Different strategies for improving the quality of the ECG signal recorded by the armband were investigated, and different morphological features were studied for artifact detection. An artifact detector based on machine learning was developed. 24-hours recordings were analyzed, and obtained results suggest that the armband device is suitable for a daily life monitoring, obtaining usable data during more than half of the non-bedtime and almost all the bedtime. Five traditional heart rate variability (HRV) were also analyzed. The obtained Pearson’s correlation coefficient between the measurements from the armband device and the measures from a commercial Holter were higher than 0.99 suggesting that the quality of armband-ECG signals is enough to estimate HRV parameters during stationary movement restricted conditions.
A study on deriving respiratory rate using the wearable armband was performed. The study included paced breathing stages at 10, 12, 18, 24, and 30 breaths per minute, as well as an stage during spontaneous breathing. Respiratory rate was estimated from the armband-ECG signals, and the estimates were compared to those obtained from the respiration signal. The obtained relative error that was not higher than 1.25% in median and not higher than 2.78% for all the studied stages. In addition, the methods were modified in order to reduce their computational time. The adaptation reduced the computational time up to a 82.05% assuming a lower accuracy, specifically, the highest median of relative error increased to -1.33%, and its highest IQR increased to 4.22%. Moreover, an algorithm for tracking tidal volume by using the ECG was developed. In this case, the amplitude of the EDR signals was studied as a surrogate of tidal volume. Obtained correlations between EDR amplitude and tidal volume were higher than 0.9 for some EDR methods, suggesting that it is feasible to track tidal volume changes with the armband. These results are very promising and allow us to consider the armband wearable device as a potential wearable cardiorespiratory monitor.
Alternatively, PPG signals were analyzed also, as this technology is the basis of many wearable devices in the market. Novel methods for deriving respiratory rate from PPG signals were proposed focused on noisy environments. Furthermore, methods for estimating baroreflex sensitivity from PPG signal were developed, leading to another source of information about autonomic nervous system.
An Android application that records that PPG and stores it in “the cloud” (once anonymized) was created. Simultaneous smartphone-PPG recordings and conventional PPG recordings are being performed in order to validate the Android application.
WECARMON’s results were disseminated within 13 international conference proceedings and 16 scientific journal articles, that can be found on the ‘Results’ tab.
