Development of automated neural network prediction for echocardiographic left ventricular ejection fraction

Abstract

Introduction: The echocardiographic measurement of left ventricular ejection fraction (LVEF) is fundamental to the diagnosis and classification of patients with heart failure (HF).

Methods: This paper aimed to quantify LVEF automatically and accurately with the proposed pipeline method based on deep neural networks and ensemble learning. Within the pipeline, an Atrous Convolutional Neural Network (ACNN) was first trained to segment the left ventricle (LV), before employing the area-length formulation based on the ellipsoid single-plane model to calculate LVEF values. This formulation required inputs of LV area, derived from segmentation using an improved Jeffrey's method, as well as LV length, derived from a novel ensemble learning model. To further improve the pipeline's accuracy, an automated peak detection algorithm was used to identify end-diastolic and end-systolic frames, avoiding issues with human error. Subsequently, single-beat LVEF values were averaged across all cardiac cycles to obtain the final LVEF.

Results: This method was developed and internally validated in an open-source dataset containing 10,030 echocardiograms. The Pearson's correlation coefficient was 0.83 for LVEF prediction compared to expert human analysis (p < 0.001), with a subsequent area under the receiver operator curve (AUROC) of 0.98 (95% confidence interval 0.97 to 0.99) for categorisation of HF with reduced ejection (HFrEF; LVEF<40%). In an external dataset with 200 echocardiograms, this method achieved an AUC of 0.90 (95% confidence interval 0.88 to 0.91) for HFrEF assessment.

Conclusion: The automated neural network-based calculation of LVEF is comparable to expert clinicians performing time-consuming, frame-by-frame manual evaluations of cardiac systolic function.

Keywords: artificial intelligence; atrial fibrillation; echocardiogram; ejection fraction; heart failure.

Figure 1

(A–C) were from the Stanford dataset; (D) the CAMUS dataset. (A) human-labelled coordinate points in one frame. A Euclidean distance between two pink points was the LV length; (B) mask generated from these coordinate points, which was used for training our segmentation network; (C) LV area, LV widths, LV heights, and LV length; and (D) annotations included information including the left ventricle endocardium, the left ventricle myocardium, and the left atrium.

Figure 2

(A) Flowchart of the pipeline. There were three main steps, including LV segmentation, LVEF calculation, and HFrEF assessment. The area information from segmentation could also be used for ED and ES identification, beat-to-beat analysis of the heart, as well as visualising changes in volume (for example, due to an arrhythmia such as atrial fibrillation). ED = end diastole; ES = end systole; HFrEF = heart failure with reduced LVEF; LV = left ventricle; LVEF = left ventricular ejection fraction. (B) Input of the pipeline. (C) Proposed AI system. (D) Output information, including the segmentation result and the beat-to-beat visualiser. The calculated LVEF values are presented in this visualiser, along with the results of the HF phenotype classification. (E) outcome.

Figure 3

Overall segmentation architecture. The segmentation network combined ResNet-50 (A), atrous convolutions, and atrous spatial pyramid pooling (ASPP) (B) to resample features at different scales and to capture multi-scale information. As an example, p0, r2, and s1 in the figure denote padding = 0, atrous convolution with rate = 2, and stride = 1, respectively.

Figure 4

Ensemble learning model: including Extra Tree (ET), AdaBoosting (AD), Lasso, and a stacking algorithm combining Ridge, K-nearest neighbours (KNNs), and Gradient Boosting Decision Tree (GBDT). The predicted LV lengths from these regressors were finally ensembled by a voting mechanism.

Figure 5

(A) Three scenarios are used for selecting true peaks, which are identified as ED and ES phases. (B) Improved Jeffrey’s method used to fine-tune LV areas computed from segmentation. Here, three parts were averaged to compute the final LV areas at ED or ES.

Figure 6

Correlation plots. (A–D) Results from the Stanford dataset, whilst (E,F) from the CAMUS dataset. (A) Correlation between LVEF values derived from segmentation results directly and those labelled by an experienced clinician. (B) Correlation between LVEF values derived from the proposed Jeffrey’s method and those labelled by the clinician. (C) Correlation between LVEF values computed from a single cardiac cycle and labelled LVEF values. (D) Correlation between LVEF values computed from all cardiac cycles and labelled LVEF values. (E) Correlation between LVEF values derived from fine-tuned segmentation results and labelled LVEF values. (F) Correlation between LVEF values derived from the improved Jeffrey’s method and labelled LVEF values.

Figure 7

HFrEF assessment results. (A) ROC curves of different methods, each having an AUC value. (B) and (C) Confusion matrices computed from the Stanford and CAMUS datasets, respectively.

Figure 8

Beat-to-beat analysis. (A) and (C) Two samples with normal sinus rhythm. (B) Patient with atrial fibrillation. (C) and (D) Human-labelled ED and ES were not exactly at peak or bottom positions.