Temporal-spatial modeling for fMRI data Public Deposited
Downloadable ContentDownload PDF
- Last Modified
- March 21, 2019
- Affiliation: College of Arts and Sciences, Department of Statistics and Operations Research
- By generating high quality movies of the brain in action, functional Magnetic Resonance Imaging (fMRI) helps us examine which parts of the human brains are activated by different task performances. Many techniques for fMRI analysis have been developed in the last decade. Independent component analysis (ICA) is an effective data-driven method to explore spatio-temporal features in fMRI data. It has been especially success-ful to recover brain-function-related signals from recorded mixtures of unrelated signals. Due to the high sensitivity of MR scanners, spikes are commonly observed in fMRI data sets and they deteriorate the analysis. No particular method exists yet to address this problem. In the first part of this work, we introduce a supervised singular value decomposition technique into the data reduction step of ICA. The proposed method improves the robustness of ICA against spikes and makes the computation more e±cient by using the particular fMRI experiment designs to guide the fully data-driven ICA. The advantages are demonstrated using a simulation study as well as a real data analysis. ICA aims to separate blind source signals from their linear mixture signals based on the assumptions of the statistical independence and non-Gaussian distributions of the source signals. The second part of this work studies the methodology of some most popular ICA algorithms and propose to evaluate some of the algorithms by assessing the variability of the estimates of the mixing matrix through a nonparametric bootstrap procedure. Two maximum likelihood ICA algorithms are studied in detail through a simulation study. Another popular category of statistical techniques for fMRI analysis consists of model-driven strategies. Among them, the most widely used approach is statistical parametric mapping (SPM), where the key technique is general linear model (GLM) and the temporal characteristic of the expected response is usually modeled by the convolution of the experiment stimulus and a predefined hemodynamic response function (HRF). However, the subjective assumptions of the form of HRF introduce estimation biases and subsequently reduce the detection power of activation. In the third part of this work, we propose a new nonparametric method to model the time component adaptively in the context of SPM. The idea is to start from an initial time component obtained from general SPM procedure and then apply a penalized smoothing technique to update the shape of the hemodynamic response in an adaptive way. The nice performance of the proposed method is illustrated through a simulation study as well as a real fMRI data analysis. Event-related fMRI (ER-fMRI) has played an important role in many recent brain imaging studies to explore the relationship between recorded fMRI signals and neural activity. Different from traditional block-design fMRI, ER-fMRI is very good at estimating the timing and waveform of the hemodynamic response. Various methods have been proposed in the literature to model the HRF. However, most of them have a number of limitations. In the last part of this work, we propose a novel regression approach to estimate the HRF directly. The approach is based on point processes modeling to account for the event-related designs. Compared to the existing methods, the proposed procedure yields simultaneously the nonparametric estimate of the HRF and a test for the linearity assumption. To illustrate its usefulness and the scientific implications, we applied this procedure to study the spatial variation of the HRF, and the extent to which the linear relationship holds in various regions of interest for Parkinson's disease patients.
- Date of publication
- December 2007
- Resource type
- Rights statement
- In Copyright
- Truong, Young
- Smith, Richard L.
- Degree granting institution
- University of North Carolina at Chapel Hill
- Open access
This work has no parents.
|Temporal-spatial modeling for fMRI data||2019-04-09||Public||