MIMIC PERform: Physiological Signals from Critically-Ill Patients

Recordings from critically-ill patients during routine clinical care, extracted from the MIMIC III Waveform Database. The collection includes separate datasets for atrial fibrillation detection (35 subjects) and training/testing (200 subjects each, split between adults and neonates). The data was made available by Peter Charlton of the University of Cambridge.

Use Cases

• Detecting atrial fibrillation based on electrocardiogram (ECG) and photoplethysmogram (PPG) signals.
• Developing signal quality assessment algorithms based on the included sample of noisy data.
• Benchmarking beat detection algorithms using the PPG signals as described in the associated publication.
• Comparing physiological patterns between adult and neonate patient groups.

Strengths

• Includes 35 subjects specifically categorized for atrial fibrillation (AF) research (19 AF, 16 non-AF).
• Provides training and testing sets of 200 patients each, with a balanced split of 100 adults and 100 neonates.
• Offers data in multiple formats (Matlab, WFDB, CSV) for accessibility.
• Contains curated 1-minute samples for specific conditions (AF, neonate, noisy) to aid initial exploration.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

MIMIC III Waveform Database

Collection Method

Extracted from physiological signals recorded during routine clinical care.