JMIR Publications

ABSTRACT

Background:

Even before the onset of the COVID-19 pandemic, children and adolescents were experiencing a mental health crisis, partly due to a lack of quality mental health services. The rate of suicide for black youth has increased by 80%. By 2025, the healthcare system will be short of 225K therapists, further exacerbating the current crisis. Therefore, it's of the utmost importance for providers, schools, youth mental health, and pediatric medical providers to integrate innovation in digital mental health to identify problems proactively and rapidly for effective collaboration with other healthcare providers. Such approaches can help identify robust, reproducible, and generalizable predictors and digital biomarkers of treatment response in psychiatry. Among the multitude of digital innovations to identify a biomarker for psychiatric diseases currently, as part of the macro-level digital health transformation, speech stands out as an attractive candidate with features such as affordability, non-invasive, and non-intrusive.

Objective:

The protocol aims to develop speech-emotion recognition algorithms leveraging AI/ML which can establish a link between trauma, stress, and voice types, including disrupting speech-based characteristics, and detect clinically relevant emotional distress and functional impairments in children and adolescents.

Methods:

Informed by the theoretical foundations (the Theory of Psychological Trauma Biomarkers and Archetypal Voice Categories), we developed our methodology to focus on five emotions: Anger, Happiness, Fear, Neutral, and Sadness. Participants will be recruited from Georgia HOPE and Family Ties in Atlanta, GA. Speech samples, along with responses to the SFSS, PHQ-9, and ACE scales, will be collected using an Android mobile app. Our model development pipeline is informed by Gaussian Mixture Model, Recurrent Neural Network, and Long Short-Term Memory.

Results:

We tested our model with a public dataset. The GMM with 128 clusters showed an evenly distributed accuracy across all five emotions. Using utterance-level features, GMM achieved an accuracy of 79.15% overall, while frame selection increased accuracy to 85.35%. This demonstrates that GMM is a robust model for emotion classification of all five emotions and that emotion frame selection enhances accuracy, which is significant for scientific evaluation.

Conclusions:

This study contributes to the literature as it addresses the need for reliable and objective tools to detect clinically relevant emotional distress and functional impairments in children and adolescents. The results show that our algorithm has the potential to aid mental health providers in diagnosis and treatment planning for pediatric patients.