Clinical study on feasibility and repeatability of left ventricular systolic function assessment in patients with chronic kidney disease using artificial intelligence-based automatic strain technique
Related Experiment Video
Updated: Nov 9, 2026
Evaluation of Left Ventricular Structure and Function using 3D Echocardiography
Published on: October 28, 2020
Fang Tian1, Zhuole Wu1, Lianli Zhong1
1Department of Ultrasonography, Hainan General Hospital, Hainan Affiliated Hospital of Hainan Medical University, Haikou, Hainan 570311, China.
Background:
Cardiovascular damage is a common complication in patients with chronic kidney disease (CKD), and left ventricular longitudinal strain (GLS) is superior to LVEF in evaluating systolic function. However, it has not been widely adopted clinically because it requires proficiency and is time-consuming. This study aimed to investigate the feasibility, reproducibility, and predictive value of automated GLS compared with manual GLS, providing a reference for clinical evaluation and reduction of cardiovascular events in CKD patients.
Methods:
A total of 285 CKD patients (aged 52 ± 12.85) without dialysis were enrolled from Hainan Provincial People's Hospital. GLS was measured using three methods on identical apical three-, two-, and four-chamber views: (1) fully automatic GLS analyzed by machine functions, (2) semi-automatic GLS corrected by researchers, and (3) manual GLS measured by experts. Five cases were excluded due to poor image quality. Clinical outcomes were followed up by telephone and outpatient visits.
Results:
About 35% of automatic GLS results required manual correction, with significant differences among the three methods (P < 0.01). The correlation and consistency between semi-automatic and manual GLS were higher than those of automatic GLS (P < 0.01). During 2-year follow-up, 55 patients (19.6%) experienced cardiovascular events. Automated GLS predicted events but with lower accuracy than semi-automatic GLS. The analysis time for automatic (15.23 ± 0.75 s) and semi-automatic GLS (75.06 ± 19.01 s) was markedly shorter than manual analysis (236.81 ± 45.41 s, P < 0.01).
Conclusions:
Automated GLS assessment in CKD patients is feasible and reproducible. As some images still require manual correction, the semi-automated approach currently offers the optimal balance between efficiency and predictive value.
