Table 3 Test accuracies and F1 Scores for LVH classification models

From: Deep echocardiography: data-efficient supervised and semi-supervised deep learning towards automated diagnosis of cardiac disease

Model

Test Accuracy

F1 Score

CNN (Stage 1)

84.07% (+/−1.23)

0.6027

CNN (Stage 2)

87.91% (+/−0.57)

0.7381

CNN with Segmentation (Stage 1)

81.32% (+/−0.98)

0.5952

CNN with Segmentation (Stage 2)

91.21% (+/−0.41)

0.8139

  1. In stage 1, weights for convolution layers were fixed and model was trained over 20 epochs. In stage 2, weights of fully-connected layers were fixed and convolution layers were fine-tuned.
Back to article page