Multimodal Imaging-Based Cerebral Blood Flow Prediction Model Development in Simulated Microgravity.

Area of Science:

• Neuroscience
• Space Medicine
• Biomedical Engineering

Background:

• Cerebral blood flow (CBF) alterations are linked to cognitive decline and neurodegeneration.
• Maintaining adequate CBF is critical for astronaut brain health during long-duration spaceflight.
• Quantitative CBF assessment in microgravity presents significant challenges.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting CBF changes under simulated microgravity.
• To investigate the relationship between internal carotid artery Doppler ultrasound and brain MRI data for CBF mapping.
• To create a practical tool for monitoring astronaut brain health in space.

Main Methods:

• A 90-day head-down tilt bed rest (HDTBR) protocol simulated microgravity conditions.
• Multimodal imaging data, including Doppler ultrasound and brain MRI, were collected.
• Various ML algorithms were employed to build predictive models for regional CBF.

Main Results:

• Significant regional CBF decreases were observed in specific brain areas after 90 days of HDTBR.
• The CatBoost ML model demonstrated high predictive accuracy for CBF in affected regions (AUCs ranging from 0.82 to 0.92).
• The prediction model was successfully deployed as an interactive web application for potential in-orbit use.

Conclusions:

• ML-driven CBF prediction models can be feasibly constructed using multimodal imaging data under simulated microgravity.
• These models show potential as early warning systems for monitoring astronaut brain health during space missions.
• The developed application offers a practical solution for in-orbit CBF monitoring.