Project Artemis, a Big Data analytics platform developed jointly by IBM and the University of Ontario’s Institute of Technology, can detect nosocomial infections in premature babies 24 hours before the symptoms appear.
Big data is getting bigger in healthcare. It promises to herald a new age of data-driven care wherein healthcare professionals will be able to improve the effectiveness of care by analyzing huge amounts of data derived from millions of cases and use it to define the needs of subpopulations. This will enable more personalized decisions tailored to individual needs, and help to identify and intervene for population groups at risk.
A prime example of such data-driven healthcare is Project Artemis. Named after the Greek goddess associated with protecting child-bearing women and infants, Project Artemis is a collaboration between the University of Ontario’s Institute of Technology and the technology giant IBM.
What is Artemis?
Artemis is a real-time data gathering and analytics platform. It acquires and stores patients’ physiological data streams along with clinical information system data to perform online real-time analytics, retrospective analysis and data mining to detect clinically significant conditions and their onset behaviors.
Artemis and Nosocomial Infections
Premature babies born before the normal gestational period of 37 weeks can face many challenges to survival—such as underdeveloped immune systems and immature lung development—putting them at extreme risk.
The very hospital environments in which babies are cared for can harbor viral, bacterial or fungal infections, which can be deadly due to weak immune systems. These hospital-acquired infections, also called nosocomial infections, cause nearly 100,000 deaths each year in the US alone.
One of the worst of this class of infections is Late Onset Neonatal Sepsis (LONS), which can be fatal for premature babies. But the real problem with LONS is that it’s difficult to diagnose clinically. By the time a premature baby exhibits the symptoms, the infant’s skin, respiratory tract, conjunctivae, gastrointestinal tract, and umbilicus may all be colonized with bacteria.
Blood tests typically don’t reveal much and false negative results are a problem. Pulse rates are usually within range and other vital signs vary within acceptable limits—even as infection continues to build. Even skilled nurses and experienced physicians have difficulty identifying the symptoms early enough to intervene and avoid severe complications.
A retrospective analysis of data by physicians at the University of Virginia surprisingly revealed that premature babies with more stable heartbeats were actually more susceptible to nosocomial infections. Starting 12 to 24 hours before the onset of noticeable symptoms, the heart rates of infected babies became too regular and stopped varying, as they should in a healthy state.
However, such indications are too subtle to be identified by healthcare workers in a neonatal intensive care unit. While there is a trend to these subtle changes and symptoms, it can only be identified in retrospect after a thorough analysis of medical data.
"We looked at the heart rate for trends; when babies become infected, the baseline heart rate tends to increase," says Dr. Andrew James, the Associate Clinical Director at The Hospital for Sick Children in Toronto where the Artemis was first tested. "What I've learned is that when you look at more data and you look at it in a more granular sense, not only is there more to be seen, but you actually see more. We're beginning to see abrupt changes in heart rate variability and that really makes us think of infection," he notes.
Thanks to Artemis’ stream analytics software platform that can continually analyze the streaming data from multiple monitors using advanced, targeted algorithms, physicians now have near real-time decision support in neonatal intensive care units.
A Small Baby is a Big Data Problem
Medical devices used in neonatal units—such as monitoring systems, ventilation support, and smart infusion pumps—generate a lot of data.
The pulse rate, respiration rate and blood oxygen levels that are displayed each second generate 86,400 readings each per day. Electrocardiogram graphs are generated based on 1000 readings every second. A premature infant’s heart beats more than 7000 times an hour or 170,000 times a day.
With a baby typically having more than 10 infusions at the same time, drug and nutrition infusion data from pumps alone can cross 1GB per patient per day. Such a large data set can only be analyzed using advanced stream processing applications such as Artemis’ SPADE.
Using a temporal data mining approach, Artemis processes 1256 readings every second and identifies the condition onset behaviors. It gives early warning signs of infection 24 hours before they can be manually identified, providing the opportunity for timely interventions that could save precious lives.
Increasing Popularity of Artemis
Since its first introduction at The Hospital for Sick Children in Toronto in August 2009, the algorithms used to detect subtle changes in the underlying data streams have been consistently improved. A cloud computing version of Artemis was also launched at the Women and Infants Hospital in Providence, Rhode Island.
Recently, Artemis has been deployed in a 200-bed neonatal intensive care unit at a hospital in Shanghai, China and a 100-bed unit in Shenzhen, China. Two more installations are also being planned, one in Canada and another in Australia.
“The interest in Artemis is amazing. People are coming to us saying that they want to be involved. They have new ideas for research projects. It’s growing: more hospitals, more conditions and more medical devices. It’s a multidimensional environment and we’re expanding on every dimension,” says Dr. Carolyn McGregor, the research chair of health informatics at the University of Ontario's Institute of Technology.
Future Applications for Artemis
The flexible platform of Artemis means that it can be used to detect early warning signs of many conditions, as long as data streams are fed into it. Currently, the additional possibilities for using Artemis are to detect low oxygen levels, slow breathing, retinopathy of prematurity and late-stage, at-risk pregnancies.
The cloud computing nature of Artemis supports use both inside and outside of the ICU and even outside the hospital. Using remote sensors and wireless transmission of data streams, it is possible to monitor patients in remote locations and provide medical alerts in near-real time.
“We have created an integrated solution tailored to neonatal and potentially all critical care,” says Dr. McGregor. “Artemis gives us an infrastructure where the physiological data can be transmitted and we can analyze it remotely, which also has huge implications for rural areas.”
“I think the framework would also be applicable for any person who requires close monitoring—children with leukemia, for example,” says Dr. James. “These kids are at home, going to school, participating in sports—they’re mobile. It leads into the whole idea of sensors attached to or even implanted in the body and having wireless connectivity. Theoretically, we could ultimately monitor these conditions from anywhere on the planet.”
As big data continues to get bigger in healthcare, the new age of data-driven care through technologies such as Artemis will support personalized interventions based on earlier identification of warning signs for populations at risk.
Shiva Gopal Reddy has a Bachelor's degree in Physics and a Master's in Applied Psychology and writes frequently on the latest research, impact, happenings and trends in digital health technology.
