Classification of ADHD Individuals and Neurotypicals Using Reliable RELIEF: A Resting-State Study

• Published in: IEEE access Vol. 7; pp. 62163 - 62171
• Main Authors: Miao, B., Zhang, L. L., Guan, J. L., Meng, Q. F., Zhang, Y. L.
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Attention deficit hyperactivity disorder; Deep learning; Entropy; Entropy (Information theory); Feature extraction; Feature selection; Functional magnetic resonance imaging; image classification; Linear regression; Magnetic resonance imaging; Medical diagnosis; Principal component analysis; Principal components analysis; Regression analysis; Reliability
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2915988

Fractional amplitude of low-frequency fluctuation (fALFF) can reflect the intensity of spontaneous neuronal activity. Feature selection on functional magnetic resonance imaging (fMRI) data combined with fALFF can be well used to study the pathology of attention deficit hyperactivity disorder (ADHD) and assist in its diagnosis. However, the unsatisfactory effect of feature selection limits the study of ADHD. In this study, a novel method is proposed to classify ADHD individuals and neurotypicals. This work introduces multiple linear regressions for reduction of confounding effects. After that, fALFF combined with principal component analysis (PCA), Shannon entropy (ShEn), and sample entropy (SampEn) are used to construct features, and a reliable RELIEF (R-RELIEF) algorithm is proposed to find the most prominent features. Public ADHD-200 dataset is used in this study to evaluate our method. The results in this study suggest that fALFF is a reliable fMRI marker for investigation of ADHD, and R-RELIEF shows good performance compared with RELIEF algorithm. Moreover, fALFF of many brain regions are found to differ between ADHD and neurotypicals, thus coinciding with the results of previous studies.