Medical News Observer

Key points

• Researchers evaluated Brainomix e-Stroke software on 1163 examinations on 551 patients.
• Varied accuracy rates: 77% for detecting ischemia, 69.1% for thrombus, and 97.8% for hemorrhage.
• The study concludes that e-Stroke should be used alongside expert analysis, aligning with the manufacturer’s guidance.

In a recent study, researchers conducted an in-depth evaluation of Brainomix e-Stroke, an AI-powered tool designed to diagnose acute stroke. This evaluation occurred at the University College London Hospital’s Hyperacute Stroke Unit. Researchers analyzed patients admitted from October 2021 to April 2022.

The main objective was to assess the effectiveness of Brainomix e-Stroke in interpreting CT images for stroke diagnosis. To achieve this, the tool’s analyses were compared against authoritative diagnoses made by a neuroradiologist equipped with comprehensive clinical and imaging data. Additionally, the study compared the measurements of core infarct and ischemic penumbra volumes made by the Brainomix e-Stroke software with those obtained from syngo.via.

Brainomix e-Stroke Software’s Effectiveness

The study covered 1163 examinations on 551 patients, with Brainomix e-Stroke successfully processing 97.2% of these within an average of 4 minutes. The accuracy of e-Stroke varied across different diagnostic aspects: it identified acute middle cerebral artery territory ischemia with 77.0% accuracy, detected hyperdense thrombus at a 69.1% rate, and showed a notable 97.8% accuracy in identifying acute hemorrhage. The tool was also 91.5% accurate in detecting large vessel occlusions.

While the results were largely positive, the researchers concluded that Brainomix e-Stroke should be used in conjunction with expert interpretation, as recommended by the manufacturer, rather than as a standalone decision-making tool. This conclusion underscores the importance of combining AI technology with human expertise in medical diagnosis.

Reference

Mallon, Dermot, Matthew Fallon, Eirini Blana, Cillian McNamara, Arathi Menon, Chak Lam Ip, Jack Garnham, et al. 2023. “Real-World Evaluation of Brainomix e-Stroke Software.” Stroke and Vascular Neurology, December, svn-2023-002859. https://doi.org/10.1136/svn-2023-002859

Also, read about AI/ML: Brainomix 360 Triage ICH