Machine learning for the real-time assessment of left ventricular ejection fraction in critically ill patients: a bedside evaluation by novices and experts in echocardiography

Rita Varudo; Filipe A Gonzalez; João Leote; Cristina Martins; Jacobo Bacariza; Antero Fernandes; Frederic Michard

doi:10.1186/s13054-022-04269-6

Machine learning for the real-time assessment of left ventricular ejection fraction in critically ill patients: a bedside evaluation by novices and experts in echocardiography

Crit Care. 2022 Dec 14;26(1):386. doi: 10.1186/s13054-022-04269-6.

Authors

Rita Varudo¹, Filipe A Gonzalez^{1

2}, João Leote¹, Cristina Martins¹, Jacobo Bacariza¹, Antero Fernandes^{1

2

3}, Frederic Michard⁴

Affiliations

¹ Intensive Care Department, Hospital Garcia de Orta, Almada, Portugal.
² Faculdade de Medicina da Universidade de Lisboa, Lisbon, Portugal.
³ Faculdade de Ciencias da Saude da Universidade da Beira Interior, Covilha, Portugal.
⁴ MiCo, Vallamand, Switzerland. frederic.michard@bluewin.ch.

Abstract

Background: Machine learning algorithms have recently been developed to enable the automatic and real-time echocardiographic assessment of left ventricular ejection fraction (LVEF) and have not been evaluated in critically ill patients.

Methods: Real-time LVEF was prospectively measured in 95 ICU patients with a machine learning algorithm installed on a cart-based ultrasound system. Real-time measurements taken by novices (LVEF_Nov) and by experts (LVEF_Exp) were compared with LVEF reference measurements (LVEF_Ref) taken manually by echo experts.

Results: LVEF_Ref ranged from 26 to 80% (mean 54 ± 12%), and the reproducibility of measurements was 9 ± 6%. Thirty patients (32%) had a LVEF_Ref < 50% (left ventricular systolic dysfunction). Real-time LVEF_Exp and LVEF_Nov measurements ranged from 31 to 68% (mean 54 ± 10%) and from 28 to 70% (mean 54 ± 9%), respectively. The reproducibility of measurements was comparable for LVEF_Exp (5 ± 4%) and for LVEF_Nov (6 ± 5%) and significantly better than for reference measurements (p < 0.001). We observed a strong relationship between LVEF_Ref and both real-time LVEF_Exp (r = 0.86, p < 0.001) and LVEF_Nov (r = 0.81, p < 0.001). The average difference (bias) between real time and reference measurements was 0 ± 6% for LVEF_Exp and 0 ± 7% for LVEF_Nov. The sensitivity to detect systolic dysfunction was 70% for real-time LVEF_Exp and 73% for LVEF_Nov. The specificity to detect systolic dysfunction was 98% both for LVEF_Exp and LVEF_Nov.

Conclusion: Machine learning-enabled real-time measurements of LVEF were strongly correlated with manual measurements obtained by experts. The accuracy of real-time LVEF measurements was excellent, and the precision was fair. The reproducibility of LVEF measurements was better with the machine learning system. The specificity to detect left ventricular dysfunction was excellent both for experts and for novices, whereas the sensitivity could be improved.

Trial registration: NCT05336448. Retrospectively registered on April 19, 2022.

Keywords: Artificial intelligence; Echocardiography; Left ventricular ejection fraction; Machine learning; Novice; Point of care ultrasound.

Publication types

Clinical Study

MeSH terms

Cardiomyopathies*
Critical Illness
Echocardiography
Humans
Machine Learning
Reproducibility of Results
Stroke Volume
Ventricular Dysfunction, Left* / diagnostic imaging
Ventricular Function, Left

Associated data

ClinicalTrials.gov/NCT05336448