Impact of RAPID-Based Patient Selection on Endovascular Therapy Outcomes in Acute Ischemic Stroke: A Retrospective Comparative Study

Abstract

Objective

Perfusion imaging is widely used in acute ischemic stroke to guide endovascular thrombectomy (EVT). This study evaluated clinical outcomes among patients selected for EVT using perfusion software-based analysis compared with those selected without it.

Methods

We conducted a retrospective comparative analysis of patients with large- or medium-vessel occlusion who underwent EVT between 2024 and 2025. Patients were categorized into a perfusion software group (RAPID implementation period) and a non-perfusion software group (non-RAPID period) based on the use of RAPID imaging software (iSchemaView, Menlo Park, CA, USA). The primary outcome was the proportion of patients achieving a good clinical outcome at 90 days, defined as a modified Rankin Scale (mRS) score of 0–3. Secondary outcomes included a shift analysis of mRS scores, procedural time metrics, all hemorrhagic events including symptomatic intracerebral hemorrhage (sICH), and 90-day mortality.

Results

A total of 54 patients were included (RAPID implementation period, 26; non-RAPID period, 28). At 90 days, the proportion of patients achieving a good outcome (mRS 0–3) was similar between the RAPID implementation period group and the non-RAPID period group (50.0% vs. 46.4%; P = 0.72). In the ordinal shift analysis of mRS scores, there was no significant difference in the overall distribution between groups (common odds ratio, 0.91; 90% confidence interval [CI], 0.41–1.99; P = 0.84). The median time from hospital arrival to groin puncture was also similar—45 min (interquartile range [IQR], 40–58) versus 46 min (IQR, 39–63; P = 0.96). The incidences of any intracerebral hemorrhage (30.8% vs. 32.1%), sICH (0% vs. 3.6%), and 90-day mortality (15.0% vs. 7.1%) were likewise comparable between the RAPID and non-RAPID periods.

Conclusion

These findings suggest that automated, perfusion-based patient selection enhances workflow standardization and can be seamlessly integrated into acute stroke management to optimize both speed and safety.

Introduction

The landscape of acute ischemic stroke (AIS) management has evolved substantially over the past decade, particularly with the advent of endovascular thrombectomy (EVT) as the standard of care for patients with large vessel occlusion (LVO). Early randomized controlled trials, such as EXTEND-IA¹⁾ and SWIFT PRIME²⁾ demonstrated the efficacy of EVT in the early time window (≤6 h from onset), primarily using perfusion imaging for patient selection. However, landmark trials such as DEFUSE 3³⁾ and the DAWN trial⁴⁾ expanded the treatment window up to 24 h, relying on advanced imaging criteria, particularly perfusion imaging analyzed by automated software like RAPID (iSchemaView, Menlo Park, CA, USA),⁵⁾ to identify patients with salvageable penumbra.

In this context, perfusion imaging has gained widespread international acceptance as a valuable tool for patient selection, especially in the extended time window. RAPID enables automated and standardized quantification of core infarct (e.g., cerebral blood flow <30%) and hypoperfused tissue (e.g., Tmax >6 s), thereby facilitating reproducible decision-making across institutions. Despite its demonstrated utility in Western settings, the real-world adoption of RAPID in Japan has been limited. This is largely due to structural and economic barriers, including a lack of public reimbursement, the high cost of software licensing, and limited availability of compatible imaging infrastructure.

Consequently, many Japanese stroke centers, including comprehensive stroke centers, have continued to rely primarily on conventional imaging modalities such as non-contrast CT, CTA, or MRI. In settings where automated perfusion quantification is not available, recent Japanese guidelines recommend semi-quantitative estimation of ischemic core using the Alberta Stroke Program Early CT Score (ASPECTS), even in the extended time window. In real-world practice, patient selection for EVT beyond 6 h has therefore often been based on a combination of ASPECTS assessment, CTA findings, and neurological severity, rather than on quantitative perfusion-based thresholds.⁶⁾ While this approach has proven effective in experienced centers with imaging expertise, it may not be easily generalizable to all stroke centers. In this context, automated perfusion imaging software such as RAPID may offer an additional value by reducing the cognitive burden of image interpretation and enabling more standardized and reproducible patient selection across different clinical settings, independent of individual physician experience.

