesla.²³ However, the above scan durations are merely based on the sum of the duration of each sequence, and therefore exclude intersequence time or events that might occur during the MRI examination. The recent endovascular trials provide detailed timings for door-to-imaging and imaging to the next workflow steps,⁴ but the precise duration of the complete MRI session is lacking. The 13-minute median MRI scan duration we recorded over a large panel of centers is representative of a real-world setting, since there was no attempt to standardize MRI protocols across centers. In contrast to MRI, CT is commonly considered as a snapshot in the stroke workflow. Although this was reasonable for noncontrast CT, it is no longer true when CTA is added, a requirement before endovascular treatment decision. As with MRI, the whole CT scan duration is longer than the sum of CT and CTA acquisition times. We here provided a measurement of total scan duration (CT/CTA: median, 7 minutes; CT/CTA/CTP: median, 10 minutes), which matches that in the SWIFT PRIME trial (Solitaire With the Intention for Thrombectomy as Primary Endovascular Treatment; CT/CTA: median, 9 minutes).²

Of note, 61% of patients in the CT group had perfusion imaging, as opposed to 23% in the MRI group. In MRI centers, the large amount of information available with standard sequences, such as direct visualization of the infarct core extent on DWI may have obviated the need for additional perfusion. To avoid any loss of time in tailoring the imaging protocol to each individual, standardized imaging stroke protocols (with or without perfusion imaging) are used in some centers,^24,26 while others add advanced imaging techniques in selected cases. On a large scale, we showed, like others,²⁷ that performing multimodal CT did not delay intravenous tPA administration, and this was also true for multimodal MRI, thus complying with the 2018 American Heart Association guidelines.²⁸

Although it did not impact the time from onset-to-intravenous tPA or to endovascular treatment, time from imaging-to-intravenous tPA was surprisingly shorter in the MRI group. This was related to both an imaging modality effect and a center effect in the post hoc analysis. About the imaging modality effect, one possible explanation is that the treatment decision was made while MRI acquisition was still in progress. In patients with a puzzling clinical presentation, a treatment decision based on straightforward DWI images with clear-cut signal changes might have been faster than one based on subtle early CT ischemic changes. About the center effect, it likely lies in organizational differences between CT and MRI centers. In centers that prioritized round-the-clock immediate access to MRI for stroke patients, the global workflow might have been better organized and more time efficient, thus reducing time from imaging-to-intravenous tPA when compared with CT centers. Of note, in CT-selected patients, imaging-to-intravenous tPA time was similar to that in the IMS III international multicenter trial (Interventional Management of Stroke III).²⁷

Times from stroke onset and from clinical examination on admission to imaging were similar in the CT and MRI groups. This suggests that none of the following steps are major obstacles to the use of MRI, in centers where it is available: access to MRI, including transport to the scanner and immediate access to the MRI room, completion of the MRI safety questionnaire to exclude contraindications, and placement of the patient on the MRI table. The systematic use of MRI in acute stroke implies round-the-clock priority access to MRI with direct admission in the imaging department, and close geographic localization between the stroke center, MRI unit, and angiography suite. The main clinical restrictions to the use of MRI remain contraindications and restless or unstable ischemic stroke patients, accounting for 9% of patients.¹⁴ This acceptable low rate might even decrease in the near future, as the vast majority of metal implants are MRI safe, and even concerns related to pacemakers may be overstated.²⁹ Taken together, our findings support the notion that MRI can be accomplished rapidly and within a similar timeframe as CT before endovascular therapy, in line with several studies showing that MRI-based selection is feasible and safe.^13–15,30 One should however keep in mind that, although there is sufficient evidence that MRI is as reliable as CT for the detection of acute intracerebral hemorrhage, dedicated training with interpretation of gradient echo sequence is needed for MRI to be safely used before treatment decision.³¹

Our study has several limitations. (1) The THRACE trial was not designed for the purpose of the current study and randomization was not stratified on imaging modality. As a consequence, the choice of imaging modality may have depended on confounding variables. For instance, in centers using MRI as a prime screening imaging modality in stroke patients, the most severely ill patients might be directed towards CT. However, in THRACE, stroke clinical severity did not differ between the CT and MRI groups, likely because the majority of patients medically ineligible for MRI suffer from intracerebral hemorrhage,³² whereas THRACE focused on ischemic stroke. We nevertheless post hoc verified that all results were similar when we excluded patients imaged with CT in MRI centers, and patients imaged with MRI in CT centers. (2) There was no synchronization across all clocks and imagers, which may have led to some degree of inaccuracy in the times recorded. However, there is no reason why clocks should be more inaccurate in one group. (3) Some of the workflow times were not recorded in the case report form of the THRACE trial (arrival to hospital door and groin puncture), therefore preventing a comprehensive evaluation of the workflow, as well as comparisons with others and current guidelines for door-to-imaging and door-to-groin puncture times.^28,33 However, times from stroke onset and from clinical examination on admission to imaging were not longer in the MRI group, suggesting that access to MRI was not detrimental in terms of delay in patient management. Times of entry in angiography suite and first angiographic run provided an estimation of the groin puncture time. (4) Admission status (direct admission versus transfer to endovascular-capable centers) was not available, preventing comparison of workflow between these 2 groups. However, given the THRACE inclusion criteria (initiation of intravenous tPA within 4 hours and thrombectomy within 5 hours of symptom onset), most patients were directly referred to endovascular-capable centers.

In conclusion, although MRI scan duration is a few minutes longer than CT, workflow in THRACE demonstrates that real-world application of MRI for acute stroke evaluation before treatment can be accomplished as rapidly as CT-based selection paradigms. This should help to promote wider use of MRI in acute stroke by counterbalancing concerns about delay and eligibility.