Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

Schematic overview of the study.

(a) Flow diagram of patient selection and dataset configuration. (b) Structural diagram of the DL model for LVH classification. Abbreviations: ECG, electrocardiography; LVH, left ventricular hypertrophy.

More »

Fig 1 Expand

Fig 2.

Structural diagram of model.

(a) Architectural diagram of the CoAt-Mixer. (b) Architectural diagram of the ResNet-CBAM. Abbreviations: ECG, electrocardiography; GeLU, gaussian error linear unit; FNN, feedforward neural network; ReLU, rectified learning unit.

More »

Fig 2 Expand

Table 1.

Dataset characteristics.

More »

Table 1 Expand

Fig 3.

Visualization of statistical metrics.

(a) Confusion matrix and statistical metrics for binary classification. (b) ROC curve and AUROC. Abbreviations: LVH, left ventricular hypertrophy; ROC, receiver operating characteristic; AUROC, area under receiver operating characteristics curve.

More »

Fig 3 Expand

Fig 4.

Performance of LVH prediction model for gender.

(a) Receiver operating characteristic curve with 95% CI for predicting LVH in entire dataset. (b) Receiver operating characteristic curve with 95% CI for predicting LVH in male dataset. (c) Receiver operating characteristic curve with 95% CI for predicting LVH in female dataset. Abbreviations: AUROC, area under receiver operating characteristics curve; LVH, left ventricular hypertrophy; ECG, electrocardiography; CI, confidence interval.

More »

Fig 4 Expand

Table 2.

Performance metrics for classification task with Youden index.

More »

Table 2 Expand

Table 3.

Performance metrics for single-lead electrocardiography (lead I, II excerpt) with Youden index.

More »

Table 3 Expand

Fig 5.

Class activation maps.

(a, b) Grad-CAM for male with LVH class (≥132 g/m2). The ECGs of male subjects without LVH (<132 g/m2) in panel (a) attracted minimal attention from the model predicting LVH. On the other hand, the model showed strong attention to ECGs with LVH. (c, d) Grad-CAM for female with LVH class (≥109 g/m2). The ECGs of female subjects without LVH (<109 g/m2) in panel (c) attracted minimal attention from the model predicting LVH. On the other hand, the model showed strong attention to ECGs with LVH. Abbreviations: Grad-CAM, gradient-weighted class activation map; ECG, electrocardiography; LVH, left ventricular hypertrophy.

More »

Fig 5 Expand

Table 4.

Comparison with previous studies.

More »

Table 4 Expand

Table 5.

Acronym table.

More »

Table 5 Expand