The aim of this study was to reconstruct a 12-lead electrocardiograph (ECG) with a universal transformation coefficient and find the appropriate electrode position and shape for designing a patch-type ECG sensor. A 35-channel ECG monitoring system was developed, and 14 subjects were recruited for the experiment. A feedforward neural network with one hidden layer was applied to train the transformation coefficient. Three electrode shapes (5 cm × 5 cm square, 10 cm × 10 cm square, and right-angled triangle) were considered for the patch-type ECG sensor. The mean correlation coefficient (CC) and minimum CC methods were applied to evaluate the reconstruction performance. The average CCs between the standard 12-lead ECG and reconstructed 12-lead ECG were 0.860, 0.893, and 0.893 for a 5 cm × 5 cm square, 10 cm × 10 cm square, and right-angled triangle shape. The right-angled triangle showed the highest performance among the considered shapes. The results also suggested that the bottom of the central area of the chest was the most suitable position for attaching the patch-type ECG sensor.
Keywords: 12 lead ECG; artificial neural network; reconstructed electrocardiogram; universal coefficient; wearable patch device.