NASA and Google AI Medical Assistant for Mars Astronauts

The Growing Need for Autonomous Medical Care in Space

As the duration and distance of human spaceflight missions increase, maintaining astronaut health presents escalating challenges.

Currently, astronauts aboard the International Space Station benefit from immediate communication with mission control in Houston, consistent delivery of medical supplies, and a relatively swift return to Earth after a six-month deployment. However, this support structure is anticipated to change significantly.

NASA, alongside commercial partners such as SpaceX, is planning extended missions to the Moon and Mars. This necessitates a shift towards more self-sufficient medical capabilities for crews in deep space.

Introducing the Crew Medical Officer Digital Assistant (CMO-DA)

In response to this evolving need, NASA is developing a proof-of-concept AI medical assistant in collaboration with Google. This initiative aims to enhance on-orbit medical care, reducing reliance on Earth-based support.

The Crew Medical Officer Digital Assistant (CMO-DA) is designed to aid astronauts in diagnosing and treating medical issues when direct access to physicians is unavailable or communication with Earth is disrupted.

This multimodal tool leverages speech, text, and image analysis, operating within Google Cloud’s Vertex AI environment.

Technical Details and Development

The project operates under a fixed-price subscription agreement with Google Public Sector, covering cloud services, application development infrastructure, and model training.

David Cruley, a customer engineer at Google’s Public Sector business unit, explained to TechCrunch that NASA retains ownership of the application’s source code and has actively participated in model refinement.

The Google Vertex AI platform grants access to a diverse range of models, including those developed by Google and other third-party providers.

Initial Testing and Performance

CMO-DA has undergone evaluation through three simulated medical scenarios: an ankle injury, flank pain, and ear pain.

A team of three physicians, including a practicing astronaut, assessed the assistant’s performance across key areas, including initial evaluation, medical history acquisition, clinical reasoning, and treatment recommendations.

The results demonstrated a promising level of diagnostic accuracy:

• Flank pain evaluation: 74% likelihood of correct diagnosis and treatment.
• Ear pain evaluation: 80% likelihood of correct diagnosis and treatment.
• Ankle injury evaluation: 88% likelihood of correct diagnosis and treatment.

Future Development and Potential Applications

NASA’s development roadmap for CMO-DA is intentionally phased. Future plans include integrating additional data sources, such as readings from medical devices.

Furthermore, the model will be trained to exhibit “situational awareness,” specifically recognizing and responding to medical conditions unique to the space environment, like those induced by microgravity.

While Google has not confirmed plans to seek regulatory approval for terrestrial use, the potential for adapting this technology to improve healthcare on Earth is significant.

Cruley emphasized that the insights gained from this tool could extend beyond space medicine, offering benefits to various healthcare domains.