Brainomix 360 Triage Stroke

Brainomix 360 Triage Stroke is an AI-powered software designed to expedite stroke care by rapidly analyzing non-contrast head CT scans to identify suspected large vessel occlusions and intracranial hemorrhages for workflow prioritization. It utilizes machine learning algorithms, including ASPECTS and hyperdense vessel sign analysis for LVO and hyperdensity detection for ICH, to provide timely notifications to clinicians, thereby aiming to reduce delays in critical stroke interventions.

Brainomix 360 Triage Stroke is a radiological computer-aided triage and notification software designed to analyze non-contrast head CT (NCCT) images from adult patients presenting with acute stroke symptoms within 24 hours of onset or last known well time. Its primary purpose is to assist hospital networks and trained clinicians in workflow triage by rapidly flagging and communicating suspected positive findings for large vessel occlusion (LVO) of the intracranial ICA and M1, and intracranial hemorrhage (ICH). The software operates in parallel with standard care image interpretation, using an artificial intelligence algorithm to analyze images and highlight critical cases on a server, which can be on-premise or cloud-based. It is intended as an adjunctive tool for prioritizing medical images and does not replace primary diagnostic procedures or alter the original medical images; clinicians remain responsible for reviewing full images. The software employs machine learning, advanced non-adaptive imaging algorithms, and large data analytics. For ICH detection, it is optimized to identify hyperdense volumes in the parenchyma typically associated with acute hemorrhage. For LVO suspicion on NCCT images, it utilizes a combined analysis of ASPECTS (Alberta Stroke Program Early CT Score) and hyperdense vessel sign (HDVS) algorithms. Clinicians receive notifications for suspected LVO or ICH findings via a web user interface or a mobile application, which includes compressed preview images for informational purposes only and not for diagnostic use. Performance data from a retrospective study of 267 cases indicated an ICH detection sensitivity of 92.5% and specificity of 87.22%. For NCCT LVO, the software demonstrated a sensitivity of 68.75% and specificity of 89.57%. A reader study further showed that the device's LVO detection sensitivity on NCCT was higher than that of radiologists. The typical time-to-notification was reported to be between 62 and 134 seconds.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the LVO detection algorithm's performance (sensitivity of 68.75% on NCCT) impact the overall stroke workflow, particularly regarding the decision to proceed to CTA, given that the device is not a rule-out tool?; What are the specific performance characteristics (sensitivity, specificity) for detecting LVO in different M1 segments (proximal vs. distal) or ICA locations using the NCCT-based algorithm?; The document mentions the ICH algorithm is optimized to identify hyperdense volume; how does it differentiate ICH from other hyperdense mimics on NCCT, such as calcifications or contrast extravasation if prior contrast was administered for other reasons?; Given the device's validation on Siemens, GE, and Philips scanners, are there known variations in performance across these manufacturers, and what is the strategy for validating on scanners from other manufacturers or newer models not included in the initial validation?; The reader study indicated superior LVO detection sensitivity for the device compared to radiologists on NCCT; what were the specific experience levels of the 'expert' neuroradiologists in that study, and what factors might contribute to the device's higher sensitivity in this context?