JMIR Publications

ABSTRACT

Background:

Obsessive compulsive disorder (OCD) in youth is characterized by behaviors, emotions, physiological reactions, and family interaction patterns. Machine learning (ML) models are available that can use behavioral, physiological, and speech signals that reflect autonomic nervous system activation to differentiate individuals with and without psychiatric symptoms and classify clinical severity with relatively high levels of accuracy. An outstanding question is whether ML models are accurate enough to be used as reliable, objective assessment tools in clinical practice. Furthermore, is it plausible that an automatic assessment tool using physiological signals from a wearable biosensor would enable continuous symptom monitoring inside and outside of the clinic. However, in-the-wild signal processing is still a challenge. Thus, we aim to develop ML models that can detect and predict OCD episodes and objectively classify clinical severity in and outside of the clinic.

Objective:

Thus, we aim to develop ML models that can detect and predict OCD episodes and objectively classify clinical severity in and outside of the clinic.

Methods:

In a pilot study, 10 children and adolescents with OCD and one of their parents and 10 children and adolescents with no psychiatric diagnoses and one of their parents wear biosensors on their wrist that measures pulse, electrodermal activity, skin temperature, and acceleration. Patients and their parents mark OCD episodes and controls mark fear episodes. Continuous, in-the-wild, data collection lasts eight weeks. Controlled experiments designed to link physiological signals to mental states are conducted at baseline and after eight weeks. Interpersonal interactions in the experiments are filmed and coded for behavior. The films are also processed with computer vision and for speech signals. Participants complete clinical interviews and questionnaires at baseline, Week 4, 7, and 8.

Results:

The purpose of the Wrist Angel pilot study is to test the feasibility of implementing psychotherapy and research enhanced with artificial intelligence in a public child and adolescent mental health center. If we find the study feasible in terms of recruitment, retention, biosensor functionality and acceptability, adherence to wearing the biosensor, safety related to the biosensor, and using physiological, audio, and visual signals as markers of OCD further studies will be conducted to integrate biosensor, audio and visual sensor output into ML algorithms that can provide patients, parents and therapists with actionable insights into their OCD symptoms and treatment progress.

Conclusions:

Future definitive studies will be tasked with testing the accuracy of ML models to detect and predict OCD episodes and classify clinical severity. Clinical Trial: ClinicalTrials.gov: NCT05064527, registered October 1, 2021.