4 Minutes
Spaceflight vision risks and a new predictive tool
Space travel poses many physiological challenges, and the eyes are particularly vulnerable. A condition known as spaceflight associated neuro-ocular syndrome (SANS) causes progressive vision changes in a substantial fraction of astronauts after extended missions. While some visual deficits improve after return to Earth, others can persist, making early identification of at-risk crewmembers essential for safe long-duration exploration.
A research team at the University of California, San Diego has developed an artificial intelligence (AI)–based screening method that predicts which astronauts are likely to develop SANS before launch. The model analyzes preflight eye imaging to detect structural patterns associated with vision changes that previously were observable only after exposure to microgravity.
Study design, data and AI approach
The investigators trained a deep learning model on a limited but carefully curated dataset of ocular scans. Because actual astronaut scans are few, the researchers augmented their data using images from terrestrial microgravity simulations as well as clinical eye exams from individuals who had not flown to space. To increase the number of training examples, each volumetric eye scan was divided into thousands of two-dimensional slices, enabling the neural network to learn fine-grained features across retinal layers.
Training was performed on a high-performance supercomputer at UC San Diego. The model learned to associate subtle changes in retinal anatomy—such as the retinal nerve fiber layer and the retinal pigment epithelium at the back of the eye—with later SANS diagnosis. Cross-validation on held-out preflight scans resulted in a predictive accuracy of roughly 82 percent, demonstrating that preflight imaging contains detectable biomarkers of susceptibility to microgravity-related vision loss.
Key findings and scientific implications
The AI highlighted nearly identical patterns of change in eyes exposed to true spaceflight and in those subjected to Earth-based microgravity analogs, indicating similar underlying mechanisms. This finding supports the use of ground-based simulations to study space health and to refine screening protocols before missions.
Beyond prediction, the model sheds light on possible SANS pathophysiology by identifying the retinal regions that most strongly influence risk classification. These regions point to fluid shifts, intracranial pressure changes, and structural retinal remodeling as probable contributors to vision decline in orbit. Understanding these mechanisms is crucial for designing targeted countermeasures such as improved in-flight monitoring, optical devices, or physiological interventions (for example, lower-body negative pressure or optimized exercise regimens).
Although the system is not yet approved for operational use, the researchers see it as the foundation for an integrated astronaut health toolchain. Incorporating AI-driven preflight screening into mission planning could enable personalized preventive strategies, informed crew selection, and real-time risk mitigation for long-duration missions to the Moon, Mars, or other deep-space destinations.
Broader context and next steps
SANS is one of several health risks associated with extended weightlessness; others include bone density loss, cardiovascular deconditioning, and neurocognitive changes. Expanding the training dataset with more astronaut scans, multimodal imaging, and longitudinal follow-up will be necessary to improve model robustness and generalizability. Future research will also examine combining AI predictions with physiological monitoring and countermeasure trials to lower vision risk in space.
Conclusion
AI-based analysis of preflight eye scans offers a promising approach to predict who is most likely to experience spaceflight-related vision decline. With continued data collection and model refinement, predictive screening could become a practical component of astronaut health planning, helping to protect vision on the next generation of long-duration missions.
Source: sciencealert
Leave a Comment