Motion-sensing tools deployed in war on pathogens
By Tony Craine
With fists at chin level, a young video game player focuses on the television in front of him. He begins shadow boxing, throwing a series of punches and kicks. Meanwhile, on the screen, an animated character engaged in hand-to-hand combat faithfully reproduces all of the player’s punches and kicks in real time.
The player is not holding a controller. Instead, his movements are captured by Microsoft Kinect, a motion-detecting component of the Xbox gaming system. Kinect’s built-in camera and depth sensor use “computer vision” that allows the player to use his entire body as a controller. For gamers, Kinect turns video game play into a full-body experience, with all of the excitement and none of the danger.
But for a group of University of Iowa researchers, Kinect is not a game—it’s a tool to help understand how infections spread in health care settings. By employing Kinect’s computer vision in a hospital room, the research team is pioneering an automated approach to track interactions between health care workers and patients, capturing previously elusive data to support the work of hospital epidemiologists.
Since the 1990s, the health care community has become increasingly vigilant in efforts to reduce health care-acquired infections. Driving much of this activity is a simple fact: Many health care-acquired infections are highly preventable. In fact, according to the World Health Organization, proper hand hygiene alone can prevent most health care-acquired infections. Another key safety measure is the use of proper procedures when putting on or taking off personal protective equipment, such as gloves, gowns, and masks.
Awareness campaigns around these issues seem to be working. In March, the U.S. Centers for Disease Control and Prevention (CDC) reported a 50 percent drop in central line-associated bloodstream infections between 2008 and 2014; a 17 percent decrease in surgical-site infections among 10 select procedures during the same period; and recent declines in almost every category tracked, including hospital-onset MRSA bacteremia and Clostridium difficile infections. The Agency for Healthcare Research and Quality estimates that in 2013 about 35,000 lives were saved by improving safety measures in health care settings.
Yet plenty of other emerging threats, such as the 2014 Ebola outbreak in West Africa, jeopardize the safety of patients and health care workers in different ways and require redoubled efforts at infection control.
For these reasons, the CDC in 2015 expanded its Prevention Epicenters program, an initiative to spur new research at a handful of academic medical centers that specialize in health care epidemiology. The UI Carver College of Medicine joined five other centers added to the roster of Prevention Epicenters, bringing the total to 11. As a Prevention Epicenter, the UI will receive $2.2 million over three years to investigate new ways of preventing the spread of pathogens in health care environments.
“This funding extends the work we’re already doing,” says Eli Perencevich, MD, UI professor of internal medicine and an infectious disease epidemiologist with UI Hospitals and Clinics and the Iowa City Veterans Affairs Health Care System. “Iowa has been at the center of infection-prevention research for several decades, focusing on making the hospital as safe as possible from health care-associated infections.”
Tracking provider activity
Perencevich leads the UI Prevention Epicenter team, which is using the CDC funding to support four projects that demonstrate in various ways the university’s leadership in the movement to safeguard health care environments.
“We have a broad range of multidisciplinarily trained faculty members, all focusing in this one area,” Perencevich says. “And we’re incredibly collaborative as a group.”
Emblematic of that collaborative ethos is the Kinect project. That work is a product of the UI Computational Epidemiology Group (CompEpi). Begun by Philip Polgreen, MD (’00 R, ’02 F, ’04 F), associate professor of internal medicine, and Alberto Maria Segre, PhD, chair and departmental executive officer of computer science, CompEpi unites computer scientists and medical researchers in an effort to develop and refine informatics that measure health behaviors and health outcomes.
“Several of our projects demonstrate how infections can spread through contact networks,” Polgreen says.
Those studies tracked the movement patterns of health care workers on the job, mapping their social networks by using data collected from human observation, from electronic medical records, and from readings provided by small sensors worn on health care workers’ uniforms.
The next step is to take a closer look at what happens in the hospital room, which is where the most contact between health care workers and patients takes place. CDC Prevention Epicenter funding will help Polgreen and his collaborators do that, using Kinect as an innovative automated observer.
Currently, the only effective means of monitoring those interactions inside patient rooms is through human observation, either in person or via video recording. But with computer vision and depth sensing, a room could be monitored automatically for hours or even days at a time, yielding much more detailed and precise data.
“Before, we could tell if someone entered a room,” Polgreen says. “Now, we’ll know not only if they were in the room, but where they were in the room and where their hands were. We can measure contacts at a granular level.”
The Kinect effect
When Kinect “sees” a person, it translates that image into an axial skeleton—a stick-figure abstract of the body, defined by 25 points that represent major joints. A depth sensor enables a 3-D rendering of that axial skeleton. Kinect can also detect the presence of colors in specific areas of the skeleton. This abstraction of the images guarantees anonymity, because the only data saved is in the form of Cartesian (x, y, and z) coordinates. No information that would personally identify patients or health care workers is collected.
The CompEpi team devised a system using three Kinect units to monitor activity in a room. Two Kinects are placed behind the head of the bed—one on the left side, one on the right—facing diagonally across the bed toward its foot. These two Kinect units follow the axial skeleton of a person moving around the bed. Any time the health care worker’s hands come within touching distance of the patient, that is recorded as a contact, and the location and duration of the contact are noted. The third Kinect, pointed toward the room’s doorway, uses color recognition to detect protective equipment on health care workers entering the room. Each Kinect is capable of tracking up to six skeletons at once.
The team used simulated hospital room settings and empty patient rooms to develop the system. For this project, the system ultimately will be deployed in an actual medical intensive care unit room for patients with influenza. The primary objectives are to determine whether the system can reliably detect personal protective equipment worn by health care workers and whether it can reliably identify moments of close or direct contact between workers and patients.