In the Japanese clinical setting, the role of automated perfusion imaging in EVT triage remains incompletely defined, despite its established value in randomized trials. We hypothesized that automated perfusion imaging could primarily contribute to more standardized and efficient patient selection in real-world practice and secondarily might translate into improved functional outcomes, although any incremental benefit was expected to be modest in experienced centers. To better understand the role of automated perfusion imaging in Japanese clinical practice, we conducted a retrospective, single-center comparative study analyzing EVT cases across 2 equivalent time frames: April to June 2024 (RAPID implementation period) and April to June 2025 (non-RAPID period), following the implementation timeline of RAPID at National Cerebral and Cardiovascular Center. The study aimed to evaluate whether RAPID-based patient selection influenced procedural efficiency and short-term clinical outcomes. By comparing 2 consecutive cohorts treated under similar institutional conditions, we sought to clarify the practical impact of perfusion imaging–guided triage on workflow and treatment strategies for AIS in a Japanese comprehensive stroke center.

Materials and Methods

Study design and setting

We conducted a retrospective, single-center observational study at the National Cerebral and Cardiovascular Center in Suita, Osaka, Japan, comparing consecutive patients who underwent EVT during 2 periods: April 1 to June 30, 2024 (RAPID [iSchemaView] implementation period), and April 1 to June 30, 2025 (non-RAPID period). The primary objective was to evaluate the impact of RAPID software implementation on patient selection, procedural metrics, and early clinical outcomes. The institutional review board of the National Cerebral and Cardiovascular Center approved the study (IRB No. M23-073-13, M30-081-3), and the requirement for patient consent was waived due to the retrospective design.

Patient selection

Eligible patients were those aged ≥18 years who underwent EVT for AIS due to large- or medium-vessel occlusion, including the common carotid artery, internal carotid artery (ICA), M1 or M2 segment of the middle cerebral artery (MCA), vertebral artery (VA), or basilar artery (BA), within 24 hours from symptom onset or last-known-well during the study periods. Patient selection for EVT was generally based on contemporary Japanese guidelines and institutional practices.⁶⁾ There was no predefined upper age limit for EVT eligibility. In principle, patients with extensive early ischemic changes (ASPECTS ≤2) were not considered candidates for EVT. Patients with mild neurological deficits (National Institutes of Health Stroke Scale [NIHSS] score ≤5) were also generally excluded; however, EVT could be considered in selected cases when the symptoms were judged to be disabling. In the non-RAPID period, the decision was based on clinical and conventional imaging criteria (non-contrast CT, CTA, and, in selected cases, MRI), while in the RAPID implementation period, perfusion imaging with automated analysis using RAPID was incorporated into the decision-making process. RAPID provided quantitative estimates of ischemic core (cerebral blood flow <30%) and penumbra (Tmax >6 s) evaluated during the RAPID implementation period. Patient selection was generally guided by the presence of a conventional target mismatch (TMM) (core <70 mL, mismatch [MM] ratio >1.8, MM volume >15 mL), where MM ratio was defined as the ratio of Tmax >6 s volume to the ischemic core, and MM volume as the volume of tissue with Tmax >6 s min the ischemic core. A limited number of patients with relatively large ischemic cores were treated based on individual clinical considerations.

Data collection

Clinical, demographic, and procedural data were collected from electronic medical records. Baseline characteristics included age, sex, NIHSS score, comorbidities, and pre-stroke modified Rankin Scale (mRS) score. Imaging variables included the ASPECTS and occlusion site. Procedural details included time metrics (onset-to-door, door-to-puncture, and puncture-to-reperfusion), successful reperfusion defined as an expanded Thrombolysis in Cerebral Infarction (eTICI) grade of 2b–3, and the use of intravenous thrombolysis. First-pass effect (FPE) and modified FPE (mFPE) were defined as achieving complete or near-complete reperfusion (eTICI ≥2c) and successful reperfusion (eTICI ≥2b50), respectively, after a single pass.

