EFAI Intelligent Cardiothoracic Ratio (iCTR) Assessment System

The EFAI Intelligent Cardiothoracic Ratio (iCTR) Assessment System is software designed to automatically calculate the cardiothoracic ratio from adult posterior-anterior chest X-ray images. It employs an artificial intelligence algorithm to measure cardiac and thoracic diameters, providing quantitative data and annotated images to assist clinicians.

The EFAI Intelligent Cardiothoracic Ratio (iCTR) Assessment System is a software medical device designed for hospitals and clinics to automatically assess the cardiothoracic ratio (CTR) from posterior-anterior (PA) view chest X-ray images. Utilizing an artificial intelligence algorithm, the iCTR system measures the maximal transverse diameter of the heart and the maximal inner transverse diameter of the thoracic cavity to calculate the CTR. The system outputs a structured report including a preview of the compressed chest X-ray image with an annotation line indicating these measurements, the CTR result, and a trajectory of CTR records. It is intended for use by physicians or other licensed practitioners in various healthcare institutions for adult patients aged 20-80 years, as an assistive tool, not for standalone clinical decision-making. Clinical performance testing involved 840 eligible PA CXR images from three sites, demonstrating an accuracy of 0.95 for the EFAI iCTR-derived CTR compared to physician-derived CTR. The system incorporates quality control models to filter non-chest X-ray images (Sensitivity 0.99, Accuracy 0.99) and non-PA view images (Sensitivity 0.99, Accuracy 0.97), and an image clarity assessment. The annotation models for heart diameter and thoracic cavity diameter showed root-mean-square errors of 8.81mm and 14.3mm, respectively. Non-clinical testing included software verification and validation according to IEC 62304 and usability engineering per IEC 62366-1 and IEC 62366-2.

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 AI algorithm's performance in calculating CTR vary across different X-ray imaging equipment manufacturers and exposure settings?; What are the specific image quality parameters (e.g., inspiration level, rotation, penetration) that most significantly impact the accuracy of the iCTR system, and what are its tolerance levels?; What was the range and standard deviation of the root-mean-square-errors (8.81mm for heart, 14.3mm for thoracic cavity) for the annotation models, and how do these compare to manual measurement variability?; Beyond the stated 0.95 accuracy, what is the sensitivity and specificity of the EFAI iCTR system for detecting cardiomegaly (defined by a specific CTR threshold) compared to expert human readers?; How does the system handle atypical anatomies or chest X-rays with significant pathologies (e.g., pleural effusions, lung masses) that might obscure cardiac or thoracic borders?