Comparison of simple algorithms for estimating respiration rate from electrical impedance pneumography signals in wearable devices

Respiration rate (RR) is considered as a useful parameter in characterizing the health condition of a person. Among the methods used for respiration measurement, Electrical Impedance Pneumography (EIP) can be easily obtained in wearable applications due to the possibility of using the electrocardiography (ECG) electrodes for the EIP measurement. In the fast growing field of wearable devices, having clinically valuable and reliable information along with providing the convenience of the user, is probably the most important and challenging issue. To address the need of small sized devices for ECG (and EIP) measurements, EASI electrode configuration is an acceptable solution. The signals from EASI system not only provide useful information by themselves when directly used for cardiological analyses, but can also be converted to the standard 12-lead ECG information. With aforementioned advantages of EASI system, the question then arises how suitable the electrode locations of the system are for EIP measurements and what algorithms perform better for respiration rate derivation. In this work, we evaluated eight methods for deriving respiration rate from EIP signals measured from 15 subjects (10 males +5 females) in three conditions: standing, walking slowly, and walking fast. The algorithms were autoregressive (AR) modeling (three different approaches), Fast Fourier Transform (FFT), autocorrelation, peak detection and two counting algorithms. Our results show that advanced counting method is the most promising approach among the ones studied in this work. For this algorithm, the concordance correlation coefficients of the respiration rate estimates between EIP and the reference measurement were 0.96, 0.90 and 0.97 for standing, walking with 3 km/h speed, and walking with 6 km/h speed, respectively.