Development of a multimodal magnetic resonance imaging-based machine learning prediction model for flight cadets

Area of Science:

• Neuroscience
• Aerospace Medicine
• Machine Learning

Background:

• Current flight cadet selection and training methods in civil aviation are lengthy and subjective.
• Objective assessment tools are needed to enhance the accuracy and efficiency of evaluating flight aptitude.
• Understanding the neural underpinnings of flight-related skills can inform selection and training protocols.

Purpose of the Study:

• To develop and validate machine learning models using multimodal MRI data for distinguishing flight cadets from ground cadets.
• To identify neuroimaging features associated with advanced cognitive functions, visual processing, and attention allocation relevant to flight skills.
• To offer a data-driven approach for improving flight cadet selection and training evaluation.

Main Methods:

• Collected multimodal MRI data: structural MRI (sMRI), diffusion tensor imaging (DTI), and functional MRI (fMRI) from flight and ground cadets.
• Extracted and fused representative features from each MRI modality.
• Employed machine learning classifiers (logistic regression, random forest, support vector machine, Gaussian naive Bayes) with five-fold cross-validation.

Main Results:

• The multimodal fusion model combining sMRI, DTI, fMRI, and logistic regression achieved optimal performance.
• Achieved high accuracy (0.838), AUC (0.942), sensitivity (0.835), and specificity (0.834) in differentiating cadets.
• SHapley Additive exPlanations identified key features linked to cognitive functions, visual processing, and attention.

Conclusions:

• Multimodal MRI combined with machine learning offers a robust and objective method for flight cadet assessment.
• The study highlights the potential of neuroimaging and AI in understanding and enhancing flight-related skills.
• This approach can revolutionize flight cadet selection and training evaluation by reducing subjectivity and improving efficiency.