The Smart Neuro ICU

Managing brain injury in a more individualistic manner will require data from diverse sources (physiology, imaging, biomarkers, genomics, etc.). The data will need to be integrated, annotated, analyzed, and visualized. It will need to be archived in a manner that can be shared for retrospective analysis, research, and education. A seamless architecture for neurocritical care data is highly desirable but has been very elusive with many barriers slowing progress. Our project will address shortcomings in three areas of technology:

Data Aggregation

The connected Neuro ICU must gather data from multiple sources to provide a comprehensive view of the patient as well as input to analytics and visualization.  These sources include medical devices, lab data, imaging, therapeutic measures (nursing action, drugs, etc.), and the medical record.

One of these areas, medical device connectivity, has been particularly problematic.  There is no widely adopted standard for medical devices to output their data. Additionally, in medical device communication protocols there is a wide range of quality (high quality to error-prone) and capabilities (robust communications to none at all). Today’s solutions for recording medical device data are generally focused on populating the electronic medical record and not on obtaining data at the resolution required for advanced analytics.

Some of the data we want to collect (labs, images, medical record data) reside in separate repositories and access to it ranges from easy (images using the DICOM standard) to difficult (medical record data).

This focus area will attempt to identify, facilitate, and/or create methods for the collection of comprehensive data required for managing the neurocritical care patient.

Data Management, Standards, and Sharing

There is an acute need to develop standards for identifying data and metadata both physiologic and phenotypic.  The same holds with methods and techniques for collecting, annotating, and archiving data. Multi-center trials generally spend an inordinate amount of time and money trying to accomplish this…many of them re-inventing the wheel. Solving this problem will provide a cost benefit to funding agencies and will accelerate advances in the quality of care due to the information that can be mined from a unified database.

In this topic we will work with data repositories as well as helping to expand the NINDS Common Data Elements.  We will review and work with various standards and data organizations to provide recommendations. We will also work with data repositories such as FITBIR.

Analytics and Visualization

Once the data is collected, there is a need to extract value.  The value may differ depending on the consumer of the data and thus the analytics will likely span a range of techniques from simple metrics to machine learning.  This topic will cover these techniques as well as associated regulatory processes.