We have the ability to transform healthcare for children who are not geographically or socially lucky using artificial intelligence (AI) applications. The effort to build accurate, real-time, privacy-preserving AI applications for children’s medicine will require access to data from all 1,000,000 healthcare machines in all 500 children’s hospitals in the world — a pediatric moonshot. To make progress towards reaching the “moon,” we are building a Federated Learning Lab designed to answer a wide range of research questions.

Can the software developed for consumer AI work in pediatric AI? What strategy should we use for learning? Given that it’s impossible to develop AI methodologies in the absence of data, with the Federated Learning Lab we intend to address and overcome that challenge. Focusing first on the specific area of cardiology, the Federated Learning Lab for Children’s Medicine will start with five components.

  1. A decentralized in-the-building edge cloud service

As with the original moonshot, we too need a new rocket. BevelCloud has engineered decentralized, in-the-building edge cloud service. Initially, there are five component services: edge compute, edge storage, edge network, edge data, and edge application services. These services were engineered with 37 security-by-design features, including fine-grain data-sharing mechanisms. The edge servers also provide image sanitization with software partner, Glendor. Much like Apple has created a way to develop and deploy consumer applications, BevelCloud is doing the same, first with pediatric cardiology and then with orthopedic, radiology, cancer, and neonatology applications.

This BevelCloud infrastructure offers solutions to some of the challenges that have previously plagued consumer federated learning applications. In particular, BevelCloud edge servers have continuous high bandwidth communication with the healthcare machines within the edge zone, as well as secure network communications outside the edge zones. Second, the edge servers are always powered on —this is important if you consider the hours of use for a typical ultrasound machine leave at least 16 hours, 7 days a week of available compute power that can be dedicated to federated learning. Finally, because of edge data services architecture— every edge cloud application accesses identically formatted data whether the edge servers are in Vatican City, Orange County, or Sao Paolo.

  1. One hundred servers, one hundred twinned ultrasounds

The second component of the Federated Learning Lab is to deploy the edge cloud in-the-building of six children’s hospitals on two continents. Furthermore authorized applications will have real-time access to non-compressed images from 100 ultrasounds by 100 dedicated edge servers.

  1. Large, continuous diverse data sharing

Twinning 100 ultrasound machines will provide AI researchers access to real-time and raw (non-compressed) ultrasound data. As the edge servers are twinned directly, applications will not suffer from the current challenges of pulling compressed data from a variety of picture archiving and communication system (PACS) applications. During the first year of operation, a federated learning application can be trained on over 100,000 Terabytes of diverse ultrasound data.

  1. Open-source cardiology AI app as an edge cloud application

The Federated Learning Lab will also be supplied with an open-source, AI cardiology neural network application designed to measure ejection fraction on an ultrasound image. While this may not be the most trail-blazing AI application, it will enable every experiment with the same starting point to benchmark improvements. The application will be available as an edge cloud application.

  1. Multi-site, multi-country IRB

Finally, in coordination with Dr. Reddy at Stanford, we are establishing a multi-site, multi-country, IRB protocol to govern the lab. Research users of the Federated Learning Lab will have access to the collective expertise of pediatric cardiologists who can provide domain expertise to the users of the lab.

It then remains for a variety of federated learning researchers to answer a wide range of research questions. Starting with the most basic, will the software developed for consumer AI be applicable to pediatric AI? While most work in federated learning for healthcare has occurred in the cross-silo model, the Federated Learning Lab will operate in the more familiar cross-device model. So how well will they work in a new domain? There are many software stacks to consider.

  •  Acuratio is an enterprise platform with solutions for horizontally and vertically partitioned data
  • DynamoFL simplifies model training across privacy-critical datasets using Federated Learning and Differential Privacy
  • FedML provides an open-source community, as well as an enterprise platform for open and collaborative AI
  • FLARE (Federated Learning and AI for Robotics and Edge) is a software stack built for deploying AI models on edge devices
  • Flower an open source research platform for training models in a federated manner
  • HP Swarm Learning focused on de-centralizing the aggregation step
  • OpenFL is a Python* 3 library for community supported projects, originally developed by Intel Labs and the Intel Internet of Things Group
  • NimbleEdge focuses on hyper-personalized machine learning on mobile edge
  • PySyft is an open-source library developed for secure and private federated learning
  • TensorFlow Federated (TFF) is an open-source framework developed by Google for implementing federated learning

Even beyond the question of which federated learning software lie many other important questions that computer scientists actively working in this area will be able to address, including:

  • What are the implications of relaxing the traditional constraints of consumer federated learning on federated learning for medicine? 
  • Does relaxing the constraints change any of the features provided by the federated learning software?
  • Does one aggregation strategy work better than others? Should you aggregate neural network weights within a zone before aggregating globally?
  • Should we consider split learning (learning on just half of the neural network model) before sending results to the aggregation server, as proposed by the MIT team?
  • When should a model change be declared as the production version, given the potential for continuous learning?

These are but a small fraction of the important research questions to be answered by the Federated Learning Lab.

This fascinating topic of federated learning, along with others will be discussed at the annual AIMed Global Summit, scheduled for June 4-7th 2023 in San Diego. Book your place now! 

We believe in changing healthcare one connection at a time. If you are interested in the opinions in this piece, in connecting with the author, or the opportunity to submit an article, let us know. We love to help bring people together! [email protected]