Validating an AI Model to Screen for Heart Valve Diseases Using Routine Non-Contrast Chest CT Scans in Adults

What this Trial Is About

Researchers are evaluating a deep learning model designed to screen for valvular heart diseases using routine, non-contrast chest CT scans. This prospective, multicenter study aims to validate the models performance, focusing primarily on its sensitivity to detect moderate-to-severe heart valve disease. Additional measures include accuracy, specificity, and the area under the receiver operating characteristic curve AUC. Participants from physical exam and outpatient clinics will undergo a standard non-contrast chest CT scan. The AI model will analyze these images in real-time to identify those with moderate-to-severe valvular heart disease. Individuals flagged by the model will immediately receive a confirmatory echocardiogram, which serves as the reference standard for diagnosis. The study will use statistical methods to assess the AI models diagnostic accuracy over a 12-month period. During the study, participants will have routine chest CT scans and possibly an echocardiogram if indicated by the AI model. Researchers will collect and analyze imaging results and clinical records. The primary outcome measured is the models sensitivity within one year, along with secondary outcomes of AUC, specificity, and accuracy. The study involves no additional safety risks beyond standard imaging procedures and is expected to last about 12 months.

AI Assisted Screening for VHD Using Routine Chest CT Scans