Neurofeedback by real time functional MRI (rt-fMRI) has potential for addiction research and treatment that will be realized only if the feedback given the subject is related meaningfully to the cognitive states that must be controlled. The mental operations of the brain are too distributed to be represented by the raw rt-fMRI signal in any one brain region or small group of regions.
Our aims are to: 1) Use computational machine learning to rapidly detect patterned activation in the rt-fMRI signal that better expresses cognitive state;2) augment these data with concurrently-collected electroencephalographic (EEG) data;3) develop an atlas of brain data that identifies brain patterns with cognitive states relevant to addiction and drug abuse research and 4) to explore rt-fMRI neurofeedback using this rt-fMRI/EEG machine learning method. Our approach will be to first create rapid algorithms for pattern matching that are fast compared with the imaging, thereby allowing "real-time" application. To do so we will select features from the images that express the differences among state concisely (more technically, we will use a method known as independent components analysis to reduce the data dimensionality.) We will similarly condense the EEG features by studying them by the location of their sources within the brain, and by examining the frequencies that they contain. We will run experiments on volunteers designed to help us see their tendency to make impulsive choices - which is known to relate to their likelihood to become drug users, as well as experiments that track changes in their brain as they control their craving urges. For these studies we will look at heavy cigarette users. Cigarette use on its own is a serious health burden to the nation, and it is also an excellent model for addiction more generally, as it is known to have many neural features in common with use of other drugs of abuse, such as cocaine and methamphetamine. This is a phased innovation proposal. The first phase will be focused on the developments of the rt-fMRI analysis and instrumentation technology. On its successful completion, based on specific milestones, we will move to the more applied work with human subjects.
Our research aims to develop and characterize a means of rapidly detecting brain states relevant to addiction research through the use of magnetic resonance imaging and electroencephalography. We are interested specifically in states and markers of impulsivity and cigarette craving. Our goal ultimately is to have a tool that can be used in the context of neurofeedback, allowing human subject or patient to receive an indication of activity in their brains associated with these states and to enable them to learn to control these cognitive/affective states by controlling the brain activity.
|Kerr, Wesley T; Hwang, Eric S; Raman, Kaavya R et al. (2014) Multimodal diagnosis of epilepsy using conditional dependence and multiple imputation. Int Workshop Pattern Recognit Neuroimaging :1-4|
|Anderson, Ariana; Douglas, Pamela K; Kerr, Wesley T et al. (2014) Non-negative matrix factorization of multimodal MRI, fMRI and phenotypic data reveals differential changes in default mode subnetworks in ADHD. Neuroimage 102 Pt 1:207-19|
|Kerr, Wesley T; Douglas, Pamela K; Anderson, Ariana et al. (2014) The utility of data-driven feature selection: re: Chu et al. 2012. Neuroimage 84:1107-10|
|Lenartowicz, Agatha; Simpson, Gregory V; Haber, Catherine M et al. (2014) Neurophysiological signals of ignoring and attending are separable and related to performance during sustained intersensory attention. J Cogn Neurosci 26:2055-69|
|Anderson, Ariana; Bramen, Jennifer; Douglas, Pamela K et al. (2011) Large Sample Group Independent Component Analysis of Functional Magnetic Resonance Imaging Using Anatomical Atlas-Based Reduction and Bootstrapped Clustering. Int J Imaging Syst Technol 21:223-231|
|Culbertson, Christopher S; Bramen, Jennifer; Cohen, Mark S et al. (2011) Effect of bupropion treatment on brain activation induced by cigarette-related cues in smokers. Arch Gen Psychiatry 68:505-15|
|Douglas, P K; Harris, Sam; Yuille, Alan et al. (2011) Performance comparison of machine learning algorithms and number of independent components used in fMRI decoding of belief vs. disbelief. Neuroimage 56:544-53|
|Anderson, Ariana; Labus, Jennifer S; Vianna, Eduardo P et al. (2011) Common component classification: what can we learn from machine learning? Neuroimage 56:517-24|