Even though they’re so small and vulnerable, premature babies generate more data than most current hospital IT infrastructures can decipher. That’s why a cloud-based, big data effort in Canada is being applied to help make sense out of streaming physiological data and give those babies a better chance to survive the numerous health threats they face so early in life.
The effort, called Project Artemis (named after the Greek goddess of childbearing), is led by Carolyn McGregor, Ph.D., the Canada Research Chair in Health Informatics at the University of Ontario Institute of Technology (UOIT). Artemis initially was deployed in 2013 as a research partnership between UOIT, IBM Canada and The Hospital for Sick Children in Toronto. It has since expanded across the globe and is supported by the Canada Foundation for Innovation and the Canadian Institutes for Health Research, among other organizations. The group behind the effort recently announced a $3 million project to develop a commercial version of the system that can be used as a software-as-a-service application.
Project Artemis sifts through the streams of real-time data being generated by monitoring devices to analyze heart rate, respiration, blood oxygen levels, blood pressure and other information to determine when premature babies are entering danger zones and are susceptible to infection and other health complications. According to research from the World Health Organization and other health agencies, premature birth is the leading cause of infant death worldwide. About 25 percent of premature babies end up with infections, and about 10 percent end up dying from health complications.
The effort currently monitors more than 1,000 premature babies and has expanded beyond the initial deployment at The Hospital for Sick Children in Toronto. The Artemis cloud computing environment, built on a platform from IBM, is linked to Women & Infants Hospital in Providence, R.I., and The Children’s Hospital of Fudan University in Shanghai, China. Additionally, McGregor is working with space agencies as well as SWAT teams in Toronto to apply physiological data monitoring systems to their efforts.
For hospitals treating premature babies on a fine line between life and death, the knowledge generated by Project Artemis is critical. The Artemis project has demonstrated through medical research that it:
- Can combine information from heart and respiration rates with other physiological data to reduce false positives in sepsis detection when compared with solely using heart rate information.
- Can automatically identify what type of neonatal “spell” or apnea a baby is having with greater than 98 percent accuracy.
- Can automatically classify sleep and wake states in neonates to help doctors assess the way a baby’s brain is maturing.
- Can provide information on oxygen levels that can be used to help ensure babies don’t receive too much oxygen, which can lead to permanent eye damage.
In neonatal wards, the stakes couldn’t be higher, and the chasm between data and actionable information couldn’t be wider, McGregor says.
There’s no shortage of data being generated on premature babies, she notes: babies’ hearts beat up to 8,000 times per hour, and they take about 2,000 breaths in the same time span. But when McGregor was first asked by a visiting neonatologist from the The Hospital for Sick Children to try to apply UOIT’s computer science expertise to examine the issue, she realized there was a fundamental piece missing from the architecture.
“All that second-by-second data being generated was just numbers on a monitoring screen—when people look at critical care they think of it as highly advanced, but what’s being provided to caregivers is an enormous amount of data that’s not in a form they can easily use at the bedside,” McGregor says. “There was no piece of the architecture that could ingest all that data and actually make use of it—all that second-by-second information was being used to generate one aggregate number an hour that was supposed to indicate a patient’s condition, which wasted so much information, because so many changes can occur in an hour.”
A physician at the hospital told her they often relied on the instincts of well-trained nurses, instead of the monitoring data, to identify babies who were heading toward a serious health problem. “When you have real-time streaming data on hand but the most reliable indication of a problem is a nurse saying, ‘that’s not the same baby I treated yesterday, something’s wrong,’ then as a computer scientist you know there’s a technological issue that needs to be addressed.”
McGregor set about bridging that chasm by designing a big data system to analyze massive amounts of high-velocity data to pinpoint when changes in the health data are correlated to certain conditions, and alert clinicians to impending health threats so they can take quick action before a baby is harmed. For example, Project Artemis has detected real-time patterns for conditions that infants in the neonatal intensive care unit can develop, such as:
• Apnoea of prematurity (pauses in breathing due to prematurity)
• Retinopathy of prematurity (eye damage from premature birth)
• Anemia of prematurity (deficiency of red blood cells often as a result of drawing blood for medical tests)
The project utilizes mining techniques, patented by McGregor, that are designed for the extraction of implicit, non-trivial and potentially useful abstract information from large collections of temporal data, in this case the numerical data being generated by monitoring devices.
The analytics system creates abstractions from the streams of data and looks for patterns, and then checks if patients who have had health events—such as infections, respiratory distress syndrome or forms of sleep apnea—have had those same types of patterns occur in their physiological data. Once those patterns are determined to be medically significant and have relationships with certain conditions, the analytics system can look for those types of patterns across the patient population and alert caregivers when the data indicates a baby is in danger of having a health problem.
Big Data Foundation
Project Artemis uses three off-the-shelf medical connectivity systems—from Capsule Tech, ExcelMedical and True Process—to feed data in real-time to a cloud-based database and analytics environment built on the InfoSphere analytics platform and the DB2 relational data management and warehouse system, both from IBM. There are two copies of the data, one which resides in the hospital and the other sent via real-time connections to a cloud environment for processing. The information in the data warehouse is used for retrospective analysis by applying temporal data mining techniques to come up with hypotheses that can be evaluated by medical researchers. Once those hypotheses are proved to be medically sound, they can be applied as rules in the hospitals’ own monitoring systems.
McGregor says Project Artemis is designed to be a “white box” effort that keeps clinicians informed about its discoveries and provide details about the patterns being unearthed via the analytics effort. “There are a lot of problems created in the medical environment when technology is a black box and the people caring for the patients don’t know how or why alerts are being generated. We are following an approach of traditional medical research that is transparent about how a certain pattern has been associated with a certain condition.”
Beyond the Hospital
The ability of Project Artemis to break down second-by-second physiological data into actionable health information is already expanding beyond the hospital. McGregor is working with space agencies to apply her techniques to monitoring astronauts, and also working with tactical law enforcement groups in Canada to better understand how high-stress situations affect performance.
For the space programs, astronauts on the International Space Station have weekly check-ups with Earth-based physicians. McGregor is now working with space agencies, and the Russian space agency in particular, to combine her monitoring techniques with a wellness algorithm developed by the Russians to provide a real-time picture of the overall health of astronauts/cosmonauts and to better identify the effects of body fluid load changes, bone degradation and other health issues associated with a weightless environment. The effort is designed to demonstrate how Artemis could be applied to better monitor the health of people on the International Space Station and as well as NASA’s proposed 2030 mission to Mars and other exploration efforts where space travelers would be out of contact with Earth-based physicians for extended periods.
In addition, McGregor has teamed with tactical operators in SWAT teams to analyze their physiological data while they train on virtual reality games. The project seeks to understand what kinds of changes in heart rate and other data can lead to problems such as fainting during operations, as well as what health data patterns are associated with post-traumatic stress syndrome and other long-term issues.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access