Original Research ARTICLE

Front. Syst. Neurosci., 06 August 2012 | http://dx.doi.org/10.3389/fnsys.2012.00058

Individual classification of ADHD patients by integrating multiscale neuroimaging markers and advanced pattern recognition techniques

  • 1Centre for Computational Systems Biology, Fudan University, Shanghai, P.R. China
  • 2Mathematical Department, Zhejiang Normal University, Jinhua, Zhejiang Province, P.R. China
  • 3Department of Computer Science, Warwick University, Coventry, UK

Accurate classification or prediction of the brain state across individual subject, i.e., healthy, or with brain disorders, is generally a more difficult task than merely finding group differences. The former must be approached with highly informative and sensitive biomarkers as well as effective pattern classification/feature selection approaches. In this paper, we propose a systematic methodology to discriminate attention deficit hyperactivity disorder (ADHD) patients from healthy controls on the individual level. Multiple neuroimaging markers that are proved to be sensitive features are identified, which include multiscale characteristics extracted from blood oxygenation level dependent (BOLD) signals, such as regional homogeneity (ReHo) and amplitude of low-frequency fluctuations. Functional connectivity derived from Pearson, partial, and spatial correlation is also utilized to reflect the abnormal patterns of functional integration, or, dysconnectivity syndromes in the brain. These neuroimaging markers are calculated on either voxel or regional level. Advanced feature selection approach is then designed, including a brain-wise association study (BWAS). Using identified features and proper feature integration, a support vector machine (SVM) classifier can achieve a cross-validated classification accuracy of 76.15% across individuals from a large dataset consisting of 141 healthy controls and 98 ADHD patients, with the sensitivity being 63.27% and the specificity being 85.11%. Our results show that the most discriminative features for classification are primarily associated with the frontal and cerebellar regions. The proposed methodology is expected to improve clinical diagnosis and evaluation of treatment for ADHD patient, and to have wider applications in diagnosis of general neuropsychiatric disorders.

Keywords: ADHD, functional brain networks, pattern classification, fALFF, ReHo, BWAS

Citation: Cheng W, Ji X, Zhang J and Feng J (2012) Individual classification of ADHD patients by integrating multiscale neuroimaging markers and advanced pattern recognition techniques. Front. Syst. Neurosci. 6:58. doi: 10.3389/fnsys.2012.00058

Received: 31 March 2012; Accepted: 19 July 2012;
Published online: 06 August 2012.

Edited by:

Stewart H. Mostofsky, Kennedy Krieger Institute, USA

Reviewed by:

Volker Steuber, University of Hertfordshire, UK
Ani Eloyan, Johns Hopkins University, USA

Copyright © 2012 Cheng, Ji, Zhang and Feng. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in other forums, provided the original authors and source are credited and subject to any copyright notices concerning any third-party graphics etc.

*Correspondence: Jie Zhang and Jianfeng Feng, Centre for Computational Systems Biology, Fudan University, Handan Road 220, Shanghai, 200433, P.R. China. e-mail: jzhang080@gmail.com; jianfeng64@gmail.com

These authors equally contributed to this work.