ADHD diagnosis guided by functional brain networks combined with domain knowledge

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Functional magnetic resonance imaging (fMRI) is crucial for modeling functional brain networks (FBNs) in attention-deficit/hyperactivity disorder (ADHD) research.
• Current FBNs-based methods struggle to capture regional correlations and long-distance dependencies (LDDs), impacting diagnostic accuracy.
• Limited sample sizes and class imbalance hinder the performance of ADHD diagnostic models.

Purpose of the Study:

• To propose an automated diagnostic framework integrating modeling, multimodal fusion, and classification for ADHD.
• To enhance the extraction of representative FBNs and incorporate domain knowledge for improved ADHD classification.
• To address limitations in capturing dynamic brain activity and data imbalance in ADHD research.

Main Methods:

• A multi-head attention-based region-enhancement module (MAREM) was developed to capture both regional correlations and LDDs in brain activity.
• A multimodal supplementary learning module (MSLM) was proposed to integrate neuroimaging FBNs with phenotype data, addressing data scarcity and imbalance.
• An ADHD automatic diagnosis framework guided by FBNs and domain knowledge (ADF-FAD) was implemented and tested on the ADHD-200 dataset.

Main Results:

• MAREM effectively extracts FBNs suitable for modeling and classification tasks.
• The integrated ADF-FAD framework achieved high diagnostic accuracies: 92.4% (NYU), 74.4% (PU), and 80% (KKI).
• The study demonstrates the framework's ability to capture critical information for ADHD diagnosis, outperforming existing methods.

Conclusions:

• The proposed ADF-FAD framework offers a robust and accurate method for ADHD diagnosis using fMRI-derived FBNs.
• Integrating multimodal data and advanced attention mechanisms significantly improves diagnostic performance and addresses data limitations.
• This approach holds promise for aiding clinicians in making more accurate ADHD diagnostic decisions.