CINA-iPE

CINA-iPE is a software tool intended to expedite the detection of incidental pulmonary embolisms in non-dedicated contrast-enhanced CT scans by alerting radiologists to suspicious findings, thereby optimizing workflow triage for potentially critical conditions. The system utilizes a sophisticated deep learning algorithm, trained on a substantial and varied international dataset, to autonomously analyze CT images for indicators of iPE and provide rapid notifications integrated into the existing radiological workflow.

CINA-iPE is a radiological computer-aided triage and notification software designed to assist radiologists in managing workflow by identifying suspected incidental pulmonary embolisms (iPE) in adult and transitional adolescent patients. These patients are undergoing contrast-enhanced CT scans for clinical indications other than pulmonary embolism suspicion, where the scan includes at least part of the lung. The software employs a deep learning algorithm, trained on 5,429 diverse cases from the US and France, to analyze these non-dedicated CT scans. It flags potential iPE cases and sends notifications, including a non-diagnostic compressed preview image, to a worklist application. This system operates in parallel with the standard image interpretation process, aiming to prioritize potentially critical findings without altering original medical images or replacing the radiologist's ultimate diagnostic responsibility. CINA-iPE is not designed to detect PE in subsegmental arteries but uses models to locate major vessels for contrast assessment and segment lung parenchyma to define the region of interest. The performance of CINA-iPE was evaluated in a retrospective, multicenter, multinational, blinded study using 381 clinical anonymized cases. The software demonstrated a global sensitivity of 87.8% and a specificity of 92.0% in identifying iPE when compared against a ground truth established by a consensus of three US board-certified radiologists. Subgroup analysis showed varying sensitivities for PEs in different arterial locations: 96.3% for main, 94.5% for interlobar, 92.9% for lobar, and 88.3% for segmental arteries. The system also proved efficient, with a mean time-to-notification of 1.5 minutes for all cases and for true positive cases. These results indicate the software's capability to assist in the early triage and prioritization of medical images with suspected iPE findings, enhancing the efficiency of radiological practice.

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 CINA-iPE algorithm's performance (sensitivity/specificity) vary with different levels or timing of contrast opacification in the pulmonary arteries, and what is the minimum threshold for reliable analysis?; Given the training dataset included cases from the US and France, were there any notable performance differences observed when the software was validated exclusively on US patient populations or specific demographic subgroups not detailed in the summary?; What specific deep learning architectural components (e.g., CNN types, attention mechanisms) are employed by CINA-iPE, and how do they contribute to the identification of iPE features and the segmentation of relevant anatomy like the lung parenchyma and major vessels?; Since CINA-iPE is not designed to detect PE in subsegmental arteries, what is the quantified miss rate for subsegmental PEs that were present in the validation dataset, and what is the potential clinical impact or follow-up recommendation for these cases?; How does the CINA-iPE system handle challenging CT imaging scenarios, such as those with significant motion artifacts, beam hardening, low-dose protocols, or co-existing lung pathologies (e.g., pneumonia, tumors) that might mimic or obscure pulmonary emboli?