Rapid ICH

Rapid ICH is a software device utilizing an artificial intelligence algorithm to analyze non-enhanced head CT images for the detection and notification of suspected intracranial hemorrhages. Its purpose is to expedite clinical workflow by providing rapid notifications to radiologists, thereby enabling prioritization of potentially critical cases and significantly reducing the time to image assessment compared to standard care.

Rapid ICH is a radiological computer-aided triage and notification software designed for the analysis of non-enhanced head CT (NCCT) images. Its primary function is to assist hospital networks and trained radiologists in workflow triage by flagging and communicating suspected positive findings of various intracranial hemorrhages (ICH), including intraparenchymal (IPH), intraventricular (IVH), subarachnoid (SAH), and subdural (SDH) types. The system employs an artificial intelligence (AI) algorithm to analyze these images, identifying and highlighting cases with detected ICH. This process occurs on a server or standalone desktop application, operating in parallel with the standard image interpretation workflow. Users receive notifications for cases with suspected ICH, which include compressed preview images intended for informational and triage purposes only, not for direct diagnosis, and the device does not alter original medical images. The performance of Rapid ICH was evaluated in a retrospective, multicenter, multinational study involving 314 NCCT head images, demonstrating an overall sensitivity of 96.8% and specificity of 100% in identifying ICH, with an AUC of 0.98632. A significant clinical benefit is the reduction in time-to-notification, averaging 0.65 minutes with Rapid ICH compared to a standard care time-to-exam-open of 72.6 minutes. Performance was further detailed across various ICH subtypes, patient demographics (gender and age), hemorrhage volume (e.g., 100% sensitivity for volumes ≥ 5mL), CT slice thickness, and different CT scanner manufacturers and models, generally showing robust results. The software is intended to be used in conjunction with other patient information and professional judgment to assist with triage and prioritization of medical images, with notified radiologists responsible for viewing full images per standard care.

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 the various specified ICH subtypes (IPH, IVH, SAH, SDH) with varying degrees of subtlety or co-occurrence?; What were the specific characteristics of the 'additional data for training and validation' used for this version compared to K193087, and how did this data impact the algorithm's performance, particularly for challenging or rare ICH presentations?; How does the system handle variations in CT image quality, noise levels, or artifacts from different scanners or protocols, and what is the minimum image quality required for reliable performance?; Beyond time-to-notification, has there been any investigation into the downstream impact on clinical decision-making, radiologist reading times for positive cases, or patient outcomes when Rapid ICH is integrated into a clinical workflow?; What are the performance limitations, if any, for detecting very small volume hemorrhages (e.g., below the 0.4 mL threshold mentioned in performance metrics) or hemorrhages in anatomically complex locations?