fMRI-Based Multi-class DMDC Model Efficiently Decodes the Overlaps between ASD and ADHD
Zahra Zolghadr1, Seyed Amir Hossein Batouli2, Hamid Alavi Majd1
1Department of Biostatistics, School of Allied Medical Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran.
Basic and Clinical Neuroscience
|October 15, 2024
Summary
A new machine learning model accurately distinguishes between Attention Deficit Hyperactivity Disorder (ADHD), Autism Spectrum Disorder (ASD), and healthy individuals using neuroimaging data. This computational approach offers improved diagnostic capabilities for neurodevelopmental disorders.
Area of Science:
- Neuroscience
- Computational Psychiatry
- Machine Learning
Background:
- Neurodevelopmental disorders like ADHD and ASD share overlapping symptoms, complicating diagnosis and treatment.
- Current diagnostic methods rely on behavior, which can be imprecise due to symptom overlap.
- Neuroimaging offers potential for objective diagnostic tools.
Purpose of the Study:
- To develop and evaluate a computational framework for discriminating between ADHD, ASD, and healthy controls using functional neuroimaging data.
- To compare the performance of a novel Data Maximum Dispersion Classifier (DMDC) with Support Vector Machine (SVM).
Main Methods:
- Applied a two-level multi-class Data Maximum Dispersion Classifier (DMDC) algorithm.
- Utilized functional neuroimaging datasets from ADHD-200 and ABIDE.
- Classified participants into ADHD, ASD, or healthy control groups based on functional connectivity values.
Main Results:
- The DMDC model achieved an overall accuracy of 62%, with specific accuracies of 51% for healthy controls, 61% for ASD, and 84% for ADHD.
- The SVM model showed lower accuracy for healthy controls (46%) and ASD (46%) but matched ADHD accuracy (84%).
- The DMDC model demonstrated superior discrimination power, particularly for the ASD group.
Conclusions:
- The novel DMDC method shows acceptable performance in classifying neurodevelopmental disorders and healthy individuals, outperforming the SVM in certain aspects.
- Functional connectivity patterns, especially those involving the cerebellum, are key discriminators for these conditions.
