Libby Echo:Prio

Libby Echo:Prio is an advanced software tool designed to enhance the analysis of cardiovascular ultrasound images by utilizing machine learning for automated view classification, cardiac segmentation, and quantification of key parameters like ejection fraction. It aims to provide clinicians with accurate, consistent, and efficient data to support diagnostic decision-making in adult patients undergoing echocardiographic examinations.

Libby Echo:Prio, from Dyad Medical, Inc., is an image post-processing analysis software designed for viewing and quantifying cardiovascular ultrasound images from adult patients. Its primary purpose is to aid qualified healthcare practitioners, such as cardiologists and sonographers, in diagnostic review, analysis of echocardiographic data, patient record management, and reporting. The software processes previously acquired DICOM-compliant cardiac ultrasound images, allowing sonographers to perform necessary markups, including identifying cardiac segments and taking measurements of distance, time, and area. It supports both manual markup and leverages machine learning for automated view classification and border segmentation, which can be accepted, modified, or rejected by the user. The software provides functionalities for 2D image and cine loop review, and importantly, quantifies key cardiac parameters. These include echocardiographic view classification, identification of end-diastolic (ED) and end-systolic (ES) frames for heartbeat rate (HR) estimation, and calculation of ejection fraction (EF). Ejection fraction assessment is achieved through automatic segmentation of the left ventricle (LV) chamber endocardium, left atrium (LA), and LV myocardium (LVMC), allowing muscle thickness calculations. Performance testing demonstrated high accuracy, with view classification achieving 97% accuracy (F1 >96.6%, Sensitivity 96.8%, Specificity 98.5%), and heart rate estimates showing minimal bias (slope of 0.98) compared to 12-lead ECG. The automated ejection fraction prediction showed a strong correlation (slope of 0.79, 95% CI 0.52-0.98) with annotations by four human experts, within the range of typical inter-observer variability.

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 machine learning algorithm's performance for automated segmentation and view classification vary across different ultrasound machine vendors and image qualities?; What specific features or parameters beyond ejection fraction and heart rate can Libby Echo:Prio quantify, and what is the validation data for these additional measurements?; The performance data mentions comparison to "four human experts" for EF prediction; what was the specific protocol for these expert annotations and how was inter-expert variability managed or accounted for in the comparison?; How does the software handle challenging or atypical cardiac anatomies or pathologies, and are there known limitations in its automated analysis capabilities for such cases?; What is the typical processing time for a standard echocardiographic study using Libby Echo:Prio, and how does this integrate into existing clinical workflows for sonographers and cardiologists?