Outcomes

The primary outcome was the proportion of patients achieving a good clinical outcome at 90 days, defined as an mRS score of 0–3. Secondary outcomes included an ordinal shift analysis of mRS scores, the proportion of patients with an mRS score of 0–2 at 90 days as a measure of functional independence, procedural time metrics, symptomatic intracerebral hemorrhage (sICH), and 90-day mortality. Because approximately one quarter of the cohort (13 of 54 patients) had large ischemic cores (≥50 mL), reflecting real-world heterogeneity, the primary clinical outcome was defined as mRS 0–3 rather than the conventional mRS 0–2, based on evidence from the large-core trial RESCUE-Japan LIMIT.⁷⁾

Statistical analysis

Comparative analyses were performed between the RAPID implementation period group and the non-RAPID period group. Continuous variables were expressed as medians with interquartile ranges and compared using the Mann–Whitney U test. Categorical variables were expressed as counts and percentages and analyzed using the chi-squared test or Fisher’s exact test, as appropriate. For functional outcomes, the distribution of 90-day mRS scores was compared between groups using ordinal logistic regression (shift analysis), and results were expressed as common odds ratios (ORs) with 90% confidence intervals (CIs). In addition, univariate and multivariable logistic regression analyses were performed to estimate ORs for achieving a good clinical outcome and functional independence. Baseline NIHSS score was included as a prespecified confounder in multivariable models, given its strong association with clinical outcomes. Given the small sample size, a 90% rather than 95% CI was used to improve power. As a sensitivity analysis, unadjusted comparisons of proportions for functional and safety outcomes were performed in patients treated beyond 6 h from the last known well. A P-value <0.05 was considered statistically significant. Statistical analyses were performed using SPSS version 27 (IBM Corp., Armonk, NY, USA).

Results

Patient characteristics

A total of 54 patients were included (RAPID [iSchemaView] implementation period: 26; non-RAPID period: 28). Baseline characteristics were well balanced between the 2 groups, with no significant differences in age (median, 80 vs. 82 years), sex, comorbidities, or stroke severity (median NIHSS score, 21 in both). Serum creatinine levels were higher in the RAPID implementation period group (1.08 vs. 0.77 mg/dL; P = 0.01), whereas other laboratory parameters and occlusion sites were comparable (Table 1). The proportion of patients who underwent EVT within 24 h from the last known well did not differ substantially between the 2 periods (Supplementary Table S1).

Table 1. Clinical characteristics.

	RAPID implementation period group (n = 26)	Non-RAPID period group (n = 28)	P-value
Age (years), median (IQR)	80 (74–87)	82 (71–87)	0.64
Female, n (%)	11 (42.3)	14 (50.0)	0.57
Medical history
Hypertension, n (%)	16 (61.5)	20 (71.4)	0.44
Diabetes mellitus, n (%)	2 (7.7)	3 (10.7)	0.70
Dyslipidemia, n (%)	8 (30.8)	10 (35.7)	0.70
Atrial fibrillation, n (%)	16 (61.5)	16 (57.1)	0.74
Ischemic heart disease, n (%)	3 (11.5)	3 (10.7)	0.92
Stroke history prior to index event, n (%)	5 (19.2)	5 (17.9)	0.90
Premorbid antiplatelet, n (%)	5 (19.2)	3 (10.7)	0.38
Premorbid anticoagulant, n (%)	6 (23.1)	10 (35.7)	0.31
Index stroke data
Premorbid mRS, median (IQR)	0 (0–2)	0 (0–3)	0.48
Premorbid mRS 0–1, n (%)	19 (73.1)	17 (60.7)	0.34
NIHSS score, median (IQR)	21 (13–27)	21 (17–28)	0.41
Systolic blood pressure (mm Hg), median (IQR)	150 (123–161)	149 (135–158)	1.00
WBC (103/µL), median (IQR)	8.2 (6.2–9.3)	9.4 (7.3–10.9)	0.16
Hemoglobin (g/dL), median (IQR)	12.6 (11.2–14.1)	12.9 (11.2–14.8)	0.44
Platelet (103/µL), median (IQR)	213 (160–244)	205 (180–229)	0.81
Serum creatinine (mg/dL), median (IQR)	1.08 (0.81–1.54)	0.77 (0.64–1.17)	0.01
BNP (pg/dL), median (IQR)	205 (127–604)	154 (82–486)	0.34
CRP (mg/dL), median (IQR)	0.29 (0.10–2.47)	0.54 (0.12–4.9)	0.47
D-dimer (µg/dL), median (IQR)	2.2 (1.5–5.2)	2.0 (1.0–7.7)	0.49
Occlusion site			0.80
ICA, n (%)	7 (26.9)	7 (25.0)
MCA M1, n (%)	11 (42.3)	10 (35.7)
MCA M2, n (%)	4 (15.4)	6 (21.4)
Basilar, n (%)	2 (7.7)	4 (14.3)
Baseline ASPECTS, median (IQR)	8 (6–9)	8 (7–9)	0.58
Intravenous thrombolysis, n (%)	10 (38.5)	6 (21.4)	0.17
Core volume (m), median (IQR)	10.5 (4.8–55.3)	NA
Tmax >6 s (m), median (IQR)	177 (107–259)	NA
Time metrics
Time from hospital arrival to groin puncture (min), median (IQR)	45 (40–58)	46 (39–63)	0.96
Time from the last known well to groin puncture (min), median (IQR)	171 (101–593)	156 (102–552)	0.99
Time from last-known-well to groin puncture >6 h, n (%)	10 (38.5)	9 (32.1)	0.63

