Identification of Autism Spectrum Disorder Using Topological Data Analysis

Area of Science:

• Neuroscience
• Data Science
• Medical Imaging

Background:

• Autism spectrum disorder (ASD) is a complex neurodevelopmental condition with increasing incidence.
• Effective feature representation is crucial for accurate ASD diagnosis and understanding.
• Traditional methods face challenges in capturing intrinsic data information and reducing complexity.

Purpose of the Study:

• To apply Topological Data Analysis (TDA) for feature extraction from brain imaging data in ASD.
• To evaluate the performance of TDA-based feature representation in classifying ASD subjects.
• To compare the efficacy of TDA with existing methods using large-scale datasets.

Main Methods:

• Utilized Regional Homogeneity (ReHo) data from the Autism Brain Imaging Data Exchange (ABIDE) database.
• Applied TDA techniques to extract topological features from ReHo data.
• Performed cross-validation on ABIDE I and ABIDE II datasets to assess classification accuracy.

Main Results:

• Achieved an average cross-validation accuracy of 95.6% on the ABIDE I database, surpassing existing methods (highest 93.59%).
• Attained an average accuracy of 96.5% on the ABIDE II database, significantly outperforming prior approaches (highest 75.17%).
• Demonstrated TDA's superior capability in identifying discriminative features for ASD.

Conclusions:

• TDA offers a powerful and effective method for ASD feature representation and classification.
• The TDA approach significantly enhances diagnostic accuracy for ASD compared to current techniques.
• This study highlights the potential of TDA in advancing neuroimaging-based research for developmental disorders.