CINA

Cina is an AI-powered software designed to enhance radiological workflow efficiency by rapidly analyzing head CT and CT angiography scans for suspected Intracranial Hemorrhage and Large Vessel Occlusion in specific anterior circulation arteries, providing notifications to prioritize radiologist review. Its scientific basis lies in an artificial intelligence algorithm that identifies imaging characteristics consistent with these critical conditions from DICOM data, thereby enabling faster triage without altering the primary diagnostic images.

Cina is a radiological computer-aided triage and notification software designed for the analysis of non-enhanced head CT images to detect Intracranial Hemorrhage (ICH) and CT angiography of the head to identify Large Vessel Occlusion (LVO) within the anterior circulation, specifically the distal ICA, MCA-M1, or proximal MCA-M2 segments. The system's purpose is to assist hospital networks and trained radiologists by improving workflow triage through flagging and communicating suspected positive findings. It functions in parallel with standard care image interpretation, providing notifications on a standalone web application. Importantly, Cina does not alter original medical images and is not intended as a diagnostic device; its notifications, which include compressed preview images for informational purposes only, are meant to be used alongside other patient information and professional judgment, with clinicians retaining ultimate responsibility for full image review. The core methodology of Cina relies on an artificial intelligence algorithm to process DICOM images. The system, comprising Cina Image Processing Applications for ICH and LVO and the Cina Platform server, receives and analyzes images chronologically. Upon detecting a suspected ICH or LVO, it generates notifications containing patient details, the suspected finding, and the non-diagnostic preview. A retrospective, multicenter, multinational performance study involving 814 non-contrast CT head images for ICH and 476 CT angiography head series for LVO demonstrated high efficacy. For ICH detection, sensitivity was 91.4% and specificity was 97.5% (AUC 0.94). For LVO detection, sensitivity reached 97.9% with a specificity of 97.6% (AUC 0.98). The mean time-to-notification was reported as 13.2 seconds for ICH and 25.8 seconds for LVO, facilitating earlier attention to potentially critical cases.

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 artificial intelligence algorithm differentiate between true positive findings of Intracranial Hemorrhage or Large Vessel Occlusion and potential mimics, and what are the most common sources of false positive notifications?; What were the specific imaging parameters (e.g., scanner manufacturers, slice thickness, contrast timing for CTA) of the 814 NCCT and 476 CTA cases used in the performance study, and is there data on performance variability across these parameters?; Given the device provides compressed preview images, what measures are in place or recommended to ensure these non-diagnostic previews do not inadvertently introduce anchoring bias for radiologists before they review the full diagnostic study?; Could you elaborate on the performance of the LVO detection specifically for each of the mentioned anterior circulation segments (distal ICA, MCA-M1, proximal MCA-M2), including any differences in sensitivity or specificity among these locations?; How was the 'time-to-notification' measured in the performance study (e.g., from scan completion, from image arrival at the Cina server), and how does this translate to expected time savings in a real-world clinical radiology workflow?