Validation of machine learning models for estimation of left ventricular ejection fraction on point-of-care ultrasound: insights on features that impact performance

Table 2 Single video imaging data split by type of rater

Rater	Number of videos assigned an LVEF	Number of videos of insufficient quality for LVEF estimation	Mean estimation of LVEF ± SD
ML model	341	916	0.39 ± 0.13
Level III expert visual LVEF on randomized single videos	851	406	0.41 ± 0.13
Level III expert segmentation LVEF on randomized single videos	245	1012	0.40 ± 0.14
Level III expert visual LVEF, accounting for all videos for a participant	1175*	82	0.40 ± 0.13
Level III expert segmentation LVEF, accounting for all videos for a participant	754^#	503	0.41 ± 0.13
Derived LVEF from echo reports	N/A	N/A	0.39 ± 0.12

^*All videos for a participant were included in this category if at least one video in the study was assigned an LVEF by visual assessment
^#All videos for a participant were included in this category if at least one video in the study was assigned an LVEF by segmentation

ISSN: 2055-0464