Viz SDH

Viz SDH employs an artificial intelligence algorithm to analyze non-contrast CT scans of the head for the prompt detection of suspected subdural hemorrhages. The system serves as a notification-only, parallel workflow tool that alerts specialists through a mobile application, aiming to facilitate quicker image review and streamline triage processes, supported by clinical data demonstrating high sensitivity and specificity.

Viz SDH is a software-based, notification-only tool designed to augment clinical workflow in hospital environments for patients with suspected subdural hemorrhage (SDH). It utilizes an artificial intelligence algorithm to analyze non-contrast CT (NCCT) images of the head, searching for indicators of SDH. Upon identifying a potential finding, the system dispatches alerts to designated specialists, such as neurovascular or neurosurgical experts, via a mobile application. This application permits a preview of the images; however, these are compressed and intended solely for informational notification, not for primary diagnostic interpretation. Clinicians are explicitly instructed to review the original, non-compressed images on a certified diagnostic viewer before making any patient care decisions, positioning Viz SDH as an assistive technology within the standard care pathway. The core technology of Viz SDH is its AI machine learning algorithm, which processes NCCT scans uploaded to Viz.ai's servers to detect possible SDH occurrences. Clinical performance evaluation using 542 NCCT scans showed strong efficacy, achieving 94% sensitivity and 92% specificity in SDH detection, with an Area Under the Curve (AUC) of 0.96. The system demonstrated rapid notification capabilities, alerting specialists in an average of approximately 1.15 minutes. Comprehensive performance data was stratified by various factors, including clinical site, patient demographics (age, gender), hemorrhage characteristics (subtype, thickness, volume, location), and imaging parameters (slice thickness, scanner type, reconstruction method), generally indicating robust performance, though sensitivity was notably lower for posterior fossa SDH (33%) and SDH volumes under 1 mL (33%).

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 AI algorithm differentiate between acute, non-acute (chronic), and mixed subdural hemorrhages, and does its performance vary significantly across these subtypes beyond the provided sensitivity figures?; Given the reported lower sensitivity (33%) for posterior fossa subdural hemorrhages, what specific imaging characteristics or algorithmic limitations contribute to this, and are there strategies to improve detection in this critical anatomical region?; What is the clinical impact of the device's 33% sensitivity for subdural hemorrhage volumes less than 1 mL, and how does this detection threshold balance against potential false positive rates for very small findings?; How does the Viz SDH system integrate with diverse hospital Picture Archiving and Communication Systems (PACS) and Electronic Health Record (EHR) systems to ensure efficient data transfer and timely notifications within established clinical IT infrastructures?; Beyond faster notification times, what evidence exists from real-world clinical usage regarding the impact of Viz SDH on measurable patient outcomes or definitive time-to-intervention for subdural hemorrhage cases?