Classification of functional brain images with a spatio-temporal dissimilarity map

Area of Science:

• Neuroimaging
• Machine Learning
• Clinical Neuroscience

Background:

• Functional magnetic resonance imaging (fMRI) data analysis is crucial for understanding brain function.
• Classifying subjects into predefined groups (e.g., patient vs. control) has significant clinical diagnostic potential.
• Existing methods may lack automated feature extraction for group classification.

Purpose of the Study:

• To present an automated methodology for classifying subjects into groups based on whole-brain functional MRI data.
• To develop a generalizable framework for identifying and utilizing spatio-temporal features for subject categorization.
• To validate the proposed method through simulations and a clinical application for schizotypy classification.

Main Methods:

• Utilizing preprocessed time series data from the whole brain volume.
• Employing a machine learning approach to identify distinguishing spatio-temporal features from a training set.
• Applying the trained model to categorize new subjects into predefined groups.

Main Results:

• The automated method successfully identified relevant spatio-temporal features differentiating study groups.
• Demonstrated efficacy through simulations, confirming the method's robustness.
• Successfully classified individuals into schizotypy and control groups in a clinical application.

Conclusions:

• The proposed automated method offers a viable approach for group classification using fMRI data.
• This methodology can aid in clinical diagnostics by leveraging neuroimaging biomarkers.
• The framework is adaptable for various clinical applications requiring subject group differentiation.