Introduction: The Critical Challenge of Acute Stroke Management
In the management of acute ischemic stroke, the adage “time is brain” remains the fundamental guiding principle. Every minute of delay in restoring cerebral blood flow results in the loss of approximately 1.9 million neurons. While endovascular thrombectomy (EVT) has emerged as the definitive standard of care for patients with large vessel occlusion (LVO), the clinical pathway from symptom onset to arterial puncture is fraught with logistical complexities. Identifying eligible patients requires rapid, high-fidelity neuroimaging interpretation—a task that often demands specialized neuroradiological expertise that may not be available 24/7 at all primary stroke centers.
Artificial intelligence (AI) imaging decision support systems have been proposed as a solution to this bottleneck. These technologies aim to automate the detection of LVO and quantify ischemic damage using scores like the Alberta Stroke Program Early CT Score (ASPECTS), potentially accelerating the triage and transfer process. However, until recently, large-scale, real-world evidence of their impact on treatment rates has been limited. A landmark prospective observational study published in Lancet Digital Health by Nagaratnam and colleagues provides compelling evidence that AI implementation can fundamentally alter the landscape of stroke care.
Background and the Need for Scalable Solutions
Endovascular thrombectomy significantly improves functional outcomes in patients with LVO, but its delivery is highly dependent on the efficiency of the stroke network. In England, the National Health Service (NHS) operates a hub-and-spoke model, where patients often present to a primary stroke center (PSC) before being transferred to a comprehensive stroke center (CSC) for intervention. Delays in this “drip-and-ship” model are common, often stemming from delays in imaging acquisition, interpretation, and communication between sites.
AI software, such as the Brainomix 360 Stroke platform, provides automated analysis of non-contrast CT, CT angiography, and CT perfusion scans. By providing near-instantaneous alerts to clinicians and facilitating image sharing via mobile applications, these tools are designed to democratize expert-level imaging interpretation. The study in question sought to evaluate whether the systematic rollout of such software actually leads to more patients receiving life-saving thrombectomy.
Study Design and Methodology
This prospective observational study utilized data from the Sentinel Stroke National Audit Programme (SSNAP), the national stroke registry for England. The researchers analyzed data from all 107 NHS hospitals admitting acute stroke patients between January 1, 2019, and December 31, 2023. This comprehensive dataset included 452,952 patients with a primary diagnosis of stroke.
The intervention focused on 26 evaluation sites (six CSCs and 20 PSCs) where the Brainomix 360 Stroke software was systematically implemented. The primary outcome was the proportion of stroke patients receiving EVT. The researchers compared changes in EVT rates at these evaluation sites before and after implementation, using non-evaluation sites as a contemporary control group to account for national trends in stroke care improvements.
At the patient level, the study also compared outcomes for individuals whose images were reviewed with the assistance of AI versus those whose images were reviewed through standard care. This granular approach allowed for a robust assessment of the software’s direct association with clinical decision-making.
Key Findings: A Doubling of Treatment Rates
The study results demonstrate a profound shift in clinical practice following the introduction of AI imaging support. At the 26 evaluation sites, the pre-implementation EVT rate was 2.3% (376 of 15,969 patients). Following the implementation of the AI software, this rate rose to 4.6% (751 of 15,428 patients), representing a 100% relative increase.
While EVT rates also increased at non-evaluation sites—reflecting a general national trend toward more aggressive stroke intervention—the magnitude of the change was significantly smaller. Non-evaluation sites saw an increase from 1.6% to 2.6% (a 62.5% relative increase). The statistical interaction between site type and time period was significant (odds ratio [OR] 1.24; 95% CI 1.08–1.43; p=0.0026), suggesting that the AI implementation provided a benefit over and above the general improvement in national stroke services.
At the individual patient level, the data was even more striking. Use of the AI stroke software was associated with a significantly increased likelihood of receiving endovascular thrombectomy, with an odds ratio of 1.57 (95% CI 1.33–1.86; p<0.0001). This suggests that the software helps clinicians identify eligible candidates who might have otherwise been missed or deemed ineligible due to perceived delays or subtle imaging findings.
Mechanistic Insights: How AI Facilitates Intervention
The increase in EVT rates can be attributed to several factors facilitated by the AI platform:
1. Standardized Image Interpretation
Automated ASPECTS scoring reduces inter-observer variability. In high-pressure emergency settings, a standardized score can provide clinicians with the confidence to proceed with transfer or treatment, particularly when a specialist neuroradiologist is not immediately available.
2. Rapid LVO Detection
AI algorithms can identify large vessel occlusions on CT angiography within minutes. By alerting the stroke team via mobile notifications, the software bypasses traditional pager systems and phone trees, significantly shortening the time to decision.
3. Enhanced Communication Networks
Platforms like Brainomix allow clinicians at both the referring PSC and the receiving CSC to view the same images and AI analyses simultaneously. This shared mental model streamlines the referral process, reduces redundant imaging, and accelerates the “drip-and-ship” pathway.
Expert Commentary and Clinical Implications
The findings of Nagaratnam et al. underscore a pivotal shift in the digital transformation of healthcare. The 100% relative increase in EVT rates at evaluation sites suggests that AI is not merely an incremental improvement but a catalyst for systemic change. For health policy experts, these data provide a strong economic and clinical justification for the routine integration of AI in stroke networks.
However, it is important to interpret these results within the context of an observational study. While the researchers controlled for several variables, the implementation of AI often coincides with other service improvements, such as increased staffing or better training. Nevertheless, the patient-level association (OR 1.57) strongly supports the hypothesis that the software itself is a primary driver of the observed benefits.
Critically, the study highlights that AI does not replace the clinician. Instead, it serves as a decision-support tool that augments human expertise, ensuring that the right patients reach the angiosuite in the shortest possible time. The broader implication is that AI can help mitigate geographical disparities in stroke care, ensuring that patients presenting to smaller, rural hospitals have access to the same diagnostic precision as those at major academic centers.
Conclusion: A New Standard for Stroke Networks
The prospective evidence from the English NHS indicates that AI imaging software is associated with a significant increase in the delivery of endovascular thrombectomy. By doubling the relative rate of intervention at evaluation sites, this technology addresses a critical unmet need in acute stroke care: the rapid and accurate identification of patients who stand to benefit most from mechanical reperfusion.
As stroke services globally continue to evolve, the routine use of AI imaging decision support appears increasingly essential. Future research should focus on the long-term functional outcomes of these additional patients treated and the cost-effectiveness of widespread AI implementation. For now, the message is clear: AI is a powerful ally in the race against the clock in stroke medicine.
Funding and Acknowledgments
This study was funded by the AI in Health and Care Award from the Accelerated Access Collaborative within NHS England. The researchers utilized data from the Sentinel Stroke National Audit Programme (SSNAP), which is commissioned by the Healthcare Quality Improvement Partnership (HQIP).
References
1. Nagaratnam K, Neuhaus AA, Fensome L, et al. Artificial intelligence imaging decision support for acute stroke treatment in England: a prospective observational study. Lancet Digit Health. 2026; doi:10.1016/j.landig.2025.100927.
2. Goyal M, Menon BK, van Zwam WH, et al. Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials. Lancet. 2016;387(10029):1723-1731.
3. Powers WJ, Rabinstein AA, Ackerson T, et al. Guidelines for the Early Management of Patients With Acute Ischemic Stroke: 2019 Update to the 2018 Guidelines for the Early Management of Acute Ischemic Stroke. Stroke. 2019;50(12):e344-e418.