AI Assisted Screening for VHD Using Routine Chest CT Scans

Overview

This is a prospective, multicenter study designed to validate a deep learning model for screening valvular heart diseases using routine, non-contrast chest computed tomography (CT) scans. The primary objective is to evaluate the model's diagnostic performance, with the sensitivity serving as the primary efficacy endpoint. Secondary endpoints will include other performance metrics such as area under the receiver operating characteristic curve (AUC), specificity, and accuracy, etc.

This is a prospective, multicenter study designed to validate a deep learning model for screening valvular heart diseases using routine, non-contrast chest computed tomography (CT) scans from individuals in any medical context within a hospital alliance. The primary objective is to evaluate the model's diagnostic performance, with the sensitivity serving as the primary efficacy endpoint. Secondary endpoints will include other performance metrics such as area under the receiver operating characteristic curve (AUC), specificity, and accuracy, etc. Participants from the target populations will undergo a routine non-contrast chest CT scan. The deep learning model will analyze these images in real-time. For those identified by the model as having moderate-to-severe heart valve disease, a confirmatory echocardiogram will be performed immediately. The echocardiogram results will serve as the reference standard for diagnosis. Statistical analyses will be performed to assess the model's performance against this reference, including calculating the 95% confidence interval for the AUC. As this study only involves standard, low-radiation diagnostic imaging procedures (non-contrast CT and echocardiography) that are part of routine clinical care, it is considered to pose no additional relevant safety risks to participants. The total study duration is estimated to be 12 months.

Conditions