Harnessing AI for Parkinson’s Research

Standard

Harnessing AI for Parkinson’s Research

by Krishna Knabe,
January 14, 2019

Last week at the annual Consumer Electronics Show (CES), one of the big stories was health technologies. During CES, The Michael J. Fox Foundation (MJFF) and IBM announced a partnership using artificial intelligence (AI) to develop a model for Parkinson’s disease.

Though they are now gaining traction, artificial intelligence strategies, such as machine learning, have been a priority for MJFF for many years. In 2013, we launched a data challenge to spur ideas for Parkinson’s disease monitoring and treatment. The winners used machine learning and data collected from smartphones to improve communication between doctors and patients about symptoms. Since then, we have supported more research using artificial intelligence to analyze Parkinson’s data, including another data challenge in 2016 with partners at GE Healthcare.
Data is the key to this endeavor. You have to have a lot of comprehensive data from patients, collected using strict protocols, before you can apply machine learning or other types of artificial intelligence to it. MJFF has made a herculean effort over the last decade to collect this data in Parkinson’s, through the Parkinson’s Progression Markers Initiative, BioFIND, the LRRK2 Cohort Consortium and other MJFF-driven studies.

In an article in HealthLeaders, Mark Frasier, PhD, our senior vice president of research programs explains: “Today our challenge is to analyze this treasure-trove of complex imaging, clinical, and molecular data… AI and machine learning technologies can help scientists identify trends and models across Parkinson’s data. These frameworks can help researchers design more efficient and accurate clinical studies and drug trials, speeding discovery and bringing new personalized treatments to patient hands faster.”

The promise of artificial intelligence is that it can uncover fresh insights and trends in data that researchers might never find because they would not know to look for it or because the data analysis required is very complex. In our project with IBM, researchers will use our PPMI data to develop a model of the distinct stages of Parkinson’s. The next step will be to use that to develop a predictive model of how the disease might continue to progress. This understanding could improve patient care, but it will also help researchers understand what to expect during a clinical trial. When designing a trial, researchers want to make sure that patients are in the trial long enough to show a benefit. If they understand stages of progression, they can understand what is likely to happen over that span of time to patients on a placebo and, conversely, the benefit (in slowing or stopping progression) in patients taking a drug.

Our partnership with IBM Research Healthcare and Life Sciences is one of several designed to foster data analysis in Parkinson’s disease. Last year, we announced a partnership with Blackfynn to analyze data in PPMI using its data integration and analysis platform. And we are making our data sets available through our collaborations with AMP PD and BRAIN Commons. Important insights into Parkinson’s disease and how to treat it lie in the data we have collected, and we are doing everything possible to make sure they get found.

TAGS: Parkinson’s Data Challenge, Research News, Industry & Nonprofit Collaboration