EchoGo Core (2.0)

EchoGo Core 2.0 is an advanced software application utilizing artificial intelligence to provide operator-assisted automatic quantification of left ventricular function from adult echocardiograms, aiming to support physicians in cardiovascular diagnosis. Its methodology involves AI-driven view classification, LV contouring, and calculation of key metrics like ejection fraction and global longitudinal strain, with its performance validated through rigorous testing against a reference comparator.

EchoGo Core 2.0, manufactured by Ultromics, is a software application designed to assist physicians in diagnosing cardiovascular conditions in adult populations by quantifying and reporting cardiac function from echocardiogram images. The system employs artificial intelligence for operator-assisted automatic quantification of common echocardiographic metrics, operating through a sequential workflow. This workflow includes automatic view classification (A2C, A3C, A4C) and labeling, which the operator confirms or relabels; AI-proposed contouring of the left ventricle; automated heart cycle identification with end-diastolic and end-systolic frame selection, which the operator approves; and finally, the calculation of clinical report metrics. The results are provided as secondary capture DICOM files or a structured DICOM report. The core technology of EchoGo Core 2.0 relies on AI algorithms for auto-view classification, auto-contouring with contours overlayed on images, DICOM handling, technical quality control, and automated calculation of metrics including ejection fraction (EF), global and segmental longitudinal strain (GLS, SLS) from A2C, A3C, and A4C views, and left ventricular (LV) volumes (including indexed values and non-Simpson's methods), length, and stroke volume. It also provides various visual representations like horseshoe and bullseye plots. Performance data from a retrospective validation study involving 214 unseen studies demonstrated its efficacy, achieving pre-determined acceptance criteria for Root Mean Square (RMS) error against a reference comparator (TomTec Arena TTA2). Reported RMS error percentages include 3.06-4.59% for LV length, 8.57-16.59% for LV volumes, 6.69-8.50% for EF, and 3.36-4.79% for GLS. The system also showed excellent inter-operator and intra-operator variability with a Deming RMSE of 0 for all reported metrics.

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 EchoGo Core 2.0, particularly for view classification and contouring, vary across echocardiograms with suboptimal image quality or rare cardiac pathologies?; What were the specific pre-determined acceptance criteria (RMS error thresholds) for each reported metric during the validation study, and what was the rationale for selecting these thresholds for clinical relevance?; Given the operator-assisted nature, what is the typical time saving or workflow efficiency gain observed when using EchoGo Core 2.0 compared to manual quantification methods or its predecessor, EchoGo Core 1.0?; Does EchoGo Core 2.0 offer capabilities for tracking changes in cardiac function longitudinally for the same patient, and how does it ensure consistency in measurements across different studies and operators?; What types of 'cautions' can the EchoGo Core 2.0 system generate for operator approval, and how does the AI determine when a caution is necessary regarding the quality or reliability of the automated analysis?