Within the IMPreST study, patients presenting an acute first-ever hemispheric LVO and undergoing triage for an endovascular acute stroke treatment were eligible. Inclusion criteria were: 72 h first clinical ischemic stroke at hospital admission, occlusion of a M1/M2-segment of the middle cerebral artery and/or intracranial internal carotid artery, perfusion deficits with cortical involvement, 18 years of age or above, living independent before stroke (corresponding to a mRS of ≤ 3), written informed consent of the patient or a documented authorization of an independent doctor, not involved in the study, or a post-hoc written informed consent of the patient or next of kin. Exclusion criteria were age under 18 years, contraindication to MRI, and presence of major neurological, psychiatric, or medical comorbidities (such as major cardiac, psychiatric and/or neurological diseases, early seizures, known or suspected non-compliance, drug and/or alcohol abuse, contra-indications for MRI). For this sub-analysis, only patients fulfilling the criteria of TICI of 2b-3 after MT with available perfusion studies at onset and <72 h after onset were considered.

The imaging data included a pre-interventional CT or MRI scan within 24 h after clinical onset. The CT perfusion was performed respecting the standard-of-care protocols at the University Hospital Zurich. The modalities diffusion weighted imaging (DWI) and contrast-enhanced perfusion MRI were acquired post-MT. Follow-up scans were performed <72 h (visit one), 7 days (+/− 2 days; visit two) and 90 days (+/− 14 days; visit three) after onset using a 3-Tesla Skyra MRI scanner (Siemens Healthineers, Forchheim, Germany).

Within the MRI protocol, the T1 magnetization-prepared rapid gradient echo (MPRAGE), followed a repetition and echo time (TR/TE) OF 2200/5.14 ms, slice thickness of 1 mm, voxel size 1x1x1 mm3, flip angle 8°, field of view of 230 mm. The DWI included 2D EPI sequence, TR/TE of 2500/75 ms, flip angle by 90°, slice thickness of 4.5 mm, voxel size with 1.5 × 1.5 × 4.5 mm3, b-values of 0 and 1000s/mm2. For the MR PWI the following protocol was applied: TR/TE of 2040/36 ms, flip angle of 90°, slice thickness of 4 mm, voxel size 1.7 × 1.7 × 4.0 mm3, FOV of 220 mm. Other MRI sequences were acquired as part of the established imaging protocol. These were not included in the analysis for this study.

For the acquisition of the blood oxygenation-level dependent cerebrovascular reactivity (BOLD-CVR) MRI sequences, standardized CO₂ impulse was applied by RespirAct™ (Thornhill Research Institute, Toronto, Canada), allowing targeted CO₂ end-tidal pressure (P_etCO₂) while maintaining iso-oxic O₂ level (13). The used CO₂ protocol in our institute followed a 100 s phase of patient specific resting P_etCO₂, whereafter the P_etCO₂ was increased by 10 mmHg for the duration of 80s, after which it returned to the resting P_etCO₂ for 120 s.

In the case of contrast-enhanced perfusion imaging, Dotarem Gadoteric acid (Gadoterate meglumine) from Guerbet, Villepinte, France, was administered intravenously with a dosage of 0.2 mL/kg body weight at a flow of 5 mL/s. Thereafter, a saline flush between 25 and 30 mL was applied to ensure full circulation of the contrast agent.

Perfusion weighted imaging (PWI) included a cerebral blood flow (CBF), cerebral blood volume (CBV) and a time-to-maximum (Tmax) analyses calculated applying RAPID^® software (14). Based on the perfusion parameter maps post MT, patients were categorized qualitatively by three authors (MB, MZ and SW) into three different reperfusion groups (normalized, hypo-, and hyperperfusion).

Apparent diffusion coefficient (ADC) maps were calculated applying Bayesian estimation with the RAPID® software.

For the BOLD-CVR imaging processing we followed the previously described Zurich analysis pipeline (15) using MATLAB2019 (The MathWorks, Inc., Natrick, United States) and SPM12 (Welcome Trust Center for Neuroimaging, Institute of Neurology, University College London). The BOLD-CVR were calculated voxel-per-voxel as percentage of BOLD signal change, divided by the absolute change in P_etCO₂ (% ΔBOLD/mmHg).

Clinical characteristics such as age, sex, vascular risk factors, time-to-recanalization, location of the occlusion, application of intravenous thrombolysis (IVT), NIHSS on admission, 24 h and 3 months after onset and mRS at admission, 36 h and 3 months after onset were collected and analyzed.

