ADHD classification by a texture analysis of anatomical brain MRI data

Area of Science:

• Neuroimaging
• Medical Diagnostics
• Machine Learning

Background:

• Attention-Deficit/Hyperactivity Disorder (ADHD) diagnosis often relies on complex imaging techniques.
• Resting-state functional MRI (rs-fMRI) is used for ADHD classification, but requires extensive preprocessing.
• Structural MRI offers a potentially simpler alternative for diagnostic modeling.

Purpose of the Study:

• To develop and evaluate a novel ADHD classification method using only structural MRI data.
• To compare the diagnostic accuracy of a morphology-based approach with functional MRI methods.
• To identify key anatomical features contributing to ADHD classification.

Main Methods:

• Utilized isotropic local binary patterns on three orthogonal planes (LBP-TOP) for feature extraction from structural MRI.
• Employed support vector machines (SVM) to build classification models.
• Analyzed data from 436 male subjects (210 ADHD, 226 controls) using various parcellations and resolutions.

Main Results:

• Achieved a highest classification accuracy of 0.6995.
• LBP-TOP demonstrated superior discriminative power with whole-brain data.
• Higher image resolution and gray matter information improved model accuracy.
• Cortical folding patterns were identified as significant contributors to feature distribution.

Conclusions:

• A simple ADHD classification model can be effectively built using only anatomical MRI information.
• The LBP-TOP method offers a clinically feasible and less complex alternative to rs-fMRI.
• Structural MRI analysis, particularly gray matter morphology, holds significant potential for ADHD diagnosis.