Automatic identification of artifacts in electrodermal activity data.

Area of Science:

• Psychophysiology
• Biomedical Engineering
• Machine Learning

Background:

• Wearable devices enable long-term, ambulatory measurement of electrodermal activity (EDA).
• Ambulatory EDA recordings are susceptible to noise and artifacts that can be misinterpreted as physiological responses.
• Current analysis methods lack automated artifact detection, hindering data accuracy.

Purpose of the Study:

• To develop and evaluate a machine learning algorithm for automatic detection of artifacts in ambulatory EDA data.
• To provide a reliable method for distinguishing physiological signals from recording artifacts.
• To enhance the accuracy and efficiency of EDA data analysis.

Main Methods:

• Development of a machine learning algorithm trained on EDA data.
• Implementation of artifact detection and classification techniques.
• Empirical evaluation of the algorithm's classification performance.

Main Results:

• Successful development of a machine learning algorithm for automatic EDA artifact detection.
• Demonstrated empirical classification performance for artifact identification.
• Creation of a freely available web-based tool for artifact and peak detection.

Conclusions:

• The developed machine learning algorithm effectively detects artifacts in ambulatory EDA recordings.
• This automated approach addresses a critical need in physiological data analysis.
• The freely available tool supports researchers in obtaining more accurate EDA measurements.