Identifying Non-Math Students from Brain MRIs with an Ensemble Classifier Based on Subspace-Enhanced Contrastive...

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Mathematical learning influences brain plasticity and cognitive functions.
• Previous research primarily used magnetic resonance spectroscopy to study biochemical brain changes.
• Identifying non-math students via structural brain imaging remains an underexplored area.

Purpose of the Study:

• To develop and validate a method for identifying non-math students using magnetic resonance imaging (MRI) scans.
• To investigate the structural brain differences associated with mathematical education.
• To explore the application of deep learning in classifying students based on brain imaging data.

Main Methods:

• Cropping the left middle frontal gyrus (MFG) region from MRI scans.
• Employing subspace enhanced contrastive learning for robust deep feature extraction.
• Utilizing an ensemble classifier with multiple-layer-perceptron models for student identification.

Main Results:

• Achieved 73.7% accuracy in image classification and 91.8% accuracy in student classification.
• Successfully identified students lacking mathematical education using MRI data.
• Demonstrated the efficacy of the proposed deep learning workflow on a dataset of 123 MRIs.

Conclusions:

• The proposed method effectively distinguishes between math and non-math students based on structural MRI data.
• This study offers insights into the impact of mathematical education on brain development.
• Highlights the potential of MRI and machine learning for understanding educational influences on the brain.