ANALYSIS AND PROCESSING STRATEGIES FOR FUNCTIONAL MAGNETIC RESONANCE IMAGING DATA

Abstract

Functional imaging refers to the process of visualizing physiological activities associated with anatomical structures. Because this information cannot be inferred directly from looking at anatomical features in the image, many diagnostic and interventional procedures have been greatly. enhanced. Brain comes in the forefront of organs that need further understanding of relation 'among structure, function, and pathology. Functional brain imaging is currently performed using a variety of imaging modalities including Nuclear Medicine, Computed Tomography (CT), Light Absorbance, and Magnetic. Resonance Imaging (MRI). All these methods are based on direct or indirect detection of neural response to a mental or a physical activity.

Among these techniques, functional MRI (fMRl) is preferable due to its high spatial 'and temporal resolutions. The concept behind fMRl is to localize brain activity by detecting the corresponding change in average cerebral blood oxygenation. This change occurs upon neural firing, which, in tum, occurs during stimulation. Brain tissue T2*, which represents transverse relaxation time, is sensitive to cerebral blood oxygenation changes. Hence, Blood Oxygenation Level-Dependant (BOLD) effect acts just like a T2 * contrast agent. To observe these hemodynamic changes, rapid acquisition of a series of brain MR images is needed. The ultimate result is a map relating definite regions in the brain with specific functions, hence the name fMRl.

One of the main drawbacks of fMRl is the degradation in SNR due to the transient nature of the response. The maximum achievable signal change in activated brain regions at l.ST is only 5%. Hence, there are two issues of fundamental importance in the analysis of MRI time series: the modeling of the hemodynamic response (HDR) and the estimation of the underlying signal noise. HDR refers to the local changes in blood oxygenation as an effect of increased neuronal activity. Hence, it provides the input-output relationship between stimulus and activation response. Regarding to artifacts in fMRI data, there are two types, namely physiologic and random noise.