Skip to main content

Table 3 Performance comparison DL model versus cardiologist on LQTS 1 and 2

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

Training

Type

External validation (Leuven data)

Sensitivity ± SD

Specificity ± SD

AUC ± SD

All ECG approach (Amsterdam data)

LQTS1

89 ± 4%

97 ± 3%

0.93 ± 0.02

LQTS2

91 ± 3%

87 ± 2%

0.89 ± 0.02

Expert cardiologist in LQTS

LQTS1

93%

90%

0.90

LQTS2

90%

80%

0.85

  1. A subset of 30 LQTS1, 30 LQTS2, and 300 controls (150 X 2) from the Leuven dataset was selected and used to validate our DL models, which were trained on the Amsterdam data using all acquired 12-lead ECGs (i.e., ALL ECG approach) per patient (top). The mean and standard deviation (SD) of the collected metrics is reported. The same subset was evaluated by an expert cardiologist in LQTS (bottom)