Related Experiment Video
Updated: Dec 8, 2025
09:04
Uncovering Beat Deafness: Detecting Rhythm Disorders with Synchronized Finger Tapping and Perceptual Timing Tasks
Published on: March 16, 2015
13.2K
Using behavioral rhythms and multi-task learning to predict fine-grained symptoms of schizophrenia
Vincent W-S Tseng1, Akane Sano2, Dror Ben-Zeev3
1Information Science, Cornell University, Ithaca, 14850, USA. vincent@infosci.cornell.edu.
Scientific Reports
|September 16, 2020
Summary
Behavioral rhythm sensing using smartphones and machine learning can predict schizophrenia symptom changes. This approach offers improved accuracy and interpretability for early intervention and patient monitoring.
Area of Science:
- Psychiatry
- Computational Neuroscience
- Digital Health
Background:
- Schizophrenia presents complex, dynamic symptoms, often accompanied by disrupted behavioral rhythms like sleep patterns.
- Behavioral rhythm disruptions are common in schizophrenia and may correlate with symptom fluctuations.
Purpose of the Study:
- To predict fine-grained symptom changes in schizophrenia using interpretable machine learning models based on behavioral rhythms.
- To leverage smartphone-based sensing and multi-task learning for personalized symptom prediction.
Main Methods:
- Extracted rhythm-based features from 6,132 days of data across 61 participants.
- Employed multi-task learning to predict ecological momentary assessment scores for 10 distinct symptom items.
- Utilized unsupervised clustering on model feature weights to identify symptom subtypes.
Main Results:
- Multi-task learning models significantly outperformed single-task models in predicting individual symptom trajectories (e.g., depression, social interaction, auditory hallucinations).
- Identified distinct patient and symptom subtypes through cluster analysis of model feature weights.
- Rhythm-based features enhanced prediction accuracy and model interpretability compared to previous methods.
Conclusions:
- Smartphone-based behavioral rhythm sensing with multi-task learning provides a powerful tool for understanding and predicting schizophrenia symptom dynamics.
- The findings enable personalized monitoring and early intervention strategies by elucidating the influence of behavioral rhythms on symptom conditions.
- This approach facilitates proactive management of schizophrenia, potentially preventing symptom deterioration with minimal additional burden on patients and clinicians.
