NinesAI

NinesAI is an artificial intelligence-driven software that analyzes non-contrast head CT images to detect suspected intracranial hemorrhage and mass effect, functioning as a triage tool to help radiologists prioritize their review workflow. It utilizes machine learning algorithms to automatically identify critical findings from DICOM images within a PACS environment, providing rapid notifications to clinicians.

NinesAI is a radiological computer-aided triage and notification software designed for hospital networks and trained clinicians. It functions as a parallel workflow tool, employing artificial intelligence algorithms to analyze non-contrast head CT images for findings suggestive of emergent clinical conditions, specifically intracranial hemorrhage and mass effect. The system automatically processes Digital Imaging and Communications in Medicine (DICOM) images as they arrive in the Picture Archive and Communication System (PACS) using two distinct machine learning algorithms for each respective finding. Its purpose is to aid in prioritizing radiological review by identifying and alerting radiologists to images with potential emergent findings on their worklist, and it is not intended for diagnostic use beyond notification or to replace full patient evaluation. Preview images displayed are non-diagnostic. Performance testing evaluated the efficacy and safety of NinesAI's machine learning algorithms. For intracranial hemorrhage detection, the algorithm demonstrated a sensitivity of 0.899 and a specificity of 0.974. For mass effect detection, the sensitivity was 0.964 and specificity was 0.911. Both algorithms achieved primary endpoint goals of over 0.80 for sensitivity and specificity, using test datasets from over 20 clinical sites and multiple scanner manufacturers. A time-benefit analysis showed NinesAI provided significantly faster notification (mean 0.23 minutes) compared to standard of care time-to-open-dictation (e.g., 159.4 minutes for intracranial hemorrhage and 28.5 minutes for mass effect), thereby facilitating quicker case prioritization.

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 NinesAI algorithm differentiate between various subtypes of intracranial hemorrhage (e.g., epidural, subdural, subarachnoid, intraparenchymal) and does its sensitivity vary among them?; What is the performance of NinesAI in populations with subtle or atypical presentations of intracranial hemorrhage or mass effect?; How robust is the NinesAI system to variations in CT image quality, acquisition parameters, or artifacts across different institutions and scanner types beyond those tested?; What is the learning curve and training requirement for radiologists to effectively integrate NinesAI notifications into their existing workflow without over-reliance or excessive dismissal of alerts?; Beyond time-to-notification, has any data been collected on the impact of NinesAI on patient outcomes or overall radiologist reading efficiency and diagnostic accuracy for cases not flagged by the AI?