A view from the Center

Deloitte's Life Sciences & Health Care Blog

Crowdsourcing advancements in epilepsy treatment

The Mayo Clinic hosted an online competition where more than 500 teams of data scientists from all over the world analyzed hundreds of hours of brain activity in two people and five dogs before and during epileptic seizures. The results of the 2014 competition, published recently in Brain, showed that the crowd created a reliable prediction of epileptic seizures.

More than 50 million patients worldwide have epilepsy, a neurological disorder that causes unpredictable seizures. Because it is difficult to predict when a seizure may occur and they may lose consciousness suddenly, people with epilepsy may have severe limitations on driving, swimming, holding a child, and other common activities. Predicting and even preventing seizures could greatly enhance the quality of life for people with epilepsy; they could take medications only when needed and resume activities they may otherwise avoid.

Until now, researchers have not had access to large amounts of data necessary to create a predictive algorithm. Brain activity is typically measured through implanted electroencephalography (EEG). EEG recordings lasting seven days or fewer are taken before surgery when a patient’s medications are reduce to prompt seizures. However, this data has limited information about brain function under changing conditions. Researchers working separately cannot easily share the data due to cost, privacy, and intellectual property concerns.

The Mayo Clinic’s competition, however, enlisted several data scientists and encouraged them to share data and work together to develop a reliable algorithm while competing for a $15,000 prize. Many participants had little or no experience with EEG or epilepsy.

The teams tested algorithms on nearly 350 seizures over more than 1,500 days, and the winners agreed to share their computer code for free. A medical device company called Neuro Vista Corporation made the data from dogs available to Mayo and other researchers. During the contest, over half of the crowdsourced algorithms outperformed random predictions. The best performing algorithms accurately predicted more than 70 percent of seizures when tested on unseen portions of the canine data. Mayo Clinic and Medtronic, Inc. will work together to test the safety and efficacy of seizure forecasting first in canines, followed by human trials.

Analysis: Crowdsourcing can allow organizations to dynamically source specialized skills from virtually anyone, anywhere. Companies can use this knowledge to help with simple tasks, such as data entry and coding, and more specialized activities like advanced analytics and product development.

The Mayo Clinic’s crowdsourcing experiment with epilepsy may help other research groups compare their algorithms to the ones generated from this competition. Long-term data from medical devices have the potential to help researchers better understand the disease process. Combining these data with the crowdsourcing process and use of open data may allow experts from many fields to come together to solve problems.

Source:
Benjamin H. Brinkmann et al, “Crowdsourcing reproducible seizure forecasting in human and canine epilepsy,” May 27, 2016, Brain.

This weekly series explores innovative breakthroughs and new technologies that are driving momentum and change in the life sciences and health care industry.

Author bio

Doug leads Deloitte Consulting LLP’s Life Sciences and Health Care practice. With 24 years of experience, he works closely with multiple top health care organizations on major clinical and enterprise transformation efforts and on large-scale technology implementation projects. Doug has extensive experience in comprehensive quality and patient safety transformations, turnaround and performance improvement in academic medical centers as well as organization/workflow redesign and technology enablement. He has served as the lead on a number of enterprise transformation initiatives with some of Deloitte’s most largest and most complex clients.