Based on the colored map of the Tmax, CBF and CBV parameters at visit one (<24 h after onset), the patients were categorized into one of the following groups: hyperperfusion, if Tmax was shorter and CBF and CBV higher compared to the contralateral hemisphere; hypoperfusion, if Tmax was longer and CBF and CBV lower compared to the contralateral hemisphere; normalized, if Tmax, CBF and CBV were comparable to the contralateral side. The comparison of the perfusion parameters was focused on the region within the infarct and the penumbra region. The selection was performed by three authors (MB, MZ, SW). Discrepancies in the selection process were resolved through discussion.

Descriptive statistics were used to compare the demographics, clinical patterns, therapeutic modalities and advanced perfusion study parameters between the perfusion groups. Ordinal, non-dichotomous variables are presented as median and quartiles one and three (Q1–Q3). Categorical dichotomous and non-dichotomous variables are presented as percentages. Assuming a nonnormal distribution, the Kruskal-Wallis-Test was applied to compare the variables. We defined the level of significance as the probability for a type I error of less than 5%, corresponding to a p-value of <0.05.

All statistical analysis were performed in R, version 4.3.0.

Normalized perfusion after successful MT correlates with a favorable clinical outcome (14). This cohort presented accordingly a trend for the lowest mRS score in the 3 months follow up. Collateral status tended to be better and final infarct volume to be smaller. As potential confounders, younger age, more distal artery occlusions, and lower NIHSS at onset were found on average in this group of patients. This favorable clinical and radiological outcome is consistent with the results of previous studies (16, 17).

Patients presenting with hypoperfusion post-MT showed a tendency toward the second-worst clinical outcome. They also had the longest time-to-recanalization. Accordingly, in previous studies, RF was mostly linked to persistent hypoperfusion associated with longer onset-to-arrival and time-to-recanalization (11, 18–20). Additionally, the duration of the MT was significantly longer in patients with hypoperfusion, potentially indicating difficult interventons with increased likelihood of residual vessel obstructions or fragment embolism. In accordance with Ng et al., we also observed the highest rate for hemorrhagic transformation in this group (11). In our cohort, this was the group with most patients receiving IVT treatment.

This group also displayed a trend for poor leptomeningeal collaterals compared to the normalized group, which was previously suggested to be associated with RF (21) and generally perceived as an independent predictor for poor clinical outcome (3, 22).

As one key finding of our study, we found a tendency for the least clinical improvement and poorest clinical outcome within the hyperperfusion group. This cohort also presented a trend for the worst collateral status, largest infarct volumes, and, interestingly, shortest MT-duration and time-to-recanalization. In general, these three perfusion patterns have been previously described solely after thrombolysis, therefore not exclusively in patients with LVO after MT and only within a frame of 24 h after treatment (23). The poor clinical outcome and clinical recovery could be attributed to higher age, more proximally affected vessels, along with higher baseline NIHSS, and larger infarct volumes. We also observed the smallest difference in NIHSS 24 h after onset in this group, even though none of these values reached statistical significance. These findings are in accordance with Shimonaga et al.’s (19). Potreck et al. came to an opposite conclusion: their hyperperfusion cohort showed a similar outcome compared with their normalized perfusion cohort. This difference could be explained by the different assessment time points of our studies. While their study cohorts were evaluated based on perfusion patterns obtained within 24 h after successful recanalization, our analysis is based on perfusion data acquired within 3 days post-recanalization and includes follow-up until day 90. Their hyperperfusion cohort included patients with temporary hyperperfusion, with a potential to normalization within 18–36 h after MT (17). However, we observed that hyperperfusion post-MT resulted in a long-term hypoperfusion, which was associated with poor clinical outcome, and indicating that any post-MT perfusion alterations may be associated with a worse clinical outcome (11, 24, 25). Therefore, this study suggests that hypo- and hyperperfusion should both be perceived as RF, indicating a poor clinical outcome. In this regard, Lin et al. have shown the importance of post-MT PWI imaging to recognize hyperperfusion and its consequences (26). Our results emphasize the importance of advanced imaging up to 3 months post-MT since hypoperfusion, as a sign of RF, might appear also after such a latency. Accordingly, identifying patients with greater risk for poor outcome could be enabled by advanced imaging.

In our study, hyperperfusion was not significantly correlated with shorter onset-to-recanalization, indicating that merely reducing time-to-recanalization does not prevent RF. Similarly, Shimonaga et al. could not find a significant difference, even though they described a tendency toward a longer time to recanalization (19).

