Abstract Funding Acknowledgements Type of funding sources: None. Background Transthoracic echocardiography is the first-line non-invasive investigation for assessing pediatric patients' cardiac anatomy, physiology, and hemodynamics, based on its accessibility and portability, but complete anatomic and hemodynamic assessment is time-consuming. The study aimed to determine the accuracy and reliability of a new automated artificial intelligence software initially developed for adults for analyzing the main echocardiographic parameters in pediatric patients. Materials and Methods The study was performed at the University Hospital of Bordeaux between August and September 2022 and included 45 patients who underwent a 2D transthoracic echocardiographic examination, mainly for murmur (n=8), post ASD closure evaluation(N=5), chest pain, or arrhythmia as we aimed to have a homogeneous group of patients with normal or near normal heart architecture. We compared AI automatic measurements with manual measurements performed by a senior and junior pediatric cardiologist. Pearson correlation and Bland and Altman analysis were used. As the AI software has been developed for adults, we decided to subclassify our patients into two subgroups: the first group (patients aged between 0–8 years) and the second group (patients over 8 years). The decision was made because pediatric anatomy differs from adults and to see if the dedicated artificial intelligence soft for adults has better reproducibility for pediatric patients older than 8 years. Results The median age was 8.9 years. Comparison between senior and junior pediatric cardiologists for all subgroups was significant (Pearson correlation >0,7 for most measurements performed). For younger patients (<8 years old), we found excellent correlations (Pearson correlation >0.8) for heart chambers measurements measured from A4CH and A2CH views (LV end-diastolic volume, LV end-systolic volume, RV end-diastolic area, and RV area) between AI software and both junior and senior pediatric cardiologist, but poor correlation for the measurements performed from parasternal long axis view (r=0.06 for STJ measurement). We achieved stronger correlations for the measurements evaluation of patients over 8 years old ( r>0.8 for aortic annulus, STJ, E, A, S waves, LV end-diastolic volume in A4CH, LV end-systolic volume in A4CH) for both junior and senior cardiologist in relation to AI measurements. Conclusion In this preliminary study, we found that automated artificial intelligence software provided promising results for measuring mainly volumes (LV end-diastolic volume, LV end-systolic volume) in A4CH and A2CH. The software requires improvements for automated measurements within the parasternal long-axis section to achieve better accuracy. Results obtained with the AI software for patients over 8 years old were significantly improved compared to those obtained for those under 8 years of age.
Read full abstract