|
BACKGROUND
Waveform data from physiologic monitoring devices is the most informative data from a patient. This includes waveforms produced by monitoring mechanical pressures in arteries (arterial blood pressure), electrocardiograms, electroencephalograms and oxygen saturation (Fig 1).
Real time storage and access of that data for secondary uses (i.e. retrospective studies and Machine Learning), is cost-prohibitive. The lack of access to waveform data sources is a major barrier to clinical decision support, quality improvement and actionable information for translational research. Digital signatures deciphered from interrogation of waveform data can provide additional diagnostic and prognostic capabilities impacting patient care. Industries outside of healthcare, such as mining and LiDAR struggle with similar issues regarding their timeseries data.
DESCRIPTION OF THE INVENTION
AtriumDB is an innovative solution that enables secondary uses of high frequency, medical, waveform data (Fig 2). This is accomplished through simultaneous aggregation of complete waveform data from multiple devices, storing it in a cost-effective way and maintaining the data in a form that is easily retrievable and actionable.
The team has designed a custom file structure known as a time series compression (TSC) file that allows multiple compressed segments of signal to be stored in a single file, without data loss.
COMMERCIAL APPLICATIONS & ADVANTAGES
Applications: The secondary uses of AtriumDB are vast. AtriumDB can enable:
• building visualization formats
• developing shared data registries
• identification of computational biomarkers
• developing and deploying machine learning models.
This core software is well positioned to have an impact via a SaaS model to hospitals, a direct license to companies to manage their data and an app store model for community-developed apps using waveform data. This core software would be valuable for industries where massive timeseries data is collected and relatively predictable.
Advantages:
• Cost-efficient rapid data storage and retrieval
• Dataset exploration to form hypotheses
• Combining measurable biomarkers into a live feed of a computational biomarker
• Improved machine learning model creation with real time deployment
• Ability to “replay the past” to analyze various biomarkers during historical clinical events
Without AtriumDB, accomplishing these applications using currently available cloud solutions would be at least ten times more expensive.
DEVELOPMENT STAGE
• Deployed at SickKids since January 2016 and in continuous operation since.
• Largest clinical database at SickKids – beating heart is heavily optimized collection, indexing and storage - > 3.5 trillion datapoints stored in 1.1 TB able to be extracted at a billion datapoints/second.
• In the process of being deployed at four other major hospital as pilot sites.
PATENT STATUS
A US Provisional patent has been filed for AtriumDB, under 524-004USPR.
|
