StrokeSENS LVO

StrokeSENS LVO is a radiological computer-aided triage and notification software that employs a machine learning algorithm to analyze CT Angiography head images for the presence of Large Vessel Occlusion (LVO). Its purpose is to expedite clinical workflow by providing early notification of suspected LVO cases to clinicians, thereby facilitating prioritized review and potentially accelerating time-sensitive stroke interventions.

StrokeSENS LVO is a medical image processing software designed to assist hospital networks and trained clinicians in workflow triage by flagging and communicating suspected positive findings of Large Vessel Occlusion (LVO) in head CT Angiography (CTA) images. The system utilizes a machine learning-based software algorithm to analyze DICOM-compliant CTA scans. This algorithm acts as a binary classifier, identifying characteristics suggestive of LVO. If a suspected LVO is detected, StrokeSENS LVO sends a notification to pre-configured destinations, which can include a compressed preview of images for informational purposes only, prompting clinicians to review the full images on a diagnostic radiological viewer. The system operates in parallel to the standard of care workflow and does not alter the original medical images, nor is it intended for direct diagnostic use beyond notification. Performance validation of StrokeSENS LVO was conducted through a retrospective case study involving 400 independent studies (217 LVO positive, 183 LVO negative), which included a heterogenous dataset representing varied patient demographics and imaging parameters. The device achieved a mean sensitivity of 89.4% (95% CI [85.3%, 93.5%]) and a mean specificity of 87.4% (95% CI [82.6%, 92.2%]) for LVO detection. The average time for the system to process a CTA image and generate a notification for LVO positive cases was 0.75 minutes (SD ±0.17 mins), meeting the target of less than 5 minutes. Software verification and validation adhered to recognized standards including ISO 13485, IEC 62304, IEC 62366, ISO 14971, and NEMA guidelines for DICOM conformance.

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 StrokeSENS LVO algorithm's performance (sensitivity and specificity) vary with different CTA acquisition parameters, scanner manufacturers, or in the presence of motion artifacts?; What is the real-world clinical impact on door-to-needle or door-to-puncture times when StrokeSENS LVO is integrated into a hospital's stroke protocol compared to standard care?; What specific image features has the machine learning algorithm been trained to identify as indicative of LVO, and is there a mechanism to review or understand the basis for a specific positive or negative classification?; How does the notification system integrate with existing hospital IT infrastructure, such as PACS and EHR systems, beyond simple email notifications, and what are the safeguards against notification fatigue?; What is the performance of StrokeSENS LVO on different subtypes of LVO (e.g., M1, M2, ICA terminus occlusions) and in patient populations with varying comorbidities or prior cerebrovascular events?