Date of Completion

2024

Document Type

Honors College Thesis

Department

Department of Electrical and Biomedical Engineering

Thesis Type

Honors College

First Advisor

Dr. David Jangraw

Keywords

Machine Learning, Exposure Therapy, OCD, Fear, Distress

Abstract

Pediatric psychotherapy, specifically cognitive behavioral therapy (CBT), is a promising treatment for Obsessive-Compulsive Disorder (OCD). Clinical fear ratings obtained during pediatric psychotherapy provide valuable information about patient progress and have been associated with clinical outcomes. Evidence-based treatments rely on patient fear ratings, but no methods have been developed that can automatically extract fear ratings during psychotherapy. Advancements in computational methods have shown promise in detecting stress from written text that could be used to inform mental health diagnoses. We use a tool made to find the essential meaning of written text and machine learning models to predict patient fear ratings during 434 recorded pediatric OCD therapy sessions. Results indicate there are correlations between clinical fear ratings and the semantic features of words spoken during pediatric OCD therapy. Certain patients are prone to providing consistently low fear ratings, thus impacting the efficacy of machine learning algorithms applied to this dataset. When patients could be in the training and testing data of the machine learning models produced in this study, the area under the receiver operating characteristics (ROC) curve was higher than when patients were separated into either the training or testing data. These findings suggest that it may be possible to extract personality traits from the words spoken during therapy, indicating a correlation between the words spoken during therapy and patient fear.

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.

Download

Share

COinS