Radiology · Stroke neuroimaging & triage platform
Rapid
iSchemaView, Inc. (d/b/a RapidAI)
Rapid is the perfusion-imaging engine behind modern late-window stroke care: the automated core-and-penumbra measurement it produces was the mandatory selection tool in the DAWN and DEFUSE 3 trials, whose results pushed the thrombectomy treatment window out to 24 hours and were written into the 2018 AHA/ASA guidelines. On the strength of that paradigm the platform reached more than 2,500 hospitals across over 100 countries and now carries a wall of FDA clearances plus regulatory authorisation in Canada, Japan and Australia. The reservation that keeps it at two marks is a real one: the randomised evidence validates the imaging-selection strategy, not the diagnostic accuracy of the software itself, whose direct validation is largely retrospective and manufacturer-affiliated and sits alongside a substantial independent literature on perfusion core overestimation.
Performance Metrics
Clinical Evidence
Rapid's evidence base has two distinct halves, and keeping them apart matters for reading it honestly. The first half is the landmark thrombectomy and thrombolysis trials in which Rapid was the imaging instrument used to select patients — not the intervention under test. In DEFUSE 3 (New England Journal of Medicine, 2018; 182 patients) Rapid CT-perfusion or MRI defined a target mismatch of ischaemic core under 70 mL with a mismatch ratio of at least 1.8, and thrombectomy in the 6-to-16-hour window produced functional independence at 90 days in 45 percent of patients versus 17 percent with medical management. DAWN (NEJM, 2018; 206 patients) used Rapid-defined clinical-imaging mismatch to extend the window to 24 hours, with functional independence of 49 percent versus 13 percent. EXTEND (NEJM, 2019; 225 patients) used Rapid perfusion mismatch to select patients for extended-window intravenous thrombolysis, with a modest benefit (35.4 versus 29.5 percent) offset by more symptomatic haemorrhage. These are the results the 2018 and 2019 AHA/ASA guidelines translated into Class I recommendations for imaging-based late-window selection — which is what made automated perfusion analysis, and Rapid as its most-studied implementation, close to standard of care. Crucially, this evidence supports the selection paradigm; it is not a randomised test of the software's own accuracy. The second half is the diagnostic-accuracy literature on individual modules, which is thinner and skews toward manufacturer-affiliated authorship. Rapid LVO was reported at 97 percent sensitivity for intracranial internal-carotid and M1 occlusions (area under the curve 0.997 and 0.962) but 90 percent for more distal M2 occlusions, with overall specificity around 74 percent (Amukotuwa et al., Stroke, 2019) — an author group including iSchemaView staff. Rapid ICH reported 95.6 percent sensitivity and 95.3 percent specificity against a three-neuroradiologist reference (Heit et al., American Journal of Neuroradiology, 2021), again with vendor ties. Set against those figures is a genuine independent cautionary literature that editorial neutrality requires carrying. Perfusion-based core estimation overestimates the true infarct in a meaningful minority of early-window or poor-collateral patients — the "ghost infarct core" phenomenon — reported in 21 percent of reperfused patients with a median overestimation of 12 mL (García-Tornel et al., Stroke, 2021) and across a 6-to-58 percent range in a systematic review (Ballout et al., Journal of Neuroimaging, 2023). A reanalysis of the EXTEND imaging found that only 87 percent of Rapid-eligible patients still met mismatch criteria when the same scans were reprocessed with a different perfusion package (Bivard et al., 2022), a direct demonstration that these tools are not interchangeable and that a selection decision can flip with the software. Automated ASPECTS scoring likewise agrees only moderately with expert readers and is positioned as an aid rather than a replacement.