As shown in rodent models, leptomeningeal collaterals are crucial for the survival of at-risk-tissue (27). Preclinical data suggest that fast, overshooting reperfusion after recanalization in stroke is linked to poor leptomeningeal collaterals and reperfusion injury (28), which was also reflected in our patients with altered perfusion patterns. To this regard, the reperfusion injury describes a damage occurring at the vascular level due to reperfusion. The exact pathophysiological mechanisms are not fully understood yet. One of the most accepted theories proposes that following a vessel occlusion, endothelial damage at the level of the smooth muscle cells may lead to an impaired autoregulation and thus to an increased rigidity of the cerebral vessels. As a consequence, alteration of the cerebral perfusion pressure may favor increased blood–brain barrier permeability, thus favoring hemorrhagic transformation (26, 29). Also, increased vessel rigidity and a damaged endothelium, to which a thrombus might attach with greater difficulty, could explain the shorter duration of MT in the hyperperfusion group. Another theory suggests a damaging effect of free radicals during hypoxic phases, leading to altered endothelium function, vasodilatation, and in a later phase to pericyte contraction, and increased blood–brain-barrier permeability (8, 26), which also possibly could reflect the changes in the perfusion studies in our hyperperfusion group between visit one and three. A relevant role could play the reactive hyperemia, in which a phase of cerebral ischemia is followed by a transient increase in perfusion due to autoregulatory mechanisms. This phenomenon has been associated with both favorable prognostic implications (30) and adverse outcome (31) However, this does not explain the long-term differences observed at Visit two and three. Similarly, the so-called luxury perfusion, which frequently appears after restored blood-supply, may explain a perfusion alteration in the subacute phase of a stroke or after MT respectively, but not the persistent changes in the affected area (32–34).

In addition, due to mechanical stress, MT itself may lead to additional damage of the endothelium of the vessel wall (35). Damaged endothelium activates thrombocytes and potentially triggers an immune response involving neutrophils, leucocytes, and monocytes (7, 10, 35, 36), and therefore leading to a malfunction of autoregulation.

In accordance with some other studies, our dataset shows an increased rate of hemorrhagic transformation in the hyperperfusion cohort (19, 29, 37–41). Additionally, our study underlines poor clinical outcome in this group. Hypoperfusion in the affected area at visit three followed the initial hyperperfusion. To the best of our knowledge, this tendency has not been described before and emphasizes the risk of hyperperfusion post-MT.

As mentioned above, patients exhibiting hypoperfusion may have experienced residual distal vessel occlusion due to distal embolization. In such cases, a timely repeated attempt with MT and/or local application of thrombolysis could be beneficial. Given the not specified pathophysiology of the phenomenon, it is not possible to implement a clear therapeutic strategy in the case of the hyperperfusion group.

Also, some therapeutic agents have been tested in rodent models, to counteract some proposed mechanism responsible for the reperfusion failure. Therefore, Yang et al. tested an agent called NA-1 which successfully inhibited ischemia-induced pericyte constriction (42). Regarding the neutrophilic stalls, El Amki et al. successfully applied anti-Ly6G-antibodies in rodent models, which hindered the adhesion of neutrophils, preventing perfusion disturbances in recanalized vessels (7), thus providing a proof-of-concept for new therapeutic strategies.

BOLD-CVR displayed significant differences between the reperfusion groups, thus proving a suitable and sensitive tool for impaired cerebral reperfusion. Both, hypo- and hyperperfusion post-MT correlated with a reduced cerebrovascular reserve in BOLD-CVR imaging. Using the same data set, another analysis suggested poor clinical outcome 3 months post intervention in patients with impaired BOLD-CVR (43). Therefore, BOLD-CVR might be useful as an early-stage predictor of reperfusion failure and poor clinical outcomes post-MT (27, 44).

In this study, we used advanced hemodynamic imaging techniques to identify three different post-MT perfusion patterns in patients with LVO over a period of 3 months. Previous studies focused on either hyperperfusion (19, 26, 29) or hypoperfusion (11, 45) without comparing them, and only applied advanced imaging shortly after stroke (11, 19, 26, 27, 29, 45). Thus, we were able to unveil the course of tissue reperfusion after MT, particularly finding that initial hyperperfusion was followed by hypoperfusion and accompanied by reduced CVR.

However, our study has several limitations. One important limitation is the small number of patients, impeding proper statistical analysis. The small number resulted from a relatively high rate of withdrawal during the acute enrolment phase, as well as from frequently incomplete datasets, either due to missing perfusion studies prior to MT or the complexity of certain acquisitions, such as BOLD-CVR, which requires active patient participation. Furthermore, we only analyzed patients fulfilling the criteria for a successful MT (TICI 2b-3), in order to study RF after successful recanalization, which however, next to the non-randomized nature of the cohort, limits generalization of findings to other patients. While we performed a qualitive analysis of post-MT perfusion using perfusion imaging, further studies could include additional quantitative perfusion analysis of the interested areas.