In 2021, Frontier Development Lab (FDL) researchers worked with Intel AI Mentors to conduct a landmark health study on the dangers of radiation to learn t its physiological impacts on astronauts with Mars as a potential location for future long-duration space flights. With the use of Intel’s AI technology, FDL devised a first-of-its-kind algorithm to recognize the biomarkers of cancer progression. They do this by utilizing the combination of rats and human data information concerning radiation exposure.
Health risk to astronauts
Cosmic radiation can infiltrate several layers of steel and aluminum, affecting human tissue during space travel. This can affect an astronaut’s health and create eventual cancer complexities. As such, Intel and FDL collaborated to create causal machine learning across a federation of collaborator institutes to access this siloed data.
They can share an AI algorithm to train it on data stored in separate locations without sharing the data. The causal machine learning method launched the researchers’ scientific challenge to foretell the genes affected by radiation more accurately, some relating to cancer and others to immunity response.
Can we use #AI to better understand the effects of radiation exposure on a mission to Mars? Watch their results live 13 Aug https://t.co/fS23GKSEx7 @FDL_Ai @NASA @NASAGeneLab @NASAAmes @SETIInstitute @googlecloud @intel @MayoClinic #FDL #FDL2021 #machinelearning #AstronautHealth pic.twitter.com/DyH7lcAlSX
— FDL (@FDL_AI) August 7, 2021
Research framework
This research enables Intel’s Open Federated Learning (OpenFL) framework prepared by the Intel and FDL researchers. This makes it achievable to integrate CRISP 2.0 models from organizations like NASA and NASA’s Gene Lab without needing to move the data into the central node. The use of the framework proved to be significant because, even though each organization had the right to use the data, the data was private. The cost of transmitting data generated aboard a spacecraft was also expensive, so this also helps cut costs.
Each institution sent a group of global models to conduct a single round of AI training. The models were then returned to the central node to be accumulated and reshared to the partner institutions. Finally, CRISP 2.0 was utilized for further study and investigations.
Photo credit: The feature image used is owned by NASA, and the image in the body of the article is owned by Frontier Development Lab. Both photos have been provided via Intel for press usage.
