Characterizing Electrodermal Activity Using Point Processes in Young Children

Oct. 1, 2018


Subramanian, S., Barbieri, R., Johnson, K.T., Brown, E. "Characterizing Electrodermal Activity Using Point Processes in Young Children," 50th Annual Meeting of the Biomedical Engineering Society (BMES), Atlanta, Georgia, October 2018.


The sympathetic nervous system governs the body’s response to stressful situations. In children, stress can have lasting developmental impact. Tracking sympathetic activity in real-time may shed light on how young children respond to stress.

Electrodermal activity, or EDA, is a measure of sweat gland activity, which is uniquely controlled by the sympathetic nervoussystem. Because it can be measured non-invasively, it lends itself well to the monitoring of sympathetic dynamics in young children. However, current EDA analysis algorithms do not accurately describe the statistical time-varying properties of the underlying physiological processes.

Materials and Methods
We analyzed ~100 minutes of EDA data from each of three children ages 2-6, collected via a wireless wrist-worn EDA sensor using wet electrodes on the inner wrist of the non-dominant hand. All data were acquired during naturalistic play scenarios, following IRB-approved protocol. EDA is a pulsatile signal consisting of discrete events of sweat release at the skin surface due to increased nerve stimulation of underlying sweat glands. The discrete, pulsatile nature of EDA makes it a good physiological candidate for point process analysis approaches. For each of the three subjects, we first preprocessed the data by removing artifacts and the underlying slow-moving tonic component of the signal and identified EDA pulse peaks by prominence compared to local surroundings. We then computed and fit inter-pulse intervals to several right-skewed distributions, including the exponential, inverse Gaussian, Rayleigh, lognormal, and generalized inverse Gaussian, to assess the presence of regular statistical structure. We computed Kolmogorov-Smirnov (KS) distances as the maximum vertical distance between the theoretical and empirical cumulative distribution functions (CDFs) and compared them to significance thresholds. Finally, we fit a history-dependent inverse Gaussian model for each subject to track instantaneous estimates of mean and standard deviation of ‘EDA pulse rate’ and validated it with a formal goodness-of-fit analysis.

Results and Discussion
Figure 1A shows Kolmogorov-Smirnov (KS) plots for the inter-pulse intervals of one subject compared to three distributions. The KS distances and significance threshold are given. If the KS distance is under the threshold, the distribution offers a good explanation for the structure in the data. Based on Figure 1A, both the inverse Gaussian and lognormal distributions fit well to the data, while the exponential does not. This was true for all three subjects. No one has previously identified such structure in EDA, and this suggests that the physiology of EDA in young children can be described using point process approaches and very few parameters. Finally, Figure 1B shows an example of instantaneous dynamics computed for the same subject using a history-dependent inverse Gaussian model. Using this model, the level of and variability in sympathetic activity (mean and standard deviation of pulse rate respectively) can be tracked instantaneously in young children.

Figure 1. (A) The KS plots for one subject for three distributions: inverse Gaussian, lognormal, and exponential. Comparing the KS-distances to the significance threshold, the data fits well to the inverse Gaussian and lognormal, but not to the exponential. (B) Dynamic model output for a single subject, showing instantaneous estimates for mean and standard deviation of EDA pulse rate

This exploratory study highlights three main results: 1) Novel statistical structure was found in EDA data from three young children; 2) this structure suggests that EDA in children can be described with few parameters as a point process; and 3) this structure can be used to track instantaneous sympathetic dynamics in children. Further work is necessary using more subjects to validate these conclusions and analyze sympathetic dynamics in response to specific stimuli and in other contexts, such as autism or attention deficit disorders.

Related Content