The Unique Potential of Open Federated Learning in Healthcare

 

Historically, people have opted into research, shared their health data, or taken part in patient surveys. Data had to be collected in a central location with the consent of the users. These paths keep private information protected but unfortunately, it eliminates a considerable amount of data available. However, with federated learning, it’s possible to collect confidential information, maintain anonymity, and increase data quality. Intel’s Morgan Andersen spoke with Andrew Lamkin, Software Product Manager, and Patrick Foley, the Lead Architect of OpenFL, to discuss a groundbreaking use case of Open Federated Learning (OFL) and the exciting potential of the model. “This goes a little different than our other technologies at Intel, as it is an open-source one,” said Andersen.

OpenFL is a federated learning model. Google introduced federated learning in 2017 to improve text prediction without taking identifying information from Android users (Open Zone). “The short of it is that federated learning deals with sending the model to where the data resides, out at the edge, instead of sending data to a central place for the purpose of training,” said Foley.

Intel collaborated with the University of Pennsylvania and applied federated learning to healthcare. The use case applied explicitly to brain tumors, identifying the lines of the tumors, and determining which ones were operable and inoperable. The study brought in 71 research institutions from around the world. “these models were able to identify operable tumor regions 33 percent better than a model that was trained on public data alone,” said Foley. The study was instrumental in proving that the model was effective in medical research and that it protected patient information.

Python is at the core of OpenFL. “The project, very early on, had the realization of ‘your main customers are data scientists,’ right? So meeting them where they are and in the tool sets that they work, was really kind of crucial to getting things going and being as fast and sot of as robust as things are today,” said Lamkin. Python is used widely in deep learning, ideal use for OpenFL. OpenFL also links to Jupiter Notebook, where models are developed. With enough demand, Intel’s OpenFL could grow to work with additional programs. “There’s really only a few multinational companies that could really attempt to centralize all the data sets. I mean that have a presence in enough hospitals, that have a presence globally, at a global scale to put it all together,” said Lamkin. Five years ago, the option to source data on this scale was impossible. With OpenFL, researchers have a remarkable ability to develop insights from confidential data in all industries.

Learn more about OpenFL by visiting the OpenFL Github and joining the OpenFL Slack Community or connecting with Andrew Lamkin and Patrick Foley on LinkedIn. Don’t forget to subscribe to this channel on Apple Podcasts, Spotify, and Google Podcasts to hear more from the Intel Health and Life Sciences at the Edge.

 

Follow us on social media for the latest updates in B2B!

Image

Latest

No Fail Trust Methodology
Empowering Teams with the No Fail Trust™ Methodology: Insights From U.S. Air Force Pilot Jason O. Harris
February 7, 2025

On this episode of the Through the Storm podcast, Dr. Travis Hearne speaks with Retired U.S. Air Force Pilot, Global Keynote Speaker, and all-around badass, Jason O. Harris. Lieutenant Colonel Jason Harris is a highly sought-after keynote speaker, best-selling author, expert thought leader, and professional pilot. As a fourth-generation military member, Jason’s distinguished career…

Read More
The Art of Communication in Disaster Recovery: Building Trust and Delivering Results
The Art of Communication in Disaster Recovery
February 6, 2025

In this episode of Inside Restoration and Recovery, host Martha Lewis sits down with Chris Matthews, Regional Director for Large Loss at BMS CAT, to explore the critical role of communication, preparation, and adaptability in disaster recovery. Chris shares his personal journey in the restoration industry, including his passion for helping clients during their…

Read More
women’s health
AI and Advocacy Unite: Tackling Women’s Health Disparities with Hologic Mia Keeys
February 5, 2025

Women’s health remains a critical issue globally, with alarming disparities in access and outcomes. A new survey from Hologic and Gallup reveals that while nearly 90% of women recognize the importance of health screenings, less than half undergo them consistently due to barriers like cost, time, and anxiety. The stakes are particularly high in…

Read More
toast
The Framework Series Advantage: Using Toast to Visualize and Refine Business Processes
February 5, 2025

Most organizations face challenges when aligning teams on processes and strategies. According to a 2021 McKinsey Global Survey, fewer than one-third of companies successfully improve performance and sustain those improvements over time, often due to a lack of employee engagement and poor execution. The key to overcoming this? A simple exercise involving sticky notes and…

Read More