Values are n (%) or median (IQR).

ASPECTS, Alberta Stroke Program Early CT Score; BNP, brain natriuretic peptide; CRP, C-reactive protein; ICA, internal carotid artery; IQR, interquartile range; MCA, middle cerebral artery; mRS, modified Rankin Scale; NA, not applicable; NIHSS, National Institutes of Health Stroke Scale; RAPID, iSchemaView, Menlo Park, CA, USA; WBC, white blood cell

Primary and secondary outcomes

At 90 days, the proportion of patients achieving a good clinical outcome (mRS 0–3) was similar between the RAPID implementation period group and the non-RAPID period group (50.0% vs. 46.4%; P = 0.79) (Fig. 1). The proportion of patients achieving functional independence (mRS 0–2 at 90 days) was also comparable between the RAPID implementation period and non-RAPID period groups (34.6% vs. 39.0%; P = 0.72). In ordinal logistic regression (shift analysis) of 90-day mRS scores, there was no significant difference in the overall distribution between the RAPID implementation period group and the non-RAPID period group (common OR, 0.91; 90% CI, 0.41–1.99; P = 0.84), indicating that the introduction of automated perfusion imaging maintained comparable functional outcomes (Fig. 2).

Fig. 1. Ninety-day mRS outcomes for RAPID implementation versus non-RAPID periods. The proportion of patients achieving a good clinical outcome (mRS 0–3) was comparable between the RAPID implementation period (52%) and the non-RAPID period (46.4%) (P = 0.79).

mRS, modified Rankin Scale; RAPID, iSchemaView, Menlo Park, CA, USA

Fig. 2. Distribution of 90-day mRS scores between 2 groups (RAPID implementation period group vs. non-RAPID period group) for shift analysis. Distribution of mRS scores at 90 days (shift analysis) comparing the RAPID implementation period group and the non-RAPID period group. The overall distribution of mRS scores at 90 days was similar between the 2 groups (OR, 0.91; 90% CI, 0.41–1.99; P = 0.84). The analysis indicates that the use of automated perfusion imaging software did not alter the overall functional outcome distribution following endovascular therapy.

CI, confidence interval; mRS, modified Rankin Scale; OR, odds ratio; RAPID, iSchemaView, Menlo Park, CA, USA

The incidence of any intracerebral hemorrhage was similar between groups (30.8% vs. 32.1%; P = 1.00), as was the rate of hemorrhagic transformation (19.2% vs. 28.6%; P = 0.63). Mortality at 90 days did not differ significantly (15.0% vs. 7.1%; P = 0.66), and no sICH occurred in the RAPID implementation period group (Table 2). In a sensitivity analysis restricted to patients treated beyond 6 h from the last known well, functional and safety outcomes were generally comparable between the groups, with results consistent with the primary analysis (Supplementary Table S2).

