Modeling low‐frequency fluctuation and hemodynamic response timecourse in event‐related fMRI

Abstract

Functional magnetic resonance imaging (fMRI) suffers from many problems that make signal estimation difficult. These include variation in the hemodynamic response across voxels and low signal-to-noise ratio (SNR). We evaluate several analysis techniques that address these problems for event-related fMRI. (1) Many fMRI analyses assume a canonical hemodynamic response function, but this assumption may lead to inaccurate data models. By adopting the finite impulse response model, we show that voxel-specific hemodynamic response functions can be estimated directly from the data. (2) There is a large amount of low-frequency noise fluctuation (LFF) in blood oxygenation level dependent (BOLD) time-series data. To compensate for this problem, we use polynomials as regressors for LFF. We show that this technique substantially improves SNR and is more accurate than high-pass filtering of the data. (3) Model overfitting is a problem for the finite impulse response model because of the low SNR of the BOLD response. To reduce overfitting, we estimate a hemodynamic response timecourse for each voxel and incorporate the constraint of time-event separability, the constraint that hemodynamic responses across event types are identical up to a scale factor. We show that this technique substantially improves the accuracy of hemodynamic response estimates and can be computed efficiently. For the analysis techniques we present, we evaluate improvement in modeling accuracy via 10-fold cross-validation.

A Java-based fMRI Processing Pipeline Evaluation System for Assessment of Univariate General Linear Model and Multivariate Canonical Variate Analysis-based Pipelines

“…Further, the performance rank of fMRI processing pipelines changes according to different preprocessing conditions which indicates that the influence of the preprocessing steps (e.g., spatial smoothing) on pipeline performance can be significant and should not be ignored, as has been demonstrated by others (e.g., Tegeler et al 1999;Tanabe et al 2002;LaConte et al 2003;Smith et al 2005;Kay et al 2008). Therefore, the evaluation results of this study may not be readily generalized to other preprocessing settings.…”

Compressive spatial summation in human visual cortex

“…To ensure accurate GLM fits, we used a flexible model for the HRF at each voxel (Kay et al 2008a). Control points were placed every 2.5 s from 0 to 20 s after trial onset and every 10 s from 20 to 50 s after trial onset.…”

Optimizing preprocessing and analysis pipelines for single‐subject fMRI. I. Standard temporal motion and physiological noise correction methods

“…the preprocessing pipeline), or that the default settings in established software packages give near-optimal results. In recent years, though, it has been repeatedly shown that both standard and non-standard data preprocessing choices may have significant effects (negative or positive) on the quality of extracted results (e.g., [Della-Maggiore et al, 2002;Kay et al, 2007;Morgan et al, 2007;Murphy et al, 2009;Poline et al, 2006;Sarty, 2007;Strother et al, 2004;Tanabe et al, 2002;Zhang et al, 2009]. To draw reliable conclusions from fMRI results, it is thus necessary to evaluate the interactions of preprocessing and data analysis choices in a rigorous, systematic manner.…”

“…Additionally, Jones et al [2008] have shown that the effectiveness of PNC depends on its order in the preprocessing pipeline. Low-order temporal detrending has also been found to significantly affect results, depending on choice of detrending basis [Kay et al, 2007;Tanabe et al, 2002] and interactions with other parameters, such as spatial smoothing [Shaw et al, 2003]. A number of studies have also investigated basis decomposition techniques using Principal Component Analysis (PCA) and Independent Component Analysis (ICA) for data denoising [e.g.…”