A regional stroke center in California has cut the time it takes to identify and transfer large-vessel-occlusion (LVO) stroke patients by 44%, according to new data presented at the American Heart Association’s International Stroke Conference (ISC) in March 2026. The improvement, driven by AI stroke-detection software from Viz.ai, moved the hospital’s average door-in-door-out (DIDO) time from 202 minutes down to 113 minutes — beating the Joint Commission’s 120-minute national benchmark.
The Study Behind the Numbers
The findings, presented in a study titled “Bridging the Gap: Leveraging AI Technology & Partnerships to Meet DIDO Benchmarks for LVO Patients,” came from Adventist Health + Rideout, a regional primary stroke center. Caezar G. Jara, the hospital’s stroke program manager, presented the results at ISC 2026 on March 5. “Our initiative was driven by the need to eliminate manual bottlenecks that delay stroke care in regional settings,” Jara said, according to Viz.ai’s account of the presentation.
How the AI Speeds Up Detection
Door-in-door-out time measures how long a stroke patient spends at a community hospital before being transferred to a comprehensive stroke center capable of performing endovascular thrombectomy — the mechanical clot-removal procedure that is often the only effective treatment for a large-vessel occlusion. Every minute of delay matters enormously in this window: acute ischemic stroke is associated with the loss of roughly 1.9 million neurons per minute of continued occlusion, a figure Viz.ai cites in explaining why speeding up the diagnostic chain has such outsized clinical stakes. At Adventist Health + Rideout, the AI software helped cut the time from CT angiography completion to LVO detection by 84%, and reduced the time to notify the receiving care team from 45 minutes down to just 7 minutes.
A Broader Push Across the Stroke-Care Network
The Rideout results were one of two studies Viz.ai highlighted at ISC 2026 examining its stroke platform’s effect on patient outcomes and hospital economics. The company also used the conference to introduce Viz Assist, a tool aimed specifically at regional and community hospitals that integrates real-time patient context, automates care-team activation, and standardizes transfer workflows across hub-and-spoke hospital networks — the referral arrangement in which smaller community hospitals route complex stroke cases to larger comprehensive centers. Viz.ai now holds more than 50 FDA clearances and says its stroke and related detection software is deployed in more than 1,700 hospitals nationwide.
The Skeptical View
Independent clinicians and health-IT analysts covering the broader AI-in-radiology and AI-in-emergency-medicine space have cautioned that results like these, while measurable, are typically presented by the vendor or in vendor-sponsored studies rather than fully independent, multicenter randomized trials — a distinction that matters when hospitals are deciding whether to invest in a specific platform versus a broader quality-improvement effort that might achieve similar gains without new software. Broader industry reporting on hospital AI adoption in 2026 has also flagged that enterprise-level deployment does not guarantee frontline clinical uptake: across radiology broadly, only about 30% of practicing radiologists report actually using AI day to day, even as formal hospital-level adoption climbs. Emergency physicians and nurses, who must act on these alerts in real time, have likewise been described in recent scoping reviews as more skeptical of AI triage tools than the administrators who purchase them, in part over concerns about alert fatigue and how AI-flagged priorities interact with clinical judgment.
What It Means for Stroke Networks Going Forward
Stroke remains one of the clearest use cases for AI in acute care because the intervention window is short and unambiguous — get the patient to a thrombectomy-capable center faster, and outcomes improve. The Rideout results suggest that even a single regional hospital, without becoming a full comprehensive stroke center itself, can meaningfully close the gap between a patient walking in the door and receiving definitive treatment elsewhere. As more hub-and-spoke hospital networks adopt platforms like Viz Assist, the next test will be whether these DIDO-time reductions translate into measurably better 90-day functional outcomes across larger, more diverse patient populations — and whether smaller hospitals without dedicated stroke program managers like Jara can replicate the gains without the same institutional investment in workflow redesign that Rideout put in alongside the software itself. Viz.ai’s push to extend Viz Assist across more hub-and-spoke referral networks over the coming year will be an early test of whether these kinds of results are reproducible outside a single well-resourced regional program, or whether they depend as much on institutional commitment and staff training as on the underlying detection algorithm.
