Population data-based federated machine-learning improves automated echocardiographic quantification of cardiac structure and function – the AVE project

Abstract Background Machine-learning (ML)-based automated measurement of echocardiography images emerged as an option to reduce observer variability. Objectives To improve the accuracy of a pre-existing automated reading tool (“original detector”) by federated ML-based re-training. Methods AVE (Automatisierte Vermessung der Echokardiographie) was based on the echocardiography images of n = 4,965 participants of the population-based Characteristics and Course of Heart Failure Stages A-B and Determinants of Progression Cohort Study. We implemented federated ML: echocardiography images were read by the Academic CoreLab Ultrasound-based Cardiovascular Imaging at the University Hospital Würzburg (UKW). A random algorithm selected 3,226 participants for re-training of the original detector. According to data protection rules, generation of ground truth and ML training cycles took place within the UKW network. Only non-personal training weights were exchanged with the external cooperation partner for refinement of ML algorithms. Results Both the original detector as the re-trained detector were then applied to the echocardiograms of n = 563 participants not used for training. With regards to the human referent the re-trained detector revealed 1) superior accuracy when contrasted with the original detector´s performance as it arrived at significantly smaller mean differences in all but one parameter, and 2) smaller absolute difference between measurements when compared to a group of different human observers. Conclusions Population data based ML in a federated ML set-up was feasible. The re-trained detector exhibited a much lower measurement variability than human readers. This gain in accuracy and precision strengthens the confidence into automated echocardiographic readings, which carries large potential for applications in various settings.