Table 2. Clinical outcomes between 2 groups (RAPID implementation period group vs. non-RAPID period group).

	RAPID implementation period group (n = 26)	Non-RAPID period group (n = 28)	P-value
Good clinical outcome (mRS 0–3 at 90 days), n (%)	13 (50.0)	13 (46.4)	0.79
Functional independence (mRS 0–2 at 90 days), n (%)	9 (34.6)	13 (39.0)	0.72
Any ICH, n (%)	8 (30.8)	9 (32.1)	1.00
Hemorrhagic transformation, n (%)	5 (19.2)	8 (28.6)	0.63
Mortality at 90 days, n (%)	4 (15.0)	2 (7.1)	0.66
Symptomatic ICH, n (%)	0 (0.0)	1 (3.6)	NA

ICH, intracerebral hemorrhage; mRS, modified Rankin Scale; NA, not applicable; RAPID, iSchemaView, Menlo Park, CA, USA

In univariate logistic regression, there was no significant difference in the odds of achieving a good clinical outcome between the RAPID implementation period group and the non-RAPID period group (OR, 1.15; 90% CI, 0.47–2.83; P = 0.79). After adjusting for NIHSS score, the odds of achieving a good clinical outcome remained comparable between the 2 groups (adjusted OR, 1.01; 90% CI, 0.38–2.65; P = 0.99) (Table 3).

Table 3. Multivariate logistic regression between groups’ outcomes (RAPID implementation period group vs. non-RAPID period group).

	Crude OR (90% CI)	Adjusted OR (90% CI)*
Good clinical outcome (mRS 0–3 at 90 days)	1.15 (0.47–2.83; P = 0.79)	1.01 (0.38–2.65; P = 0.99)
Functional independence (mRS 0–2 at 90 days)	0.82 (0.32–2.07; P = 0.72)	0.64 (0.23–1.79; P = 0.48)

^*Odds ratios were adjusted for NIHSS score.

CI, confidence interval; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; OR, odds ratio; RAPID, iSchemaView, Menlo Park, CA, USA

Imaging profiles

Among patients with anterior circulation occlusion, the median baseline ASPECTS was 8 (interquartile range [IQR], 6–9) in the RAPID implementation period group and 8 (IQR, 7–9) in the non-RAPID period group (P = 0.58). The median ischemic core volume was 10.5 mL (IQR, 4.8–55.3 mL) in the RAPID implementation period group, and the median Tmax >6 s volume was 177 mL (IQR, 107–259 mL) in the RAPID implementation period group. The distribution of occlusion sites was comparable between groups, with the MCA M1 segment being the most frequent (42.3% vs. 35.7%), followed by the ICA (26.9% vs. 25.0%) and M2 segment occlusions (15.4% vs. 21.4%). BA occlusion was less common, observed in 7.7% and 14.3% of patients, respectively.

Procedural and time metrics

Procedural efficiency and technical performance were comparable between the 2 groups. The number of device passes was identical (median, 2 [IQR, 1–3] in both groups; P = 0.47). FPE was achieved in 42.3% of patients in the RAPID implementation period group and 35.7% in the non-RAPID period group (P = 0.62); mFPE occurred in 46.2% and 39.3%, respectively (P = 0.61). The overall rate of complete or near-complete reperfusion (final eTICI ≥2c) was high and comparable (88.5% vs. 85.7%; P = 0.76). The median time from groin puncture to successful reperfusion (eTICI ≥2b50) was 35 min in the RAPID implementation period group and 36 minutes in the non-RAPID period group (P = 0.86) (Table 4).

Table 4. Mechanical thrombectomy procedures (RAPID implementation period group vs. non-RAPID period group).

