Grant 211713 EPILEPSIAE 1
• Aims
– Implementation of large, high-quality polymodal data from patients with epilepsy for the purpose of advanced analyses leading to an improved prediction of epileptic seizures
• Planned data content
• 250 surface, 50 intracranial continuous long-term EEG data sets with annotations
• Imaging data (3D MR data sets)
• Structured metadata (clinical)
• Derived features based on EEG analysis
EUROPEAN EPILEPSY DATABASE
Grant 211713 EPILEPSIAE 2
Database: preparatory steps• Specific issues:
– Safety of personal data• Pseudonymization of patient data• Deletion of data contents with
identifying character
– Ethical issues• Individual informed patient consent• Restrictions of access and use
– Management of multimodal data• raw data (e.g. EEG/MRI data)• meta data (e.g. patient history,
seizure counts, semiological characteristics, EEG annotations, electrode positions)
– Multisite access to local databases of consortium members (data warehouse approach)
• Multiple local databases• Procedures do assure identical data
content
Coimbra Freiburg
Paris
Replicated European database
Client interfaces
SQL
Grant 211713 EPILEPSIAE 3
Clinical procedures
• Patient selection• Criteria for data quality
– Standardized annotations(types and positioning of seizure-related markers:
• Clinical seizure onset / first behavioural alteration
• EEG seizure onset / first EEG change
• Spikes• Subclinical events
– Uniform nomenclature of EEG channels
– EEG review• Metadata (types and inclusion
form)
Consortium consent on Inclusion criteria (Paris, 2008):
• Patients with focal and multifocal epilepsy• Minimum duration of continuous EEG
recordings: 96h (4 days)• Minimal number of seizures: 5
with an interseizure interval of ≥ 3.5h• Information on subclinical EEG events and
sleep stages in the preictal period• Appropriate electrode implantation for
focus identification
Grant 211713 EPILEPSIAE 4
T1.1: clinical procedures
Sleep staging
EEG onsetEarly propagation
EEG pattern morphology
Grant 211713 EPILEPSIAE 6
Database development
Storage of sample values:• External binary files: space
efficient & the way people are used to work
• Inside database tables: flexible for querying
Database system:• Open source (PostgreSQL) vs.
Oracle (commercial)• Decision for Oracle to allow the
storage of the samples inside the database
Replicated vs. distributed database• Local datatbase at every site,
replicated content• Not a single distributed database
because of immense data traffic• Estimation of data volume: > 50Tb