Comparison of Wearable and Clinical Devices for Acquisition of Peripheral Nervous System Signals

Figure 1. Illustration showing the devices used for the acquisition of physiological signals. A: Thought Technology FlexComp (picture from www.thoughttechnology.com) with four wired sensors: (a) electrodermal activity, (b) blood volume pulse, (c) electrocardiogram, (d) respiration; B: Empatica E4 (picture from www.empatica.com); C: ComfTech HeartBand.

Figure 2. Portions of signals from wearable devices (WDs) extracted from the Wearable and Clinical Signals (WCS) dataset to show examples of high (left column) and low (right column) quality signals. (A): electrocardiogram (ECG) signals collected by the HeartBand; (B) blood volume pulse (BVP) signals collected by the E4; (C) electrodermal activity (EDA) signals collected by the E4.

Figure 3. Data analysis procedure adopted in this study. IBI: inter-beat intervals; RMSSD: root of the mean of the squares of subsequent differences; HF: high frequency.

Figure 4. Signal quality indicators (SQIs) of the cardiac signals: (A). Kurtosis and spectral power ratio of the ECG collected from the FlexComp (gray) and Heart Band (white); (B). Kurtosis and spectral power ratio of the BVP collected from the FlexComp (gray) and E4 (white). Striped areas indicate the ranges of SQI values associated with low signal quality.

Figure 5. Metrics of the quality of the inter-beat intervals extracted from the cardiac signals.

Figure 6. Bland–Altman plots to assess the reliability of the physiological indicators computed from the IBI signals extracted from different signals and devices. All plots refer to the indicators computed from the ECG signals collected by the FlexComp.

Figure 7. SQIs of the EDA signals collected from the FlexComp (gray) and E4 (white). Striped areas indicate the ranges of SQI values associated with low signal quality.