Power Spectral Density Analysis of Time Series of Pixel of Functional Magnetic Resonance Image for Different Motor Activity
Nivedita Daimiwal and Revati ShriramCummins College of Engineering for Women, Pune, India.
Corresponding Author E-mail: nivedita.daimiwal@gmail.com
Abstract: Functional Magnetic Resonance Imaging (fMRI) is a non invasive modality to detect structure and function of the brain. Brain functions for various activities like motor, sensory, speech and memory process are detected using fMRI modality. This paper deals with the analysis of power spectrum of pixel time series for different motor activities. The analysis is to relate the power magnitude of the spike in the power spectrum of the fMRI time series with the activity performed. The fMRI data set consists of a sequence of images with respect to time, when the subject performs a definite task in a given block paradigm. The data set consists of four slices each of size 64×64 pixels. The power spectrum is acquired by taking the Fourier transform of the time series. The shape of the power spectrum is often referred to as 1/f or the inverse frequency function. Low frequency noise is removed by applying discrete cosine transform on time series. Data was originally, collected from General Electric Signa 1.5 T MRI system for 5 male subjects; 3 subjects: Performed lower limb movement (LL) and 2 subjects: Performed upper limb movement (UL). The power magnitude of the spike is recorded for lower limb and upper limb movement. The spike in the power spectrum at f Hz corresponds to the frequency at which the task is performed. The power magnitude amplitude for lower limb activity is around 14.31 dB and upper limb is around 4.0 dB. Power spectral density (PSD) of the time series is used for the detection of activities occurring in the brain.
Keywords: Discrete Cosine Transform; fMRI; Power Spectrum; Time Series Back to TOC