Application of Oculomics Artificial Intelligence with Astronaut Fundus Images to Assess Systemic Health During Spaceflight

Abstract

Purpose :

 With a future mission to Mars and a return to the Moon, mankind is breaking barriers in spaceflight exploration. However, current diagnostic capabilities onboard spacecraft and the International Space Station (ISS) are inconvenient for monitoring systemic health. Our current project, export controlled by NASA, is the first implementation of oculomics to fundus photos taken in the spaceflight environment. It promises to provide further insights into physiological changes occurring in spaceflight, potentially revolutionizing health monitoring in future exploration missions.

Methods : 

Using published fundus photos taken of astronauts in the pre-flight (L), in-flight (D), and post-flight R) periods, we applied established oculomics algorithms predicting systemic biomarkers, including the retinal age gap. The retinal age gap is the difference between chronological age and the age predicted by the deep learning algorithm. We obtained a statistical significance threshold of alpha = 0.05, one-way ANOVA testing, and post hoc analysis of average values with Benjamini-Hochberg correction. For the retinal age gap, we determined the astronaut’s average age using values published in the literature and performed an additional t-test comparing the retinal age gap to a null value of zero.

Results : 

Table 1 (A, B) demonstrates the retina age gap values among the fundus photos. The greatest retinal age was recorded during the in-flight period, with an age gap of 14.3 ± 1.78 years (p=0.0134). However, retina age gap during the pre-flight (8.18) and post-flight (5.81) periods were both statistically significant. In-flight comparisons against both L (p<0.0001) and R (p<0.0001) phases were statistically significant. The most significant increases were seen in the R versus D phases.

Conclusions :

 The retinal age gap is transiently elevated in spaceflight. The retinal age gap has a significant association with cardiovascular disease, cancer mortality and neurological disease. In future models, we hope to refine current algorithms and determine validity against astronaut characteristics. Given the availability of fundus imaging and OCT onboard the ISS and spacecraft, we could enhance our ability to monitor astronaut health by ultilizing the systemic biomarkers included in our analysis.

This abstract was presented at the 2025 ARVO Annual Meeting, held in Salt Lake City, Utah, May 4-8, 2025.