PhD Scientific Days 2024

Poster Session G - Mental Health Sciences 2.

Non-verbal Cues in Depression: A Multimethod Approach Research Proposal

Text of the abstract

Introduction
The analysis of non-verbal communication has always had a significant role in psychiatric care. In clinical settings, professionals pay attention not only to verbal communication but to the non-verbal aspects of communication as well. Depressed individuals often exhibit a narrowed emotional spectrum, resulting in altered emotional expressiveness. These changes are noticeable during diagnostic interviews and check-ups, but simple observation is often distorted by the subjective perception of the clinician.
With the advancement of computer science, the integration of digital technology enables monitoring of changes in motor performance, vocal patterns, and facial expressions all together. This change in method makes the analysis of non-verbal cues objective.

Aims
Our study seeks to investigate the non-verbal manifestations of depression across three dimensions: motor performance, speech patterns, and facial expressions, compared with those of a healthy control group. Using machine learning algorithms, we aim to develop a digital framework capable of differentiating between depressed and non-depressed non-verbal behaviors.

Method
Participants will engage in various motor tasks, followed by a clinical interview within an experimental setup. The entire session will be audio-visually recorded for subsequent digital analysis of motor behavior, speech dynamics, and facial alterations.

Results
Based on existing literature in this field, finding differences between groups in all three channels is expected. Depressed individuals are anticipated to exhibit slower movements with reduced gestural frequency during speech. In the facial channel, we expect more negative emotions and fewer emotions overall in the patient group compared to controls. In terms of vocal patterns, we predict depressed individuals to demonstrate decreased frequency and intensity, prolonged pauses, and reduced articulation rates.

Conclusion
The anticipated results will shed more light on the non-verbal patterns of depression. These findings will facilitate the development of a profile including information from these three channels that’ll help machine learning distinguish between depressed and healthy individuals, thus helping the diagnostic process.

Funding
Financial support was provided by the National Research, Development, and Innovation Office of Hungary (ÚNKP-23-3).