Brainomix – A Help or a Hindrance in our developing Stroke Services in South Wales?

Introduction

In 2024 we conducted a service evaluation of our new stroke AI in Wales, Brainomix. We reviewed its sensitivity and specificity at identifying large vessel occlusions (LVO’s) on CT Angiograms (CTA’s) for our stroke patients. The purpose of Brainomix being to assist in rapid detection of LVO’s and support decision making around proceeding to mechanical thrombectomy for appropriate individuals. Multiple studies in recent years have demonstrated improved functional outcomes for patients receiving mechanical thrombectomy. Brainomix was hoped to be pivotal in bringing these improvements to Wales.

Methods

We reviewed the results of all CTA’s sent to Brainomix April-June 2026 inclusive in Cardiff and Vale (CAV), Morriston and Cwm Taf Health Boards. We compared these results to the reports given by our Neuroradiologists, using this as our gold standard. Patients were excluded if they were < 18 years old or they were identified as having posterior circulation LVO’s, as these are not yet detectable via the software. 

Results 

In total 306 CTA scans were sent through to Brainomix. 12 scans were excluded. In CAV 22.4% of patients were reported by Neuroradiologists as having LVO’s. In Morriston this was 37.5% and in Cwm Taf only 13.9%. Brainomix’s sensitivity at detecting LVO’s ranged from 92.8% in CAV to 60% in Morriston, which compares to 85% calculated at its initiation in 2024 in CAV. Reasons for false negatives were mainly occlusions affecting M2 branches of MCAs. Specificity was higher across the 3 Health Boards, ranging from 84.7% in Cwm Taf to 96% in Morriston.  

Conclusion

These results show there certainly are still deficits in Brainomix detecting both posterior circulation and more distal anterior circulation clots. Though Brainomix has advantages and functions as a useful adjunct for stroke physicians, the gold standard Neuroradiologist reporting has not yet been surpassed.