syngo.CT Brain Hemorrhage

syngo.CT Brain Hemorrhage is an artificial intelligence-powered software designed to analyze non-contrast head CT scans to detect suspected intracranial and subarachnoid hemorrhages, providing notification-only output to assist radiologists in prioritizing their case worklists. The system employs an improved algorithmic model for enhanced sensitivity, particularly for smaller hemorrhages and newly included subarachnoid hemorrhage detection, and supports images from multiple CT scanner vendors.

syngo.CT Brain Hemorrhage is an image-processing software that utilizes an artificial intelligence algorithm to support qualified clinicians in analyzing and prioritizing non-contrast head CT images. It is designed to algorithmically identify findings suspicious of acute intracranial hemorrhage (ICH) and acute subarachnoid hemorrhage (SAH), facilitating a parallel workflow where suspected positive cases are flagged for prioritized review. The device provides case-level output to a CT scanner or PACS system for worklist management but does not alter the original medical image, mark specific locations, or remove cases from the reading queue, serving solely as an informational tool and not for standalone diagnostic use. This version of the software introduces an improved algorithm model for ICH triage, aiming for higher sensitivity with smaller hemorrhages, and extends its functionality to include SAH triage. It now supports DICOM conformant head CT reconstructions from various vendors, not just Siemens scanners, and offers expanded deployment compatibility with the syngo.via host system. Performance evaluation in a retrospective study of 900 cases showed a sensitivity of 95.0% and specificity of 93.1% for ICH triage, and 86.1% sensitivity with 85.2% specificity for SAH triage, both exceeding an 80% performance goal, with an average per-case processing time of approximately 13.34 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 improved algorithm's higher sensitivity for smaller intracranial hemorrhages affect the false positive rate compared to the predicate device?; What specific architectural changes or training data differences enabled the extension of the algorithm to reliably triage subarachnoid hemorrhages?; What were the inclusion/exclusion criteria for the 900 cases in the retrospective study, particularly regarding image quality and scanner types (beyond Siemens)?; How does the device handle ambiguous cases or artifacts that might mimic hemorrhages, and what is the typical user response protocol for flagged cases?; What is the range of processing times observed, and how does it vary with case complexity or image characteristics?