Identifying individuals with attention deficit hyperactivity disorder based on temporal variability of dynamic...

Area of Science:

• Neuroimaging
• Psychiatry
• Computational Neuroscience

Background:

• Attention Deficit Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder in school-aged children.
• Current diagnostic methods for ADHD can be improved with objective neuroimaging biomarkers.
• The dynamic functional connectivity patterns in ADHD are not fully understood.

Purpose of the Study:

• To develop and validate a diagnostic model for ADHD using neuroimaging features.
• To investigate the role of temporal variability between intrinsic connectivity networks (ICNs) in ADHD.
• To identify discriminative neuroimaging patterns for ADHD diagnosis.

Main Methods:

• Utilized data from 100 children with ADHD and 140 controls from the ADHD-200 Consortium.
• Extracted features based on temporal variability between ICNs, alongside demographic and covariate data.
• Employed Support Vector Machines (SVMs) for classification, validated with Leave-One-Out Cross-Validation (LOOCV) and 10-fold Cross-Validation (CV).

Main Results:

• The diagnostic model based on inter-ICN variability demonstrated superior performance compared to models using functional connectivity or phase synchrony.
• LOOCV yielded an overall accuracy of 78.75%, sensitivity of 76%, and specificity of 80.71%.
• 10-fold CVs achieved an average accuracy of 75.54% ± 1.34%, sensitivity of 70.5% ± 2.34%, and specificity of 77.44% ± 1.47%.

Conclusions:

• Inter-ICN variability serves as a potential neuroimaging biomarker for identifying individuals with ADHD.
• The developed SVM-based model effectively distinguishes ADHD patients from controls.
• Discriminative patterns identified through SVMs align with existing findings, supporting their clinical relevance.