uAI Easy Triage ICH

The uAI Easy Triage ICH is an artificial intelligence-powered software designed to analyze non-contrast head CT images for the presence of Intracranial Hemorrhage, aiming to improve radiology workflow efficiency. By employing convolutional neural networks, the device processes DICOM images to flag suspicious cases for prioritized review, with performance validated through studies demonstrating high sensitivity and specificity.

The uAI Easy Triage ICH is a radiological computer-assisted triage and notification software designed to analyze non-enhanced head CT images. Its primary purpose is to assist hospital networks and trained radiologists in optimizing workflow by flagging and prioritizing studies suspected of containing Intracranial Hemorrhage (ICH). The system utilizes an artificial intelligence algorithm based on convolutional neural networks (CNN) and advanced image processing. It functions as a component of the uAI Easy Triage platform, reading DICOM files, verifying their compatibility, executing the triage algorithm, and communicating findings in DICOM format. The software provides both active pop-up notifications and passive indicators for cases with suspected ICH, along with a non-diagnostic DICOM image preview for radiologists. The algorithm was trained and internally tested using 9791 data points from China. For validation, a retrospective study was conducted using 295 non-contrast CT scans from the USA, with ground truth established by three U.S.-board-certified neuroradiologists. This validation dataset included 147 ICH positive and 148 ICH negative cases, with diverse representation across gender, age, race, CT equipment manufacturers (GE, Philips, Siemens), and ICH subtypes. The uAI Easy Triage ICH demonstrated a sensitivity of 92% (95% CI: 86%-96%) and a specificity of 95% (95% CI: 90%-98%), surpassing the predefined acceptance criteria of 80%. The average time to notification was 41.1 seconds (95% CI: 40.1-42.1 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 algorithm's performance (sensitivity and specificity) vary across different subtypes and severities of Intracranial Hemorrhage, particularly for subtle or small volume hemorrhages?; What were the specific characteristics of the 9791 data points from China used for training and internal testing, and how do they compare demographically and clinically to the U.S. validation dataset?; Are there performance differences or limitations when the uAI Easy Triage ICH processes images from CT scanner models or manufacturers not predominantly represented in the validation dataset (e.g., beyond GE, Philips, Siemens)?; How does the uAI Easy Triage ICH integrate with various existing Picture Archiving and Communication Systems (PACS) and radiology information systems (RIS) in terms of notification delivery and workflow management beyond DICOM compatibility?; What is the expected impact on time-to-diagnosis and patient outcomes in a real-world clinical setting when implementing the uAI Easy Triage ICH for workflow prioritization?