Accuracy enhancement in low frequency gain and phase detector (AD8302) based bioimpedance spectroscopy system

Abstract

Bioimpedance spectroscopy (BIS) is the measurement of the impedance of biological tissues. In the development of a low cost, portable BIS system the gain and phase difference detection (GPD) method is usually used. However this method suffers from low accuracy at low frequencies due to the presence of spikes or peaks. Detection and removal of these peaks will improve the performance of the GPD based BIS system at low frequencies. Peak detection in physiological signals is a well developed area. A lot of algorithms have been developed in the past for the determination of peaks in signals and their time of occurrence. For real-time processing, microcontroller based peak detection algorithms have also been developed in the past. In this paper, the problem of the GPD method at low frequency is examined closely and then three algorithms were used to mitigate it. The first is based on a moving average filter, the second is a simple peak detection and elimination (SPD) and the last is a peak detection based on the frequency of the input signal to the GPD (SFPD). These algorithms were evaluated with both simulated and measured data. Four parameters were used to indicate the performance of the algorithms; run-time, RMSE, sensitivity and positive predictability. The RMSEs were less than 0.17 for moving average, 0.07 for SPD and 0.08 for SFPD. The run-times was less than 10 ms for SFPD while that for the SPD and moving average were around 30 ms and 80 ms respectively. In all it was found that the algorithms based on peak detection have better results. The computational simplicity of the algorithms makes them suitable for microcontroller based implementation.