Depression screening using a non-verbal self-association task: A machine-learning based pilot study.

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Depression screening is crucial for early intervention.
• Non-verbal behavioral assessments show promise for supplementing traditional depression screening methods.
• Distinct neuroanatomical markers and behavioral patterns in depression patients suggest physiological differences.

Purpose of the Study:

• To explore machine learning methods for optimal depression screening using behavioral data.
• To validate a novel behavioral assessment focusing on self and emotions for depression screening.
• To identify interpretable behavioral signatures associated with depression.

Main Methods:

• Collected longitudinal behavioral assessment data from 84 participants across two sessions.
• Assessed depression using the Beck Depression Inventory II (BDI-II).
• Employed machine learning, including Gradient Boosting, for data analysis and prediction.

Main Results:

• The best machine learning model achieved a 10-Fold cross-validation Area Under the receiver operating characteristic Curve (AUC) of 0.90 and balanced accuracy of 0.81.
• Prospective prediction of depression status between sessions yielded a 10-Fold CV AUC of 0.77 and balanced accuracy of 0.66.
• Identified interpretable behavioral signatures indicative of depression.

Conclusions:

• Behavioral data presents a viable and cost-effective solution for depression screening.
• The findings support the utility of non-verbal assessments in clinical settings.
• This approach has potential for wide-ranging applications in mental health screening.