Brainomix 360 Triage LVO

Brainomix 360 Triage LVO is an artificial intelligence-powered software that analyzes CT angiogram images for the rapid detection of suspected large vessel occlusions (LVOs) in the brain's ICA and M1 arteries. Its methodology relies on machine learning algorithms, trained on diverse radiological data and validated for high sensitivity and specificity, to provide timely notifications to clinicians, thereby aiming to accelerate triage in acute stroke care.

Brainomix 360 Triage LVO is a radiological computer-aided triage and notification software designed to expedite the identification of suspected Large Vessel Occlusions (LVOs) in acute stroke patients. It functions as a stand-alone, notification-only tool operating in parallel with standard clinical workflows, analyzing CT Angiogram (CTA) images of the brain to detect occlusions specifically within the intracranial internal carotid artery (ICA) and the M1 segment of the middle cerebral artery. The system employs advanced machine learning algorithms and artificial intelligence, trained on a dataset of over 1600 CTA studies from diverse scanner manufacturers, to automatically identify potential LVOs. Upon detection, it sends a priority notification to clinicians, enabling them to preview compressed images via a mobile app or view unprocessed images through a web interface, thereby assisting in case prioritization for specialist review without replacing standard diagnostic procedures. The performance of Brainomix 360 Triage LVO was validated through a retrospective study on an independent dataset of 308 CTA scans, with ground truth established by board-certified neuroradiologists. The device demonstrated high accuracy, achieving a sensitivity of 90% and a specificity of 92.9%, with an Area Under the Curve (AUC) for the Receiver Operating Characteristic (ROC) of 91.4%. Performance was further stratified across various demographic and technical variables, including patient age, gender, race, scanner manufacturer, clinical site, affected vessel, and the presence of stenosis, generally maintaining robust results. The system also proved efficient, with a time-to-notification ranging from 86.3 to 178.2 seconds, facilitating rapid communication to specialists and supporting its role as an adjunctive tool for triaging potential LVO cases.

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 algorithm's performance vary with different contrast phases (e.g., Early Arterial, Peak Arterial, Equilibrium) encountered in the validation dataset, and were specific phases prioritized or weighted differently during algorithm training?; Given the observed difference in sensitivity for detecting intracranial carotid artery (ICA) occlusions (96.6%) versus proximal middle cerebral artery (MCA M1) occlusions (88.7%), what are the potential algorithmic or anatomical factors contributing to this variation, and are there targeted strategies to enhance MCA M1 detection?; The device showed a notably lower sensitivity (70.0%) in patients with stenosis compared to those without (91.5%). What are the specific imaging characteristics associated with stenosis that challenge the AI algorithm, and how should clinicians interpret notifications in patients with known or suspected stenosis?; What are the primary factors contributing to the variability in the time-to-notification (ranging from 86.3 to 178.2 seconds), and are there identifiable bottlenecks, such as PACS integration or data transfer speeds across different hospital infrastructures, that could be targeted for further optimization?; Beyond scanner manufacturer and basic acquisition parameters (slice thickness, KvP), how does the algorithm's robustness hold against variations in CT Angiogram protocols, such as differing contrast media injection rates or volumes, or the use of more varied reconstruction kernels not explicitly detailed in the performance stratification?