Frontier Development Lab (FDL) researchers has conducted a landmark astronaut health study with Intel AI Mentors to better understand the physiological effects of radiation exposure on astronauts. Using Intel artificial intelligence (AI) technology, FDL created a first-of-its-kind algorithm to identify the biomarkers of cancer progression using a combination of mouse and human radiation exposure data.
“With help from Intel we formulated how causal machine learning models can operate on data across different locations without having to move the data between physical locations. We achieved our goal during FDL 2021 to use bespoke algorithms to better understand, improve and support astronaut health. This research is so valuable; it could one day help astronauts at the International Space Station, future space stations and on the upcoming 2024 lunar mission, as well as those affected by cancer on Earth,” said Paul Duckworth, FDL researcher.
Cosmic radiation can penetrate several layers of steel and aluminum to affect human tissue during space travel. It can lead to astronaut health problems and future cancer complications. With little data on the effects of cosmic radiation on astronauts from existing space missions, researchers leveraged datasets from human astronaut data, which are heavily protected by various institutions. To access this siloed data, Intel and FDL formulated causal machine learning across a federation of collaborator institutes whereby they can share an AI algorithm, in order to train it on data stored in separate locations, without the need to share the data.
“The FDL Astronaut Health team achieved some truly incredible results in this year’s challenge – both in their novel combination of human and mice data and the identification of several causal genes responsible for cancer,” said Patrick Foley, Intel lead technical mentor. “This work is a testament to what can happen when public and private institutions work together and how federated learning can be used to unlock discoveries that would otherwise remain buried. We are confident that this research will go on to drive better health outcomes for astronauts and enrich the lives of every person on Earth.”
How It Works: Researchers developed CRISP 2.0 by extending CRISP 1.0 from the 2020 FDL astronaut health team. With CRISP 2.0, the 2021 astronaut health team proved rodent radiation data can be used as a homologue of human radiation data, which is used to train the human algorithm. The causal machine learning method tackles the researchers’ scientific challenge to more accurately predict the genes that will be affected by radiation, some relating to cancer and others to immunity response.
This research leveraged Intel’s Open Federated Learning (OpenFL) framework, set up by Intel and FDL researchers on Google Cloud, to make it possible to train and combine CRISP 2.0 models from institutions such as NASA, Mayo Clinic and NASA’s Gene Lab without moving the data to a central place. This was crucial because, even though each organization had the necessary right to use the data, the data was private and the cost of transmitting data that could be generated aboard a spacecraft was high. Each institution was sent a group of global models that were used to do one round of AI training on each local dataset. The models were then sent back to the central node to be aggregated and reshared to the collaborating institutions. Finally, CRISP 2.0 was used to output results for further analysis and insights.
“Partnering with Frontier Development Lab is a chance to solve huge problems at scale, with cutting-edge technology and unprecedented collaboration between the public and private sectors in AI for exploration medicine,” said Shashi Jain, strategic innovation and FDL partner manager at Intel. “We believe that the FDL Astronaut Health challenge results will enable NASA to understand the mechanisms involved in protecting astronauts more effectively as we return to the moon and beyond, as well as provide a blueprint to accelerate the use of AI in healthcare applications on Earth.”