A data visualization of the Michael J. Fox Foundation's Parkinson's data from LIONSolver.
The Michael J. Fox Foundation (MJFF) has recently begun to explore how data sourced from mobile phones and analyzed with machine learning algorithms can help improve research into Parkinson's. The organization held the Parkinson's Data Challenge, where teams competed for a $10,000 prize with their approaches to analyzing a dataset collected from Parkinson's patients.
"Last year, we were approached by [entrepreneur] Daniel Vannoni. He had conducted this project with 16 individuals -- nine patients, seven control, and for 8 weeks, 4-5 hours a day, these individuals carried a smartphone. The smartphone had seven sensors that were collecting data for that period," said Laxmi Wordham, Chief Digital Officer of the MJFF. She said that Vannoni, a Cambridge-based entrepreneur who heads up Gecko Ventures, did the study in the hopes of building a commercial app for managing Parkinson's, but when the project fell through, he elected to donate the data to the Foundation, whom he had worked with previously as a fundraiser.
"We don't have a statistician on staff, and we thought there was something behind this data that we wanted to pursue," Wordham said. "So we thought about making this a challenge. One objective was to test the prize/challenge model. Objective two was, could we learn something from data collected through mobile technology? Was there a 'there' there that we could continue to pursue? And three, could we reach a statistical analysis base of people who weren't involved in Parkinson's?"
The data was collected in the background by the phones and included movement data from the built-in accelerometer, data about the user's tone of voice, how much the phone was turned on and used, data from the built-in compass and GPS, and an ambient light sensor. One of the earliest symptoms of Parkinson's is tremors, which could be detected either from voice or movement.
In fact, a number of mobile health innovators have been working to detect and analyze Parkinson's symptoms. UCLA engineers did a clinical trial that proved the effectiveness of the iPhone's built-in-accelerometer back in 2010. More recently, Max Little's team at the Parkinson's Voice Initiative claimed to be able to detect Parkinson's with 99 percent accuracy from voice analysis over the phone.
The teams in the MJFF competition were challenged to develop a model that could use the data to identify the Parkinson's patients from the control group and identify what stage of the disease users were in. But the challenge was also open-ended for teams to do more with the data.
Machine learning software company LIONSolver took home the prize. The company built an algorithm to analyze the data, working with a movement disorder specialist at Cedar-Sinai Hospital in Los Angeles, where the company is partially based (they also have an office in Northern Italy.)
"Now that we've done that, of course, with the support of MJFF and also Cedar's there's tremendous interest in seeing this go from experiment to something that would aid physicians," said Drake Pruitt, CEO, LIONSolver. "The dream is a physician would have a way of monitoring the daily life of patients. Parkinson's disrupts the daily life of people 24 hours a day, but a patient would only see their doctor once a month or once a quarter. It's a challenge to accurately recap what's happened -- if they're feeling great, it's been great and if they're feeling poorly, it's been horrible and the reality is somewhere in between. To enable a physician to see what's really going on is something the physicians are very excited about, and patietns as well. We've talked to physicians who are really excited by this possibility."
LIONSolver will be continuing to work with the Michael J. Fox Foundation, and the foundation is continuing to use crowdsourcing in other ways as well. Last year, it launched Fox Trial Finder, an online tool to help locate Parkinson's trials that has signed on over 16,000 volunteers for about 300 trials. Wordham said the organization is now working on what they're calling the Patient Centered Research Tool.
"We want to source more information from patients. As part of that we're looking at mobile applications to collect that data more readily," she said. "Patients are willing to give out information about themselves and researchers are craving that data. Surveys, mobile apps to collect information both in a passive sense and a more active sense: There's a lot of very creative things we believe we can do."
Crowdsourcing data to improve research is a popular idea lately. At TEDMED two weeks ago, Jessica Richman from uBiome talked about how the future of research is crowdsourced data. PatientsLikeMe recently opened up an open-participation research platform to share some of the data it's collected from patient users. High-profile disease-specific groups like MJFF are well-positioned to become leaders in the crowdsourced research revolution.
"I think we are in a unique position," said Wordham. "We have a brand name, we have a large following, an active, engaged community that are willing to get involved when we ask them to. I think in our position there's a lot more that we can do with building the right tools, to both have patients get engaged and then have a research community that can utilize that to better understand the disease."
For real-time updates from MobiHealthNews, Follow us on Twitter.