Movement analysis app detects level of lung disease

Nearly 16 million Americans are afflicted with chronic obstructive pulmonary disease (COPD), a progressive disease where the lung gradually loses its ability to pump enough oxygen to the rest of the body. According to statistics published by the Centers for Disease Control and Prevention, COPD is the third-leading cause of death in the United States and only half of the actual cases are diagnosed.

A team of University of Illinois researchers would like to help these millions of COPD patients better manage their health. Under the direction of CS and Medical Information Science Professor Bruce Schatz, the team is developing mobile technology that can accurately monitor COPD patients’ symptoms through a smartphone that they carry in their pocket.

Specifically, team member Qian Cheng, a CS graduate student, developed the analysis technology to use only the phone sensors to predict results of the six-minute walk test, a simple, yet reliable, tool that doctors use in the clinic to help gauge the condition of COPD patients.  In addition, he developed a predictive model to accurately compute the pulmonary function, the health status of these lung patients as measured by how well they can currently breathe.

“A doctor can tell from how a patient moves and walks during this test just how serious his/her cardiopulmonary disease is,” explained Cheng. “Our idea is to develop the app so it could evaluate [patients’] risk without them having to come into a doctor’s office to do the test.”

The team’s MoveSense app collected movement data from 24 older COPD patients who were doing the conventional six-minute walk test under medical supervision at a pulmonary rehab facility in Evanston, IL.  Cheng then used data mining and machine learning algorithms to categorize the patients into groups based on the severity of their condition. This training model could then predict with perfect accuracy which of the four standard severity levels—mild, moderate, severe, more severe—other patients would be in when they are tested with the app.   

“Our idea is to spread this app to millions of people who wouldn’t even have to go to a doctor’s office to do the test,” said Cheng, who is presenting the results of this work at the American Medical Information Association annual symposium in November in Chicago. “Our app would evaluate their risk and if it’s high, then the patient would be told to go to the hospital where a doctor could do a more thorough exam to see if they’re in danger or not.”

Recently, Cheng received a $10,000 award as a top-10 finalist in the annual Student Technology Prize for Primary Healthcare administered by Massachusetts General Hospital associated with Harvard University (formerly known as the CIMIT Prize). 

“It was very unusual to have a finalist who did predictive modeling using machine learning, rather than building a medical device,” said Schatz, noting that the other nine finalists were bioengineering students.  Schatz’s students have been finalists in four of the eight years that the national competition has been held.

Cheng used the prize money to further develop and test his app with the same patients, but this time they carried the smartphone in their pocket so they could be monitored in their everyday activity at home. This passive monitor would require nothing special from the patient, yet the model would predict their status with clinical accuracy.

“We want to capture from the patients’ daily activities a qualified walking period, transfer that data to our server, and automatically do an analysis,” he said, noting that data collection over multiple days yields important information. “A chronic disease will not usually change overnight, but our method will detect a change for the worse as soon as possible so the patient can go see their doctor for updated treatment.”

Although he is set to graduate with his doctorate this semester with a dissertation on predictive models in healthcare applications, Cheng is working now on extracting the qualified walking period from an entire day’s activity collected by the phone that patients carry in their pocket. “We are using a sampling strategy to monitor if the phone moves or not,” he said. “Our system will be able to automatically turn the recording on when it detects enough movement since these older patients don’t move much during the course of a day.”  

According to Cheng, a long-term goal is to provide patients using his technology with a daily report summarizing their condition based on qualified data collected by the smartphone in their pocket, to enable them to better manage their chronic disease.