	RAPID implementation period group (n = 26)	Non-RAPID period group (n = 28)	P-value
Mechanical thrombectomy strategy and techniques
First-line technique			0.07
Stent retriever, n (%)	15 (57.7)	7 (25.0)
Combined techniques, n (%)	10 (38.5)	16 (57.1)
Technique used to achieve successful recanalization			0.63
Stent retriever, n (%)	8 (32.0)	5 (18.5)
Combined techniques, n (%)	13 (52.0)	17 (63.0)
Procedural parameter
Number of passes, median (IQR)	2 (1–3)	2 (1–3)	0.47
First-pass effect*, n (%)	11 (42.3)	10 (35.7)	0.62
Modified first-pass effect†, n (%)	12 (46.2)	11 (39.3)	0.61
Complete or near-complete reperfusion (final eTICI ≥2c), n (%)	23 (88.5)	24 (85.7)	0.76
Time from groin puncture to successful reperfusion (min), median (IQR)	35 (28–55)	36 (22–51)	0.86

^*Achieving complete or near-complete reperfusion (eTICI ≥2c) after a single pass.

^†Achieving successful reperfusion (eTICI ≥2b50) after a single pass.

eTICI, expanded treatment in cerebral infarction range; IQR, interquartile range; RAPID, iSchemaView, Menlo Park, CA, USA

Discussion

In this single-center study comparing RAPID (iSchemaView)-based and non-RAPID patient selection for EVT, we found that the use of automated perfusion analysis neither delayed workflow nor compromised clinical outcomes. The RAPID implementation period achieved outcomes comparable to those in the non-RAPID period, with no signal of harm. Despite concerns that perfusion processing might prolong treatment times, the RAPID group maintained equivalent procedural efficiency and safety, with similar rates of reperfusion and favorable outcomes and a numerically lower incidence of hemorrhagic events.

Other early and late time window trials

The FRAME study⁸⁾ provides important insights into the role of perfusion imaging in the early time window (within 6 h) of AIS due to LVO. The investigation demonstrated that multimodal imaging, including perfusion maps, was associated with clinical response to EVT, thereby reinforcing the notion that tissue-based selection may refine treatment strategies even when onset-to-puncture times are short. Perfusion-based classification revealed that 71% of patients met the TMM criteria and 82% met the broader MM definition. Functional recovery (mRS 0–2 at 90 days) was achieved in 54% overall. Importantly, both TMM and MM profiles were independently associated with improved outcomes after thrombectomy, with adjusted ORs of 3.3 (95% CI, 1.4–7.9) and 5.9 (95% CI, 1.8–19.6), respectively. These findings suggest that even within the conventional early window, perfusion profiles may help identify patients with salvageable penumbra and favorable vascular and brain tissue status, beyond mere time thresholds.

FRAME supports perfusion-based triage but shows that its benefit may be modest if it delays workflow. Therefore, the key clinical implication for Japanese stroke centers is to ensure that the incorporation of perfusion imaging does not compromise door-to-puncture time. When integrated seamlessly, ideally in parallel with other imaging and procedural preparations, perfusion imaging serves as a valuable adjunct that enhances precision without sacrificing speed. This observation aligns with our own center’s experience, where the implementation of automated perfusion software did not prolong procedural metrics.

In the AURORA pooled analysis⁹⁾ of randomized trials including patients treated between 6 and 24 h from the last known well, the benefit of EVT was pronounced in those with either a clinical MM or a target perfusion MM profile (OR, 3.1 for the perfusion MM subgroup) compared with standard selection. Importantly, the treatment effect increased in the longest time tercile (12.8–24 h) among patients with favorable imaging, suggesting the presence of a “late-window paradox”¹⁰⁾—that is, patients who survive longer without infarct expansion are more likely to benefit from EVT.

In the Japanese context, where imaging workflows and regional stroke systems may differ—such as a higher prevalence of intracranial atherosclerosis, older patient age, and variability in hospital access—the AURORA findings suggest that implementing automated perfusion software (such as RAPID) could help reproduce favorable outcomes by providing objective core and penumbra quantification. Moreover, given that our own real-world cohort showed no delay in workflow or increased hemorrhagic risk with RAPID-based selection, adopting perfusion-based criteria may further enhance patient selection precision, particularly in the late-window setting, where tissue viability rather than strict clock time should guide treatment decisions. This imaging-driven approach may enable Japanese stroke centers to safely extend EVT access beyond the conventional early window, ensuring that treatment is directed toward patients with truly salvageable brain tissue.

