โก Quick Summary
This study developed an interpretable AI model for detecting depressive symptoms through actigraphy data from 3,304 participants. The model achieved impressive classification performance, with AUROC values of 0.679 and 0.715 for mild and moderate-to-severe depressive symptoms, respectively.
๐ Key Details
- ๐ Dataset: 3,304 participants over one week
- ๐งฉ Features used: Absolute and relative activity indicators, daytime light intensity
- โ๏ธ Technology: Machine learning models including CatBoost (CB) and XGBoost (XGB)
- ๐ Performance: AUROC values of 0.679 (CB) and 0.715 (XGB)
๐ Key Takeaways
- ๐ Actigraphy data can effectively classify depressive symptoms.
- ๐ก Machine learning enhances the accuracy and explainability of mental health assessments.
- ๐ฉโ๐ฌ Low step counts were significant predictors of depressive symptoms.
- ๐ Older men showed the highest predictive accuracy for depressive symptoms.
- ๐ Temporal patterns of activity, particularly during dawn, are crucial for classification.
- ๐ค SHAP-based analysis provides insights into model decision-making.
- ๐ Study published in the Journal of Affective Disorders.
- ๐ PMID: 40850556.
๐ Background
Depression is a prevalent mental health issue that often goes undetected, leading to significant personal and societal burdens. Traditional assessment methods can be subjective and inconsistent. The integration of actigraphyโa method of monitoring physical activityโinto machine learning models offers a promising avenue for more objective and accurate detection of depressive symptoms.
๐๏ธ Study
Conducted with a diverse cohort of 3,304 participants, this study aimed to leverage actigraphy data collected over a week to classify individuals into depressive and non-depressive symptom groups. The researchers employed six machine learning models, focusing on the performance of CatBoost (CB) and XGBoost (XGB) in predicting depressive symptoms based on activity patterns.
๐ Results
The study found that both CB and XGB models exhibited strong classification performance, with AUROC values of 0.679 for mild depressive symptoms and 0.715 for moderate-to-severe symptoms. Notably, the analysis revealed that low step counts and high activity levels during the least active hours were significant indicators of depressive symptoms, particularly in older men, who demonstrated the highest predictive accuracy (AUROC values of 0.756 and 0.833).
๐ Impact and Implications
This research underscores the potential of using actigraphy-derived data in AI-driven mental health assessments. By focusing on both absolute step counts and temporal activity patterns, the study advocates for the development of time-sensitive, explainable AI approaches that can enhance personalized mental health screening. Such advancements could lead to earlier detection and intervention for individuals at risk of depression, ultimately improving mental health outcomes.
๐ฎ Conclusion
This study highlights the transformative potential of integrating actigraphy data with machine learning for detecting depressive symptoms. The findings advocate for a shift towards more objective and explainable methods in mental health assessments, paving the way for personalized approaches that can significantly enhance patient care. Continued research in this area is essential for refining these technologies and expanding their applications in mental health.
๐ฌ Your comments
What are your thoughts on using actigraphy and machine learning for mental health assessments? We would love to hear your insights! ๐ฌ Leave your comments below or connect with us on social media:
Actigraphy-based step analysis for the detection of depressed mood: An explainable machine learning approach.
Abstract
INTRODUCTION: Our study aims to develop an interpretable artificial intelligence (AI) model for detecting depressive symptoms using actigraphy, integrating statistically significant features into machine learning models to enhance accuracy and explainability.
METHODS: We analyzed actigraphy data from 3304 participants over a one-week period, classifying them into a depressive symptom group and a non-depressive symptom group. Six machine learning models, including CatBoost (CB) and XGBoost (XGB), were trained using absolute activity indicators based on three-hour intervals, relative activity indicators (nonparametric, Cosine analysis), and daytime light intensity exposure duration. Shapley additive explanations (SHAP)-based explainability analysis was applied, and models were stratified by and gender.
RESULTS: CB and XGB demonstrated the highest classification performance in predicting mild and moderate-to-severe depressive symptoms, respectively, with AUROC values of 0.679 and 0.715 across 10 random-seed models evaluated on a fixed test set. SHAP-based explainability analysis revealed that low step counts contributed to the prediction of depressive symptoms, high average activity during the least active 5-hour period of the day (L5) was associated with depressive symptoms, and we found that the onset of L5 was mainly distributed in the dawn hours. Importantly, model performance differed across demographic groups, with the highest predictive accuracy achieved in older men (AUROCโฏ=โฏ0.756 and 0.833, respectively) for all depressive symptoms.
CONCLUSION: This study highlights the potential of actigraphy-derived step count data in AI-driven classification of depressive symptoms. Both absolute step count and temporal patterns contribute to classification, emphasizing the need for time-sensitive, explainable AI approaches for personalized mental health screening.
Author: [‘Kim JW’, ‘Lee T’, ‘Lim B’, ‘Park SH’, ‘Park JH’, ‘Jeong I’, ‘Park K’, ‘Kang HJ’, ‘Jeon E’, ‘Kim SW’, ‘Jhon M’, ‘Lee H’, ‘Kim JM’]
Journal: J Affect Disord
Citation: Kim JW, et al. Actigraphy-based step analysis for the detection of depressed mood: An explainable machine learning approach. Actigraphy-based step analysis for the detection of depressed mood: An explainable machine learning approach. 2025; (unknown volume):120104. doi: 10.1016/j.jad.2025.120104