BME Frontiers / 2022 / Article / Tab 1

Research Article

Automatic Detection of Atrial Fibrillation from Single-Lead ECG Using Deep Learning of the Cardiac Cycle

Table 1

Reconstruction error of different encoders and different numbers of dimensions of the latent trajectories based on approximately 318000 heartbeat events. Reconstruction error is measured as root-mean-square error (RMSE, equation (6)) in mV. Average RMSE over all training sequences is presented with 95% confidence interval, assuming an exponential distribution of the error values.

Encoder

DHB136.1 [11.2–480.9]129.2 [9.1–460.6]135.7 [8.9–485.5]1927.4 [53.0–7098.1]
S-DHB140.4 [11.9–494.8]137.5 [7.7–495.5]123.1 [8.6–438.9]161.3 [10.3–577.7]
P-DHB54.4 [5.7–188.7]29.8 [3.8–101.5]19.5 [3.0–65.1]18.2 [3.2–59.4]