Annalise Enterprise CTB Triage Trauma

Annalise Enterprise CTB Triage Trauma leverages a deep-learning artificial intelligence algorithm to analyze non-contrast brain CT scans, specifically identifying 'mass effect' to enhance radiology workflow by prioritizing potentially urgent studies. The system's methodology involves seamless integration with PACS/RIS for study-level notifications, with its efficacy in triage supported by robust performance data demonstrating high sensitivity and specificity for the target finding.

Annalise Enterprise CTB Triage Trauma is a software tool designed to assist in the triage and prioritization of non-contrast brain CT studies by identifying features suggestive of "mass effect." It utilizes an artificial intelligence algorithm, specifically a convolutional neural network developed through deep-learning techniques, to analyze the imaging data. Upon detecting a suspected mass effect, the device provides a study-level notification to the existing order and imaging management systems (like PACS/RIS), facilitating earlier review of potentially critical scans by elevating their position in the worklist. The system is intended for use by trained clinicians qualified to interpret brain CTs in patients aged 22 years or older. Importantly, its output is not intended for standalone diagnostic decision-making or to definitively rule out findings, but rather to augment the standard clinical workflow by highlighting studies for prioritized attention without ever de-prioritizing a study. The device's analytical capabilities were validated through a retrospective, anonymized standalone performance evaluation using an independent test dataset of non-contrast brain CT cases, distinct from its training data. Ground truth for this evaluation was established by at least two US board-certified neuroradiologists, with a third for resolving discrepancies. Performance results for detecting mass effect showed high sensitivity and specificity across various CT slice thicknesses and operating points; for example, with a slice thickness of ≤1.5mm at one operating point, the device achieved 97.0% sensitivity and 88.7% specificity. Furthermore, a bench study involving 277 positive cases demonstrated an average triage turn-around time of 81.6 seconds, indicating its efficiency in processing and notification. This performance was validated across diverse patient demographics and imaging equipment characteristics.

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. What specific imaging biomarkers or features does the AI algorithm primarily leverage to detect 'mass effect,' and how robust is its differentiation from mimics or other neurological pathologies on non-contrast CT scans?; How does the device's performance, particularly sensitivity and specificity for 'mass effect,' vary when encountering images from CT scanners with significantly different technical specifications (e.g., detector technology, reconstruction algorithms) not broadly covered in the validation datasets?; What was the detailed methodology for establishing 'ground truth,' including the inter-rater reliability among the neuroradiologists and the specific criteria used to adjudicate disagreements for the 'mass effect' finding?; Considering the device is specific to 'mass effect,' what is the protocol or expected clinical behavior if a scan contains other urgent findings (e.g., acute hemorrhage not causing significant mass effect) that this AI does not flag for prioritization?; Could you elaborate on the components included in the 'triage turn-around time' of 81.6 seconds, specifically detailing the time taken for image acquisition by the AI, processing, and the actual notification delivery to the PACS/RIS worklist in a typical hospital IT infrastructure?