In contrast to these randomized trials demonstrating the benefit of perfusion-based selection, our study did not show a clear improvement in functional outcomes associated with the implementation of automated perfusion imaging. This finding likely reflects several characteristics of real-world practice and our study setting. In routine clinical care, EVT eligibility is determined through an integrated assessment of clinical presentation and multimodal imaging rather than strict adherence to predefined perfusion thresholds. Moreover, as a high-volume center with continuous involvement of imaging-experienced physicians, patient selection based on conventional imaging may already achieve outcomes comparable to those supported by automated perfusion analysis. In addition, the relatively small sample size of our cohort limits the ability to detect modest outcome differences attributable to imaging strategy alone. Collectively, these factors may have constrained the incremental impact of automated perfusion imaging on clinical outcomes in this study.

Clinical implications of automated perfusion imaging

Integration of RAPID into the acute workflow was seamless, as perfusion analysis could be performed concurrently with other preparatory steps for EVT. These findings highlight the growing maturity of stroke systems in which imaging data are rapidly acquired, processed, and interpreted through standardized protocols. By reducing cognitive burden and interobserver variability in image interpretation, automated perfusion tools facilitate consistent and reproducible patient triage across operators and time periods. In a setting where EVT eligibility can already be determined with a high degree of accuracy through conventional imaging and expert clinical judgment, the absence of a clear outcome advantage with RAPID underscores a different aspect of its clinical value. Rather than replacing expert decision-making, automated perfusion imaging provides an objective and standardized framework that may help reproduce expert-level patient selection without requiring advanced imaging expertise. If such reproducibility can be achieved, platforms such as RAPID may enable non-high-volume centers to apply patient selection strategies comparable to those of comprehensive stroke centers, thereby contributing to the standardization and more equitable dissemination of EVT. Based on our results, a prospective multicenter registry study is warranted to further evaluate the clinical impact and optimize the use of automated perfusion imaging in diverse clinical settings.

Limitations

This study has several limitations. First, it was a retrospective analysis conducted at a single comprehensive stroke center, and the relatively small sample size limits both the generalizability of the findings and the statistical power to detect modest differences between imaging strategies. Second, patient allocation to the RAPID or non-RAPID period was not randomized but depended on the availability of imaging software at the time of evaluation, introducing potential selection bias toward patients with better imaging profiles. Third, as discussed above, the study was conducted under optimized institutional conditions with experienced neurointerventional teams, and the results may not fully reflect performance in smaller or resource-limited centers. In addition, imaging parameters and workflow metrics were obtained from electronic medical records and may have been influenced by unmeasured confounders. Finally, the study did not assess long-term outcomes beyond 90 days or evaluate cost-effectiveness, both of which warrant further investigation in prospective multicenter studies.

Conclusion

In summary, the implementation of RAPID (iSchemaView) was associated with comparable clinical and safety outcomes to those observed during the non-RAPID period. RAPID implementation may contribute to more structured and transparent patient selection by providing objective perfusion information, thereby supporting consistent decision-making without compromising workflow or safety.

Supplementary Information

Table S1

Implementation of endovascular therapy among patients with large or medium vessel occlusion within 24 hours from onset or last-known-well

Table S2

Sensitivity analysis restricted to patients treated ≥6 hours from onset or last-known-well

Funding Statement

This work was supported by a Grant-in-Aid for Scientific Research (KAKENHI) from the Japan Society for the Promotion of Science (JSPS) (Grant Number JP 25K15950).

Disclosure Statement

All of the following conflicts are outside the submitted work. Nobuyuki Sakai reports honoraria from Kaneka, Medtronic, and Terumo, and has received research funding from Century Medical, Kaneka, Penumbra, and Toro Neurovascular. Yusuke Yakushiji reports honoraria from Daiichi Sankyo and Eisai. Masatoshi Koga reports honoraria from Daiichi Sankyo and has received research funding from Boston Scientific and Daiichi Sankyo. All other authors declare no conflicts of interest.