Stephen Strother

Research Synopsis

Functional Neuroimaging Research

My lab is focused on optimizing the imaging and dataanalysis processing pipelines of functional neuroimaging studies of the brain activity associated with human cognition, disease and behaviour. My present academic research focus is on functional magnetic resonance imaging (fMRI) and associated structural MRI techniques in the middle and older age ranges most relevant as control groups for clinical studies. These groups, particularly the middle-aged range, have been little studied uasing clinically relevant neurobehavioural tasks. The hypothesis guiding this work is that standard default image acquisition and processing pipelines are not optimal, and that as a result new insights into human brain function in normals and disease are being obscured by poor and/or limited pipeline choices.

In particular we are using and developing a Java/Matlab/C-based framework for testing pipeline choices (dubbed NPAIRS: Strother et al., 2002, Neuroimage) using statistical learning theory and machine learning techniques. Using NPAIRS and related approaches we are developing optimized experimental neuroimaging protocols using multiple behavioural tasks that take only a short time to complete, and are suitable as clinical probes of cognitive brain function. Our goal is to develop a database library of such probes to assist in designing new multitask protocols that can be tailored to a particular clinical or research goal. We are particularly focused on developments that will support clinical fMRI for vascular cognitive impairment (e.g., cognitive deficits of stroke, dementias, etc) and its comparison with clinical behavioural testing.

My primary appointment is at the Rotman Research Institute of the Baycrest Centre for Geriatric Care and as a result I work closely with cognitive psychologists and research clinicians. All of this research is highly collaborative and some is funded as part of major international consortia based in the US and Denmark, and as a result involves collaborations that exchange students and data with international groups. In the past I gained extensive experience with PET imaging, which I now continue through a company I cofounded (www.Predictek.com), which is based in Chicago, and through collaborations with the Centre for Addiction and Mental Health in Toronto, and the Centre for Integrated Molecular Brain Imaging (www.cimbi.dk) in Denmark.

Current research areas include, (1) optimization of acquisition and processing pipelines for fast fMRI imaging of vascular cognitive impairment and stroke using multi-task test batteries based on clinical behavioural tests, (2) development of theoretical and practical frameworks for testing the performance of functional neuroimaging systems to increase the yield of scientific and clinically relevant information, (3) continued development and dissemination of such a framework for optimizing fMRI & PET image and data-analysis processing pipelines, (4) study of image and data-analysis algorithms for fMRI/MRI and PET with an emphasis on multivariate techniques from the machine learning community (e.g., linear classifiers, neural networks, support vectors, etc), (5) development of sequential acquisition techniques that allow a researcher to predict how much data they must collect both within and between subjects for PET and fMRI/MRI.

Recent completed and ongoing thesis and postdoctoral research topics include, (1) Assessing the Efficacy of Preprocessing Choices: Residual Motion Artifacts as a Function of Task and Age, (2) the Interaction between Age and fMRI Intenstity Normalisation across Tasks: GLM Results, (3) Partially Optimised Preprocessing Pipelines for the Clinically-Relevant Task Trails A/B, (4) Testing fMRI Signal Detection: Dependence on ROC Technique and Simulation Structure. Members of my laboratory have presented these topics as posters at the annual meetings of the International Society for Magnetic Resonance in Medicine and the Organisation of Human Brain Mapping during the last two years, and are now preparing papers for submission to the two major journals in our field: Neuroimage and Human Brain Mapping.

For further information, please go visit our lab homepage.

Recent Publications

• Strother SC, LaConte S, Hansen LK, Anderson J, Zhang J, Pulapura S, Rottenberg D. Optimizing the fMRI Data-Processing Pipeline Using Prediction and Reproducibility Performance Metrics. Neuroimage, 23S1:S196-S207, 2004.
• LaConte S, Strother SC, Cherkassky V, Anderson J, Hu X. Support Vector Machines for Temporal Classification of fMRI Data. Neuroimage, 26:317-329, 2005.
• Strother SC. Evaluating fMRI Preprocessing Pipelines. IEEE Eng. Med. Biol. Mag. 25(2):27-41, 2006.
• Poline J-B, Strother SC, Dehaene-Lambertz G, Egan G, Lancaster J. Motivation and synthesis of the FIAC experiment: The reproducibility of fMRI results across expert analyses. Hum Brain Mapping 27(5):351-359, 2006.
• Filipp, D. and M. Julius. Molecular Immunology 41:645-656, 2004. LR: Resolution of “Fyn problem” .
• Lukic AS, Wernick MN, Yang Y, Hansen LK, Arfanakis K, Strother SC. Effect of spatial alignment transformations in PCA and ICA of functional neuroimages. IEEE Trans Med Imaging 26(8):1058-68, 2007.
• Lukic AS, Wernick MN, Tzikas DG, Chen X, Likas A, Galatsanos NP, Yang Y, Zhao F, Strother SC. Kernel Methods for Analysis of Functional Neuroimages. IEEE Trans Med Img. 26(12):1613-24, 2007.
• Zhang J, Liang L, Anderson JR, Gatewood L, Rottenberg DA, Strother SC. A java-based fMRI processing pipeline evaluation system for assessment of univariate general linear model and multivariate canonical variate analysis-based pipelines. Neuroinformatics 6:123–134, 2008
• Zhang J, Anderson JR, Liang L, Pulapura SK, Gatewood L, Rottenberg DA, Strother SC. Evaluation and Optimization of fMRI Single-Subject Processing Pipelines with NPAIRS and 2nd-level CVA. Magnetic Resonance in Medicine (in press).
• Carter CS, Heckers S, Nichols T, Pine DS, Strother SC. Optimizing the Design and Analysis of Clinical FMRI Research Studies. Biol Psych. (in press).
• Schmah T, Hinton G, Zemel RS, Small SL, Strother SC. Competing RBM density models for classification of fMRI images. Proc. Neural Information Processing Systems (in press)