Affective State during Physiotherapy and Its Analysis Using Machine Learning Methods.

Area of Science:

• Psychophysiology
• Affective Computing
• Biomedical Engineering

Background:

• Healthcare procedures can induce significant emotional responses.
• Technological advancements allow for continuous monitoring of psychophysiological functions.
• Understanding affective states is crucial for patient well-being and care.

Purpose of the Study:

• To analyze the individual's affective state during potentially uncomfortable situations.
• To investigate the correlation between physiological signals and subjective emotional reports.
• To apply machine learning for classifying affective states based on psychophysiological data.

Main Methods:

• Collected electrodermal activity, cardiac, and accelerometric signals from 49 participants.
• Administered psychological questionnaires including JAWS (affective state), DST (cognitive skills), and VFT (verbal fluency).
• Utilized statistical analysis and machine learning (kNN classifier with JMI and PCA feature selection) for data processing.

Main Results:

• The kNN classifier combined with the JMI feature selection method achieved the highest accuracy (81.63%) in classifying affective states based on JAWS test results.
• Classification sensitivity reached 85.71%, and specificity was 71.43%.

Conclusions:

• Psychophysiological signals, when analyzed with machine learning, can effectively classify affective states.
• This approach offers a potential non-invasive method for monitoring patient emotions in healthcare settings.
• Further research can refine these methods for improved accuracy and broader clinical application.