Fully Connected Cascade Artificial Neural Network Architecture for Attention Deficit Hyperactivity Disorder...

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Attention Deficit Hyperactivity Disorder (ADHD) is a complex neurodevelopmental disorder diagnosed behaviorally.
• Objective neuroimaging measures can significantly aid in the accurate classification of ADHD.
• International competitions provide valuable datasets for advancing ADHD diagnostic tools.

Purpose of the Study:

• To evaluate the efficacy of a fully connected cascade (FCC) artificial neural network (ANN) architecture for ADHD classification.
• To identify discriminative neuroimaging features for distinguishing ADHD from healthy controls and between ADHD subtypes.
• To explore the utility of brain connectivity patterns in understanding ADHD pathophysiology.

Main Methods:

• Utilized functional magnetic resonance imaging (fMRI) data from an international ADHD classification competition.
• Extracted features using various directional and nondirectional brain connectivity methods.
• Employed a fully connected cascade (FCC) artificial neural network (ANN) architecture for classification.
• Compared FCC ANN performance against other classifiers, such as support vector machines (SVMs).

Main Results:

• Achieved classification accuracy close to 90% for distinguishing ADHD from healthy subjects.
• Attained classification accuracy near 95% for differentiating between ADHD subtypes.
• Demonstrated that FCC ANN provides high accuracy irrespective of the features used.
• Identified reduced and altered connectivity in the left orbitofrontal cortex and cerebellar regions in ADHD patients.

Conclusions:

• FCC ANN architecture is a highly effective tool for classifying ADHD using fMRI data.
• Brain connectivity features offer significant discriminative power for ADHD diagnosis.
• The identified connectivity alterations provide insights into the neurobiological underpinnings of ADHD.