Abstract

Background and aims: Paediatric myocardial strain analysis through echocardiography is often characterized by high variance and limited precision, highlighting the need for a standardized and vendor-agnostic approach applicable for diverse image qualities and populations, which could enhance cardiac function evaluation and enable early detection of cardiac impairment.

Methods: The Motion-Echo system was proposed, a semi-supervised deep learning framework built on 11 096 paediatric and 11 297 adult echocardiograms spanning diverse image qualities and vendors. It integrated context compensation and motion estimation modules for temporally coherent segmentation, myocardial motion estimation, and global strain assessment with minimal manual annotations. Clinical utility was further evaluated through downstream applications.

Results: Motion-Echo achieved mean absolute errors of 2.099% [95% confidence interval (CI) 1.803-2.401] and 2.665% (95% CI 2.339-3.026) for global longitudinal and circumferential strain assessments, with Pearson correlation coefficients of 0.799 (95% CI 0.715-0.871) and 0.781 (95% CI 0.687-0.844), respectively. To validate the clinical utility, automated strain values achieved an area under the curve (AUC) of 0.906 (95% CI 0.816-0.981) for cancer therapy-related cardiac dysfunction risk prediction. For late gadolinium enhancement detection, automated global longitudinal strain reached an AUC of 0.782 (95% CI 0.666-0.885). For left ventricular ejection fraction decline forecasting, the system outperformed manual strain values (DeLong P < .001). In addition, incorporating estimated motion flows yielded a remarkable AUC improvement to 0.952 (95% CI 0.917-0.980) for myocardial infarction detection.

Conclusions: Leveraging a large-scale paediatric dataset, Motion-Echo provided a reliable and generalizable framework for myocardial strain analysis, demonstrating potential to facilitate earlier detection of cardiac dysfunction and generate digital cardiac function profiles of children.

Keywords: Deep learning; Echocardiography; Motion estimation; Myocardial strain; Paediatric cardiology; Semi-supervised learning.

Figure 1 Overview of Motion-Echo. Echocardiograms from apical and parasternal short-axis views (A) were collected as inputs to Motion-Echo for motion-based pre-training and segmentation (B). Motion-Echo generated consistent segmentation boundaries and optical flows throughout entire car diac cycles. The segmentation outputs were used for global longitudinal strain and global circumferential strain estimation (C). For optical flows repre senting myocardial motion features, an acute myocardial infarction detection task was employed to evaluate their effectiveness (D). Comprehensive subgroup analyses were conducted to assess global longitudinal strain and global circumferential strain estimation performance (E). Downstream clinical tasks were further utilized to evaluate the clinical utility of automated strain values, including cancer therapy-related cardiac dysfunction risk prediction, late gadolinium enhancement detection, and left ventricular ejection fraction decline forecasting (F). Motion-Echo was capable of generating a compre hensive digital profile of paediatric cardiac progression and offering early-stage risk alerts for impaired cardiac function (G).

Figure 2 Left ventricle and myocardium segmentation performance of Motion-Echo. (A) Visualization of segmentation results on the Huaxi testing dataset. (B and C) Dice similarity coefficient on apical views (B) and parasternal short-axis views (C) compared across different semi-supervised learning methods using Huaxi testing data. (D) Dice similarity coefficient on the CAMUS dataset trained with different sample sizes of training data. (E–G) Temporal consistency analysis of Motion-Echo on segmentation boundaries using dice similarity coefficient (E), 95% Hausdorff distance (F), and average symmetric surface distance (G) compared with baseline model. Note that the reported performance is averaged across left ventricle and myocardium segmentation. Significance levels with two-sided paired t-test are ****P < 1 × 10−4, and ns for P > 5 × 10−2.

DOI: 10.1093/eurheartj/ehaf952