Characterization of Brain Noise using Multimodal Mutual Information

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): The objective of this project is to utilize concurrent multimodal recordings using functional magnetic resonance imaging (fMRI), near-infrared optical (NIRS) and measures of systemic physiology in order to separate the contributions of each of these types of noise. The novelty of this project will be the use of mutual information shared between these concurrent measurements in order to identify the sources of noise in the brain-imaging signal. This will allow us to better understand the nature of this noise, to better design filters and analysis to remove it, and to improve the detection of small changes in activity in the brain. We hypothesize that i) mutual information between concurrent optical and fMRI signals will generate more realistic models of the noise structure in these modalities and ii) that a better quantitative understanding of this noise will lead to the identification of limitations in current analysis procedures and to the optimization of improved statistical techniques. Aim 1. Identify the contributions of instrumental and systemic sources of noise to the fMRI BOLD (blood oxygen level dependent) signal using concurrent multimodal optical and fMRI recordings. Aim 2. Characterize the role of baseline physiology in systematic biases during inter- and intra-subject test-retest experiments using median-nerve stimulation with concurrent NIRS and fMRI. Aim 3. Investigate the role of subject attention in mediating single-trial variability in the fMRI BOLD signal. Aim 4. Compare approaches for reducing physiological noise in fMRI and NIRS and determine the effect of realistic noise on the assumptions of the standard analysis model. PUBLIC HEALTH RELEVANCE: Although functional MRI can produce high quality images of brain activity, these images require the averaging of data over many trial repetitions because of the high level of noise typically associated with fMRI signals. Preliminary data has suggested that much of this noise originates inside the brain and may be associated with time-dependent factors such as subject attention or vigilance in the task. The goal of this work is to improve the ability to detect small and infrequent events in the brain by better understanding the nature of this noise through concurrent multimodal methods and by optimizing data analysis methods to improve detection of functional events.