The Michael J. Fox Foundation (MJFF) for Parkinson's Research has been seeking insights on how mobile data can aid Parkinson's patients. As part of this effort, the foundation launched a Parkinson's Challenge in February with a $10,000 prize to encourage use of patient data from smartphones for monitoring and treatment of Parkinson's. Kaggle.com a crowd-sourcing data analytics service, hosted the challenge.
MJFF was interested in exploring a crowd-sourcing approach on Kaggle after the success of crowd-sourcing on its Fox Trial Finder site, launched in 2012 to allow people living with Parkinson's disease as well as control subjects to volunteer for trials, Laxmi Wordham, chief digital officer at MJFF, told eWEEK.
The contest was modeled after the innovation challenges run by the nonprofit X Prize Foundation, according to Wordham.
"We were blown away, not just by the interest in downloading the data and submitting proposals, but also the diversity of the people that participated in this competition," Wordham said.
The winner of the challenge, LIONsolver, which stands for Learning and Intelligent Optimization Solver, involved teams in 21 countries that downloaded 630 sets of the data and used a machine-learning approach to explore clues to how the onset of Parkinson's occurs and how it progresses.
"The process was fully automated using our LIONsolver software," Drake Pruitt, CEO of LIONsolver, told eWEEK in an email. "We applied machine-learning algorithms to the data to automatically cluster subjects based on the GPS and accelerometer data so we could classify individuals by how their phones moved," Pruitt said. "Next, the software applied a 'support-vector machines' approach to identify Parkinson's patients who showed combinations of tremor and movements that were distinctive from all other subjects in the data set," Pruitt said.
The challenge proved how passive mobile smartphone data could aid patients and their physicians, according to Pruitt. These machine-learning techniques could help with additional movement disorders besides Parkinson's, he added.
LIONsolver consulted a movement specialist to carry out its research on the data, Wordham said.
Wordham compared LIONsolver's machine-learning approach to that of IBM's Watson technology, which is being used by Memorial Sloan-Kettering Cancer Center in New York to develop a decision-support application for cancer treatment.
Sensors on a smartphone can provide data on how Parkinson's is affecting a patient. Experts at Gecko Ventures and Massachusetts Institute of Technology designed a mobile application that allows Parkinson's patients and control subjects to provide the data. The smartphone included audio, accelerometry, compass, ambient light, proximity and battery level and GPS sensors.
"Our hope is by taking a cross-disciplinary approach, and collecting and analyzing multiple streams of data in a way that is easy and simple to collect from PD [Parkinson's disease] patients, we will discover greater insights into the condition and the effectiveness of treatments, and inspire new approaches to finding treatments and a cure for Parkinson's disease," Daniel Vannoni, a Gecko Ventures entrepreneur and managing director who collected the data, said in a statement.
Vannoni took data from nine Parkinson's patients and seven controls over eight weeks, Wordham said.
"We looked at the data and thought this was an interesting use of smartphones to collect data from the community," Wordham said.
Later this year, MJFF plans to launch a research tool for Android and iOS that would allow patients to input data about their medication and medical history for themselves and their families.
"As part of release one, we're looking at a mobile device that would track on/off periods for Parkinson's patients," Wordham said. "We absolutely see an opportunity to use smartphones in the collection of more data and then really just being able to give patients and their caregivers tools to give more information on the disease progression that will inevitably help them manage their disease."