1 of 7 Analysis of Functional MRI Timeseries Using Statistical Parametric Mapping Mohamed A. Mohamed 1 , Fatma Abou-Chadi 1 , and Bassem K.Ouda 2 Abstract- This paper presents a general approach to the analysis of functional MRI timeseries for one or more subjects. The approach seeks a compromise between efficiency, generality, validity, simplicity and execution speed. The main differences between this technique and previous ones are: (i) a simple bias reduction and regularization of voxel wise autoregressive model parameters, (ii) the combination of effects and their estimated standard deviations across different runs/sessions/subjects, and (iii) overcoming the problem of small number of runs/sessions/subjects. The proposed method was applied to real fMRI database acquired at different fMRI acquisition parameters. The ability of the system in detecting neuronal activation was tested using standard statistical tests implemented in statistical parametric mapping (SPM). Keywords- Functional Magnetic Resonance Imaging (fMRI), Blood- Level-Oxygenation-Dependent (BOLD), Hemodynamic Response Function (HRF), and Statistical Parametric Mapping (SPM). I. INTRODUCTION Over the past few years, there have been numerous reports on the statistical analysis of fMRI timeseries [1]. The traditional approach for analyzing fMRI data utilizes the general linear model (GLM) using a hypothesized neural model convolved with a canonical hemodynamic response function (HRF). Mismatches of the data to the specified hemodynamic model can be induced by small hemodynamic delays or slice-timing differences. The use of a hemodynamic model and its temporal derivative for fMRI analysis was proposed in [2] as a parsimonious model with additional flexibility to address delay- induced modeling mismatches. In this paper, a complete analysis of fMRI timeseries using SPM2 was introduced, and compared to the results obtained using SPM99 in [3]. The comparison was based on the ability of activation detection and the statistical F-test. The remainder of this paper is organized as: Section-II presents an overview of fMRI, SPM and data collection; Section-III introduces the concepts of temporal basis functions, temporal filtering and models of fMRI temporal autocorrelation; 1 M.A. Mohamed and Fatma Abou-Chadi are with the dept. of electronics and comm., faculty of engineering, Mansoura University, Egypt. E-mail: mazim12@yahoo.com and F-abochadi@ieee.org 2 Bassem K.Ouda is with the department of biomedical engineering and systems, faculty of engineering, Cairo University, Egypt. E-mail: bkouda@yahoo.com. Section-IV illustrates the application of these ideas to event- related models; Section-V concerns the efficiency of fMRI experimental designs, as a function of the inter-stimulus interval and ordering of stimulus; Section-VI gives the results and Section-VII is the conclusion. II. OVERVIEW A. Functional MRI Functional MRI is the most recently developed modality, which distinguishes itself from earlier methods (e.g. PET, SPECT) in which no exposure to any ionizing radiation is evolved, better spatial and temporal resolution is achieved, and a relatively straight forward coregistration to anatomical MRI acquired on the same machine can be attained. The main problem associated with fMRI is the poor SNR where the intensity of the detected MRI signal is mainly dependent on the applied static field. In general, fMRI timecourses can be modeled as the summation of the activation signal, physiologic and random noise components [4]. B. Statistical Parametric Mapping (SPM) SPM is a statistical technique for examining differences in brain activity recorded during functional neuroimaging experiments. In general, one analyzes data in SPM in three steps: (i) generating a design matrix made up of column vectors, which predict physiological responses to changing task conditions. Hence, the design matrix defines the experimental design and the nature of hypothesis testing; (ii) estimating parameters for these predictor vectors using a least squares approach to linear regression and (iii) using these parameter estimates for statistical inferences about single subject or group level hypotheses. SPM2 is the major update to the SPM software [5], containing substantial theoretical, algorithmic, structural and interface enhancements over previous versions. C. Data Collection The used database was available through the web site of the department of imaging neuroscience at the Institute of Neurology at University College London [6]. All subjects are scanned using a 1.5T Siemens Sonata MRI imaging machine. The database was provided in a single session event-related