Abstract 15550: Fully Automated 'Hands-off' Validation of Left Ventricular Ejection Fraction in Children

IntroductionThree-dimensional echocardiography (3DE) has been shown to be superior to 2DE when quantifying chamber dimensions. However, the clinical utility of 3DE has been limited by time-consuming analysis and need for 3D-specific training. This study aims to assess the clinical utility of a “hand...

Full description

Saved in:
Bibliographic Details
Published inCirculation (New York, N.Y.) Vol. 142; no. Suppl_3 Suppl 3; p. A15550
Main Authors Kauffman, Hunter, Whitehead, Kevin K, O’Connor, Matthew J, Quartermain, Michael D, Convery, Michael, Banerjee, Anirban N
Format Journal Article
LanguageEnglish
Published by the American College of Cardiology Foundation and the American Heart Association, Inc 17.11.2020
Online AccessGet full text

Cover

Loading…
More Information
Summary:IntroductionThree-dimensional echocardiography (3DE) has been shown to be superior to 2DE when quantifying chamber dimensions. However, the clinical utility of 3DE has been limited by time-consuming analysis and need for 3D-specific training. This study aims to assess the clinical utility of a “hands-off,” artificial intelligence-driven software program, Philips Dynamic HeartModelA.I. (DHM), in a pediatric population.HypothesisWe hypothesize that this program will be able to accurately calculate LVEF in children when compared to Cardiac MRI (CMRI) used as a gold standard.MethodsTo date, 10 patients (Average Age 14.8±1.9) have undergone CMRI immediately followed by echocardiography at the Children’s Hospital of Philadelphia. Echocardiographic images were analyzed via DHM without manual editing and from full-volume 3D border tracing (3DEF). We excluded 3D imaging of 2 patients due to poor image quality. 2D LVEF was also evaluated visually in all patients by three experts, and by 2D Simpson’s Biplane method. All analyses were timed with a stopwatch. Each technique was correlated with CMRI.ResultsOur results showed that LVEF measured via DHM was rapid and correlated very well with CMRI (R=0.85). DHM had a stronger correlation with CMRI than Biplane, 3DEF, and visual EF (R=0.56; R=0.68; R=0.57). Moreover, DHM analysis was faster than both Biplane and 3DEF (32±8 seconds vs 100±19 seconds, p<0.005; 32±8 seconds vs 98±16 seconds, p<0.005) (Figure 1). Notably, 3DEF correlated very well with visual estimation by expert observers (R=0.83).ConclusionsBased on our pilot data, DHM analysis in children was faster than both 3DEF and biplane method likely due to its hands-off nature. Furthermore, its correlation with CMRI was superior to both 3DEF and Biplane. Based on further patient enrollment by us and others, this completely automated method will likely be clinically useful as it has potential to become a standard application in large volume, busy echo labs.
ISSN:0009-7322
1524-4539
DOI:10.1161/circ.142.suppl_3.15550