Automated Affect Detection in Deep Brain Stimulation for Obsessive-Compulsive Disorder: A Pilot Study.

Area of Science:

• Neuroscience
• Psychiatry
• Computer Science

Background:

• Automated measurement of affective behavior in psychopathology is mainly used for screening and diagnosis.
• Clinicians need objective biometrics to assess treatment response, especially for neural implant therapies.
• Deep brain stimulation (DBS) is a treatment for intractable conditions like obsessive-compulsive disorder (OCD).

Purpose of the Study:

• To use automated face analysis to evaluate treatment response to DBS in two patients with intractable OCD.
• To assess the impact of DBS on affective behavior, quantifying positive and negative valence.
• To test the domain transfer capability of convolutional neural network (CNN) classifiers trained on normative data.

Main Methods:

• Automated face analysis was employed to quantify positive and negative affect.
• A CNN, trained on data from 160 non-OCD participants, was used for affect quantification.
• Two patients with intractable OCD were assessed intraoperatively and three months post-implantation, during both DBS-on and DBS-off conditions.

Main Results:

• DBS-on was associated with markedly positive affect in both patients and contexts.
• When DBS was turned off, affect flattened, and negative affect increased in the outpatient setting.
• The CNN classifiers demonstrated successful domain transfer with a mean Area Under the Curve (AUC) of 0.87.

Conclusions:

• Parametric variation of DBS is strongly linked to affective behavior in OCD patients.
• Discontinuation of DBS may lead to a vulnerability for negative affect.
• Automated face analysis provides objective, repeatable biometrics for assessing DBS treatment response and affective changes.