| Title |
Screening of ECG Change with Serum Potassium Level based on Deep Learning for Monitoring Hyperkalemia |
| Authors |
문병진(Byung-Jin Moon) ; 변준(Joon Byun) ; 박영철(Young-cheol Park) ; 육현(Hyun Youk) ; 이희영(Hee Young Lee) ; 추연일(Yeon II Choo) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.2.59 |
| Keywords |
Hyperkalemia; ECG; Deep learning; Depthwise separable convolution |
| Abstract |
When serum potassium level (SPL) is above 5.5 mEq/L, hyperkalemia is diagnosed. Since it can affect heart rate and cause even heart failure, a quick warning is crucial. The increase of SPL typically causes deformations of electrocardiogram (ECG). Thus, in this paper, we propose a deep-learning model that can analyze ECG change with SPL and warn patients about the risk of hyperkalemia. In this paper, we adopt a convolutional recurrent neural network (CRNN) model based on a depthwise separable convolution kernel together with a long short-term memory (LSTM). Experiments conducted with 1,879 patients show that the proposed deep-learning model can relatively accurately predict SPLs, and, thanks to the depthwise separable convolutional kernel, it can maintain the accuracy even with small network parameters. |