Annalise Enterprise CTB Triage Trauma

Annalise Enterprise CTB Triage Trauma is a software tool that utilizes a deep-learning artificial intelligence algorithm, specifically a convolutional neural network, to analyze non-contrast brain CT studies for features suggestive of vasogenic edema in adult patients. Its purpose is to aid clinicians by providing study-level notifications for worklist prioritization, enabling potentially faster review of critical cases without altering standard diagnostic practices or downgrading existing study priorities.

Annalise Enterprise CTB Triage Trauma is a software workflow tool that employs an artificial intelligence (AI) algorithm, specifically a convolutional neural network trained using deep-learning techniques, to identify suspected vasogenic edema in non-contrast brain CT studies. The device is designed for use in a medical care environment, interfacing with image and order management systems like PACS/RIS. When the AI algorithm detects features suggestive of vasogenic edema, it provides a study-level notification to these systems, facilitating the prioritization of such studies in the radiologist's worklist. This is intended to enable earlier review of potentially critical findings, operating in parallel with the standard clinical workflow for interpreting non-contrast brain CTs in patients aged 22 years or older. The device does not direct attention to specific image portions and its results are not for standalone clinical decision-making or ruling out findings, nor will it ever decrease a study's existing priority. The performance of the AI algorithm was validated in a standalone, retrospective study using an independent dataset of non-contrast brain CT cases, comparing device output to a reference standard determined by US board-certified neuroradiologists. For vasogenic edema, across different operating points and slice thicknesses (≤1.5mm and >1.5mm & ≤5.0mm), the device demonstrated sensitivities ranging from 89.9% to 94.9% and specificities from 89.7% to 97.4%. For instance, at one operating point for slice thickness ≤1.5mm, sensitivity was 91.7% and specificity was 89.7%. The test dataset included diverse patient demographics, imaging equipment, and co-existing conditions. A separate bench study on 277 positive cases demonstrated a triage turn-around time of approximately 81.6 seconds. These results indicate the device's capability for effective triage within a clinician's queue.

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 algorithm's performance for detecting vasogenic edema vary with different CT scanner manufacturers, models, or acquisition protocols beyond slice thickness?; What was the inter-rater reliability among the neuroradiologists establishing the ground truth, and how were disagreements resolved beyond simply using a third truther?; Were there specific subtypes or etiologies of vasogenic edema that the algorithm performed better or worse on, and were these analyzed in the validation study?; How does the device handle studies with significant artifacts or poor image quality, and what is the failure rate or notification mechanism for uninterpretable studies?; What impact does the 81.6-second triage turn-around time have on the overall radiology workflow and patient outcomes in a real-world, high-volume setting?