| Study | Design | n | Sensitivity | Specificity | AUC | Published |
|---|---|---|---|---|---|---|
| Albers GW, Marks MP, Kemp S, et al. (DEFUSE 3 Investigators) | RCTRCT | 182 | — | — | — | New England Journal of Medicine, Feb 2018; Rapid CTP/MRI defined target mismatch (core <70 mL, ratio ≥1.8); thrombectomy 6–16 h gave 90-day functional independence 45% vs 17% for medical management; mortality 14% vs 26% |
| Nogueira RG, Jadhav AP, Haussen DC, et al. (DAWN Investigators) | RCTRCT | 206 | — | — | — | New England Journal of Medicine, Jan 2018; Rapid-defined clinical-imaging mismatch selected patients 6–24 h; 90-day functional independence 49% vs 13%; stopped early at interim analysis |
| Ma H, Campbell BCV, Parsons MW, et al. (EXTEND Investigators) | RCTRCT | 225 | — | — | — | New England Journal of Medicine, May 2019; Rapid perfusion mismatch selected extended-window (4.5–9 h + wake-up) IV thrombolysis; primary good outcome 35.4% vs 29.5% (adjusted RR 1.44); symptomatic ICH 6.2% vs 0.9% |
| Amukotuwa SA, Straka M, Smith H, et al. (includes iSchemaView staff) | RetrospectiveRetrospective | 926 | 97% | 74% | — | Stroke, Dec 2019 (50:3431-3438); Rapid LVO on CTA — sensitivity 97% for intracranial ICA and M1 (AUC 0.997 and 0.962), 90% for M2 (AUC 0.874); overall specificity ~74%; manufacturer-affiliated authorship |
| Heit JJ, Coelho H, Lima FO, et al. (manufacturer-affiliated) | RetrospectiveRetrospective | 308 | 95.6% | 95.3% | — | American Journal of Neuroradiology, Feb 2021 (42:273-278); Rapid ICH vs 3-neuroradiologist consensus; volumetric correlation r=0.98; vendor-affiliated authorship |
| García-Tornel A, Campos D, Rubiera M, et al. (independent — cautionary) | RetrospectiveRetrospective | 393 | — | — | — | Stroke, May 2021 (52:1751-1760); perfusion core overestimation in 21% of reperfused anterior-circulation patients (median 12 mL), associated with poor collaterals and early imaging — a limit on perfusion-based selection |
Clinical Pulse
Verified clinician sentiment
Verified clinician reviews launching soon · Apply to contribute →
Reviews are signed by clinicians verified against their regulator's public register.
Inside the Auris+ Listing
Five more sections complete this device’s Auris+ Listing.
Decision Ledger
ProPro unlocks a private, cross-vendor log of every case you read on this device — what the AI called, what you concluded, and a one-line reason if you overrode it. Ready for the EU AI Act's deployer logging obligations when they land in 2028.
Clinical Evidence Deep Dive
ProPro unlocks the structured clinical-evidence summary — study count, target patient population, and a tabular accuracy-metrics view drawn from peer-reviewed sources.
Peer-Reviewed Publications
ProPro unlocks the curated peer-reviewed publication list with PubMed cross-links — the citation backbone of every editorial verdict.
Post-Market & Regulatory Conditions
ProPro unlocks the post-market surveillance summary, recall record, and the conditions of approval that bound real-world use.
AI Algorithm Version History
ProPro unlocks the chronological record of algorithm version changes — what changed when, drawn from manufacturer changelogs and regulatory filings.
Regulatory Approvals
Safety Record
No FDA recall, field safety notice or safety communication for iSchemaView / RapidAI was located in any source accessible to this review. A direct MAUDE adverse-event query and a direct FDA recall-database query could not be run this cycle because register egress was blocked; the null finding therefore rests on secondary-source and search-index review as of 31 July 2026 and should be re-run against the registers directly. Care is warranted because "Rapid" is a highly generic device name that collides with many unrelated products in adverse-event databases — only events attributed to iSchemaView / RapidAI as manufacturer would be relevant. The clinical risk surface is the quality of the imaging measure the software produces: as analysis software layered on acquired imaging, it does not touch the patient or deliver therapy, so the failure mode that matters is a perfusion or triage output that misleads a transfer or treatment decision — precisely the concern the independent core-overestimation literature documents.
Intended Use & Indications
Rapid is a family of image-processing and clinical-decision-support modules for acute stroke. The perfusion engine (Rapid CTP / Rapid MR) generates colour maps and automated estimates of ischaemic core and penumbra volumes from CT-perfusion or diffusion/perfusion MRI, the measurement used to identify a target mismatch. The computer-aided triage modules — Rapid LVO, Rapid ICH, Rapid SDH, Rapid Aneurysm, Rapid NCCT Stroke, Rapid PE — analyse imaging in parallel to the standard reading workflow and, when an algorithm threshold is met, send a notification and preview to a designated specialist so that transfer and treatment decisions can begin sooner. The triage modules are notification-only: they do not remove the study from the radiologist's queue, mark the diagnostic image, or provide a diagnosis, and the treating clinician remains responsible for interpretation and treatment. Automated ASPECTS scoring (Rapid ASPECTS) is a computer-aided-diagnosis output intended to support, not replace, the reviewing physician. The software is a decision aid layered on imaging already acquired; it does not deliver therapy or steer hardware.