Sudden Unexpected Death in Epilepsy (SUDEP) is the leading mode of epilepsy related death. Recent estimates indicate that SUDEP is responsible for approximately 7,000 deaths each year in the United States and Europe, and is the second most common cause of the number of adult life years lost after stroke. To accelerate SUDEP research, the National Institute of Neurological Disorders and Stroke (NINDS) at the NIH- funded Center for SUDEP Research (CSR), a network of 14 institutions collaborating in a broad spectrum of basic science and clinical approaches to study possible biological mechanisms underlying this potentially preventable mortality and develop predictive biomarkers for interventions. Identification and communication of alterable SUDEP risk factors to affected patients is an important strategy to lower SUDEP incidence. However, systematic individualized assessment of SUDEP risk is currently unavailable due to a number of challenges. Often the required information is embedded in data residing in disparate, unlinked datasets and systems; there is a lack of a specific controlled vocabulary for precise extraction of SUDEP risk factor information with semantic uniformity; and the corresponding computational algorithms and tools needed for important risk marker extraction from clinical text and electrophysiological signals are yet to be fully developed. We propose to overcome these challenges by developing SURME, a SUDEP Risk Marker Extraction system for automated extraction of known and putative SUDEP risk markers from the multimodal CSR data repository (called MEDCIS) which contains over 1,600 patients enrolled from Epilepsy Monitoring Units in 7 medical centers.
In Aim 1 we will develop a dedicated controlled vocabulary building on our own Epilepsy and Seizure Ontology and existing SUDEP risk guidelines and reported risk factors. We will develop an extraction pipeline, leveraging our earlier epilepsy phenotype extraction tools, for detecting risk markers from clinical text.
In Aim 2 we will develop a scalable approach for detecting two significant putative physiological biomarkers from electrophysiological signals: postictal generalized EEG suppression; and root mean square differences of successive R-R intervals.
In Aim 3 we will perform pilot implementation of SURME on MEDCIS for automated risk assessment using ?SUDEP-7 Inventory? and ?SUDEP and Seizure Safety Checklist?, as well as assessment of putative SUDEP risk factors using CSR cohort. We expect SURME and its future versions to become an invaluable SUDEP risk assessment tool as a part of standard epilepsy care. The long-term goal of this study is to create evidence-based SUDEP risk assessment tools to improve epilepsy care, with individualized risk scores and recommendations for managing modifiable risks, ultimately leading to reduced SUDEP mortality and improved epilepsy patient care.
The main goal of this project is to develop an informatics approach for automated extraction of SUDEP risk markers from multimodal clinical data to enable individualized SUDEP risk assessment. Success of this study will enable systematic SUDEP risk assessment based on known and putative factors and communication of such risk factors to patients with epilepsy. Ultimately, this study can lead to evidence-based SUDEP risk assessment tools that help clinicians and patients manage potentially modifiable risks, leading to overall reduced SUDEP mortality and improved epilepsy patient care.
