SPaRCNet Data: 50,697 Labeled ICU EEG Samples for Seizure Pattern Classification

50,697 labeled electroencephalography (EEG) samples from 2,711 patients and 6,095 EEGs, annotated by physician experts from 18 institutions. The dataset was used to train SPaRCNet, a deep neural network for classifying seizures and rhythmic/periodic patterns in ICU EEG data. It is hosted on the Brain Data Science Platform under a restricted health data license.

Use Cases

• Training seizure detection models based on expert-labeled EEG samples.
• Benchmarking neural network performance against clinical expert accuracy for ICU EEG classification.
• Researching rhythmic and periodic continuum patterns in electroencephalography data from intensive care units.

Strengths

• 50,697 expert-labeled samples provide a substantial basis for model training.
• Annotations were performed by physician experts from 18 institutions, suggesting clinical validity.
• Used to train a model (SPaRCNet) reported to match clinical expert accuracy.

Limitations

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

Provenance

Source

Brain Data Science Platform

Collection Method

EEG samples collected from patients and annotated by physician experts.

Time Range

null

Freshness

Last updated: unknown

Geography

null