Abstract
According to the World Health Organization, the number of mental disorder patients, especially depression patients, has grown rapidly and become a leading contributor to the global burden of disease. However, the present common practice of depression diagnosis is based on interviews and clinical scales carried out by doctors, which is not only labor-consuming but also time-consuming. One important reason is due to the lack of physiological indicators for mental disorders. With the rising of tools such as data mining and artificial intelligence, using physiological data to explore new possible physiological indicators of mental disorder and creating new applications for mental disorder diagnosis has become a new research hot topic. However, good quality physiological data for mental disorder patients are hard to acquire. We present a multi-modal open dataset for mental-disorder analysis. The dataset includes EEG and audio data from clinically depressed patients and matching normal controls. All our patients were carefully diagnosed and selected by professional psychiatrists in hospitals. The EEG dataset includes not only data collected using traditional 128-electrodes mounted elastic cap, but also a novel wearable 3-electrode EEG collector for pervasive applications. The 128-electrodes EEG signals of 53 subjects were recorded as both in resting state and under stimulation; the 3-electrode EEG signals of 55 subjects were recorded in resting state; the audio data of 52 subjects were recorded during interviewing, reading, and picture description. We encourage other researchers in the field to use it for testing their methods of mental-disorder analysis.
The Need for Objective Biomarkers in Psychiatry
Currently, diagnosing Major Depressive Disorder (MDD) is a subjective and labor-intensive process. It relies heavily on clinical interviews and self-reporting scales, such as the PHQ-9, which can be influenced by patient bias or a clinician's experience level. The emerging field of "computational psychiatry" seeks to address this by using machine learning to identify physiological biomarkers—objective signatures in brain activity or speech that correlate with mental health states.
However, researchers face a significant bottleneck: high-quality, open-access data from clinically verified patients is scarce. Most existing datasets are small, proprietary, or lack the rigorous clinical "ground truth" necessary for training reliable models. The MODMA dataset aims to bridge this gap by providing a high-quality repository of EEG and audio data.
Methodology and Data Collection
The researchers established a rigorous protocol to ensure data quality and clinical relevance. Every patient in this dataset was diagnosed by professional psychiatrists using the MINI (Mini-International Neuropsychiatric Interview) and DSM-IV criteria.
EEG Acquisition
Two distinct setups were utilized for EEG collection:
- High-Density (128-electrode): A traditional elastic cap used for laboratory-grade research.
- Pervasive (3-electrode): A novel, wearable device targeting the prefrontal lobe (Fp1, Fpz, Fp2). This device is designed for real-world applications where high-density caps are impractical.
Experimental Paradigms
Participants were recorded during both resting states and active tasks. The resting state involved 5 minutes of eyes-closed activity. For active stimulation, the Dot-Probe Task was employed. This classic cognitive psychology task measures attentional bias by having participants react to dots appearing behind emotional (fear, sad, happy) or neutral face pairs.
Audio Tasks
Audio data was collected during three distinct speaking patterns: structured interviews (18 questions), reading a short story ("The North Wind and the Sun"), and describing pictures with varying emotional valences.
Dataset Characteristics
The MODMA dataset includes:
- 128-electrode EEG: 53 subjects (24 MDD, 29 Healthy Controls).
- 3-electrode EEG: 55 subjects (26 MDD, 29 Healthy Controls).
- Audio Data: 52 subjects (23 MDD, 29 Healthy Controls), with 29 recordings per subject.
Crucially, data was collected prior to patients taking medication. This is vital because psychotropic drugs can alter brain activity and mask underlying physiological indicators.
Key Insights and Validation
The authors validated the data quality through several analyses:
- Alpha Rhythm Asymmetry: Using the 3-electrode pervasive device, researchers found a statistically significant difference in the relative power of alpha rhythms between the left and right brain in MDD patients—a difference absent in healthy controls. This validates that even a minimal setup can capture meaningful biomarkers.
- Audio Quality: Recordings achieved a Signal-to-Noise Ratio (SNR) between 20 and 30, ensuring suitability for sensitive acoustic analysis.
- Task Synchronization: The Dot-probe task data includes precise timestamps for triggers, enabling robust event-related potential (ERP) analysis.
Conclusion and Future Work
The MODMA dataset provides a rare and valuable resource for the development of AI-driven mental health diagnostics. While the current sample size of approximately 50 subjects is a strong foundation, future work will focus on expanding the cohort and including other disorders such as schizophrenia, anxiety, and mania. Additionally, future iterations are expected to incorporate eye-movement tracking, facial expressions, and MRI data to provide a truly multi-modal perspective on mental health.
Summary updated