Kernel Principal Component Analysis for dimensionality reduction in fMRI-based diagnosis of ADHD
- Gagan Sidhu
- Nasimeh Asgarian, AICML
- Russ Greiner, Dept of Computing Science; PI of AICML
- Matt Brown
This study explored various feature extraction methods for use in automated diagnosis of Attention-Deficit Hyperactivity Disorder (ADHD) from fMRI data. Each participants data consisted of a resting state fMRI scan as well as phenotypic data (age, gender, handedness, IQ, and site of scanning) from the ADHD-200 dataset. Machine learning was used to produce support vector machine (SVM) classifiers that differentiated between ADHD patients and healthy controls. In some cases, ADHD subtypes were considered including the ADHD combined type (ADHD-c) and ADHD inattentive type (ADHD-i). We investigated six different learning tasks: differentiating between two classes (ADHD vs control) or among three classes (ADHD-c vs ADHD-i vs control), using only the phenotypic data or only the imaging data or else both phenotypic and imaging data. We tested the Fast Fourier Transform (FFT), different variants of Principal Component Analysis (PCA), and combinations of FFT and PCA. PCA variants included PCA over time (PCA-t), PCA over space and time (PCA-st), and kernelized PCA (kPCA-st). Baseline chance accuracy was 64.2% produced by guessing healthy control (the majority class) for all participants. Using only phenotypic data produced 72.9% accuracy on two-class diagnosis and 66.8% on three-class diagnosis. Diagnosis using only imaging data did not perform as well as phenotypic-only approaches. Combining phenotypic and imaging data with FFT and kPCA-st yielded accuracies of 76.0% on two-class diagnosis and 68.6% on three-class diagnosis, better than phenotypic-only approaches. Our results demonstrate the potential of using kPCA-st with resting-state fMRI data as well as phenotypic data for automated diagnose of ADHD. These results are encouraging given established challenges with diagnosis of ADHD using the ADHD200 dataset (see Brown et al. 2012, Frontiers of Systems Neuroscience 6:69).
Citation
G. Sidhu, N. Asgarian, R. Greiner, M. Brown. "Kernel Principal Component Analysis for dimensionality reduction in fMRI-based diagnosis of ADHD". Frontiers in Systems Neuroscience, (ed: Michael Milham), 6(74), October 2012.| Keywords: | fmri, diagnosis, machine learning, neuroscience, dimensionality reduction, pca, kpca, medical informatics, computational psychiatry, ADHD-200 |
| Category: | In Journal |
| Web Links: | DOI |
| article | |
| Related Publication(s): | ADHD-200 Global Competition: Diagnosing ADHD using personal characteristic data can outperform resting state fMRI measurements |
BibTeX
@article{Sidhu+al:12,
author = {Gagan Sidhu and Nasimeh Asgarian and Russ Greiner and Matt Brown},
title = {Kernel Principal Component Analysis for dimensionality reduction in
fMRI-based diagnosis of ADHD},
Editor = {Michael Milham},
Volume = "6",
Number = "74",
journal = {Frontiers in Systems Neuroscience},
year = 2012,
}Last Updated: February 10, 2020Submitted by Sabina P
