AI4CMR v2.0

AI4CMR v2.0 is cloud-based software that automatically analyzes cardiac MR images from 1.5T and 3T scanners to quantify cardiac function metrics (ventricle volumes, ejection fraction, etc.) and calculate flow measurements in blood vessels using artificial intelligence-based segmentation. Results are not intended for standalone clinical decision-making but rather to support clinician interpretation.

AI4CMR v2.0 uses the same cross-platform, client-server architecture as its predicate (v1.0) with identical DICOM 3.0 support and multi-vendor scanner compatibility. The key new feature is 2D Phase Contrast flow quantification capability (previously absent in v1.0), which uses a U-Net convolutional neural network for automated vessel segmentation and deterministic mathematical derivatives for flow measurements.

Software validation followed IEC 62304:2006+A1:2015 and FDA guidance on software in medical devices. Performance was evaluated using the Dice Similarity Coefficient for segmentation accuracy and intraclass correlation coefficients (ICC) for agreement testing with reference measurements.

AI4CMR v2.0 is substantially equivalent because it maintains the same intended use, technological platform, and operating principles as the predicate device v1.0. The new 2D Flow feature was validated against the legally marketed reference device cvi42 (K242781) with strong multi-vendor agreement (ICC 0.95 for total forward volume, 0.82 for total backward volume, 0.95 for maximum velocity), demonstrating equivalent safety and effectiveness without introducing new risks.