Skip to main content

Table 1 Model performances on Amsterdam data

From: A deep learning approach identifies new ECG features in congenital long QT syndrome

Training Type Internal validation (Amsterdam data)
Sensitivity ± SD Specificity ± SD AUC ± SD
First ECG approach (Amsterdam data) LQTS1 79 ± 9% 96 ± 1% 0.88 ± 0.04
LQTS2 89 ± 7% 90 ± 3% 0.89 ± 0.03
LQTS3 67 ± 18% 90 ± 9% 0.79 ± 0.05
All ECG approach (Amsterdam data) LQTS1 84 ± 2% 96 ± 2% 0.90 ± 0.02
LQTS2 90 ± 2% 95 ± 1% 0.92 ± 0.01
LQTS3 87 ± 6% 92 ± 4% 0.89 ± 0.03
  1. The mean of the collected metrics and the corresponding standard deviation (SD) of the 5-fold cross-validation is reported. First ECG approach: the DL models were trained using the first acquired 12-lead ECGs. All ECG approach: the DL models were trained using all acquired 12-lead ECGs (not only the first acquired) per patient