Applications of new sensing technologies have the potential to make hospitals safer for patients and providers. For example, an automated monitoring system could alert a health care worker who enters the room without proper protective equipment or detect missed opportunities for proper hand hygiene.
And the ongoing collection of in-room interaction data can have a lasting impact on future research.
“We can learn a lot about what’s going on just by watching this kind of anonymous data,” Polgreen says. “We can build much better models to understand how health care workers and patients interact, and we can use that information to understand how infections spread.”
Because this type of automated room monitoring is so new, the work is sure to raise questions that will drive innovation. How could such a system be taught to tell the difference between health care workers and visitors? What other types of contact or movements should be measured? How does room lighting affect accuracy? What new types of sensing equipment and software are available to make the system easier to deploy?
James Cremer, PhD, UI professor of computer science and a member of the CompEpi team, says one of the strengths that computer science faculty and students bring to this work is their knowledge of new technology and methods that the medical experts may not even be able to imagine.
“You can’t always ask people what they want until you show them the new tools you have and how they work,” Cremer says. “They’ll often say, ‘I don’t know what I want.’ But then you make a guess, put in a new sensor, and try a new way of gathering data. When they see it work, then they start to get their own ideas about how to use it.”
The CompEpi group applied similar techniques to analyze publicly available data, such as examining Twitter feeds to predict flu outbreaks or mining the letters-tothe- editor sections of peer-reviewed journals to spot early trends in adverse drug events.
“These long-standing interdisciplinary collaborations across different departments and colleges help bring new computational methods and approaches for solving challenging problems in the field of health,” Polgreen says.
Putting Iowa ‘on the map’
By Tony Craine
By designating the University of Iowa Carver College of Medicine as one of its 11 Prevention Epicenters for 2015-2018, the Centers for Disease Control and Prevention is asking the UI to assume a leadership position in the effort to ensure safer conditions for patients and the workers who provide their care. It’s a familiar role.
The UI participated in the first two rounds of the Prevention Epicenters program, in 1997-2000 and 2001-2005. Loreen Herwaldt, MD, professor of internal medicine and former hospital epidemiologist at UI Hospitals and Clinics, served as principal investigator for those grants. Herwaldt says Iowa’s legacy of excellence dates to the 1970s, when Walter J. Hierholzer, MD, started the program of hospital epidemiology and also inaugurated a statewide infection prevention education program that was one of the first in the nation.
“He’s one of the grandfathers of hospital epidemiology,” Herwaldt says. “He put Iowa on the map in terms of infection-prevention education.”
In the late 1980s, Iowa’s reputation continued to grow with an emphasis on research under the guidance of Richard Wenzel, MD, who served as director of the Division of General Medicine, Clinical Epidemiology, and Health Services Research.
“He was driven to use the data that we collected day to day to answer questions about infection prevention,” Herwaldt says. “We received some of the early federal funding in the field and, after he left, we continued the model he developed of service, education, and research—you could call it the three-legged stool of infection prevention.”
Iowa’s program of hospital epidemiology has trained some of the world’s infection prevention leaders, including Didier Pittet, MD (’92 F), who heads the World Health Organization’s hand-hygiene campaign.
“In general, you’ll hear people say, ‘Iowa? Where’s Iowa?’” Herwaldt says. “But it’s fun for me to go to a national or international meeting and say, ‘I’m from Iowa,’ because in our field, people know Iowa.”
UI’s CDC-funded projects cover it all
By Tony Craine
The four projects funded by the $2.2 million Centers for Disease Control and Prevention (CDC) Prevention Epicenters grant demonstrate the range of approaches University of Iowa researchers take to advance health care epidemiology. On one end of the spectrum is a high-tech video game component; on the other end, the lowly toilet seat.
“We’re always worried about aerosol-generating procedures in health care settings, and one of the worst is flushing the toilet,” says Eli Perencevich, MD, UI professor of internal medicine and an infectious disease epidemiologist at UI Hospitals and Clinics and the Iowa City Veterans Affairs Health Care System. “So we’re trying to show how flushing a toilet leads to increased spread of bacteria. It’s not as earth-shattering as cool Xbox sensors, but both are important.”
U.S. regulations prevent toilet lids in public facilities. Perencevich is leading a simple study in 13 hospitals to investigate the efficacy of a lid in reducing the dangers of the droplets and aerosolization caused by flushing.
To illustrate the current state of affairs, Perencevich notes that safetyconscious workers in a major U.S. Ebola center resorted to using cafeteria trays as makeshift toilet lids.
“If we don’t have the data to show the dangers, we’re never going to change the regulations,” he says. “We have to push back against this.”
Another project, led by Loreen Herwaldt, MD, professor of internal medicine and former hospital epidemiologist at UI Hospitals and Clinics, will investigate ways to improve procedures for putting on and removing the personal protective equipment used when caring for patients with Ebola or another dangerous pathogen. Improper procedures with this equipment can put patients and health care workers at risk.
The final project, led by Marin Schweizer, MD, PhD, assistant professor of internal medicine and epidemiology, will involve a meta-analysis of scientific literature addressing protective equipment, such as gowns and gloves, worn by health care workers. A leading expert in this type of systematic review, Schweizer will examine all published studies on the topic and generate a quantitative overview of the benefits and harms associated with the current equipment.
“This is the strength of the University of Iowa,” Perencevich says. “We have a highly collaborative and multidisciplinary team here. If you want to study infection prevention, come to Iowa.”