JLK-ICH

JLK-ICH is an AI-powered software system that analyzes non-contrast CT head scans to detect intracranial hemorrhage (ICH) and sends real-time notifications to clinicians. It operates as a parallel workflow tool independent of standard care, enabling rapid triage and specialist notification. The system is notification-only; clinicians must view full diagnostic images and conduct complete patient evaluation before making treatment decisions.

Both JLK-ICH and predicate Viz ICH use AI/ML algorithms with DICOM-compatible image processing, hosted on similar server architectures that automatically receive and analyze NCCT scans. Both include mobile notification applications with identical functions, do not alter original images, and provide non-diagnostic preview viewing only. JLK-ICH achieves faster notification time (0.19±0.04 minutes vs. 0.49±0.15 minutes for Viz ICH) while maintaining the same output and workflow integration.

Not stated in this summary. The document references FDA guidance documents ('Content of Premarket Submissions for Device Software Functions,' June 14, 2023) and mentions compliance with 21 CFR 892.2080 special controls for radiological CAD software, but no specific ISO, IEC, or ASTM consensus standards are cited.

JLK-ICH demonstrates substantial equivalence through identical intended use (ICH detection and notification), matching technological architecture (AI algorithm + DICOM-compatible mobile app), and superior clinical performance. Standalone testing on 376 NCCT scans showed sensitivity of 97.3% (95% CI 94.8–99.5%) and specificity of 97.9% (95% CI 95.5–99.5%), both exceeding the predicate's documented performance. Stratified analysis across age, gender, race-ethnicity, scanner manufacturers, and ICH subtypes confirms consistent, non-inferior performance across clinically relevant subgroups.