A multi-paradigm and longitudinal EEG dataset including the “sixth-finger” and “affected-h

A longitudinal dataset covering pre-training, post-training, and follow-up periods for 24 stroke patients performing motor imagery tasks. It includes raw EEG data, preprocessed data, and patient clinical information, collected by Zhuang Wang and last updated in May 2026. Preliminary analysis using CSP+SVM and CSP+LDA algorithms reported an average classification accuracy of approximately 85-86% between the two motor imagery paradigms.

Use Cases

• Training and evaluating motor imagery classification algorithms based on EEG signals from stroke patients.
• Studying longitudinal neuroplasticity changes based on data collected across multiple rehabilitation stages.
• Developing personalized stroke rehabilitation feedback systems based on the novel 'sixth-finger' and affected-hand motor imagery paradigms.
• Benchmarking machine learning models for EEG signal processing based on the provided preprocessed data.

Strengths

• Includes data from 24 stroke patients, providing a specific clinical population sample.
• Covers three longitudinal stages of rehabilitation: pre-training, post-training, and follow-up.
• Provides both raw EEG data and preprocessed data, facilitating different analysis approaches.
• Preliminary validation shows an average classification accuracy of 85-86% between the two motor imagery tasks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale model training.
• Data may reflect bias inherent to the specific collection context and patient cohort.

Provenance

Source

figshare

Collection Method

Collected from 24 stroke patients during motor imagery tasks in a study by Zhuang Wang.

Time Range

Covers pre-training, post-training, and follow-up periods of stroke rehabilitation.

Freshness

Last updated 2026-05-23 05:48:54; freshness should be verified.