Extracting activated regions of brain with FMRI data using a robust unsupervised learning approach

Functional Magnetic Resonance Imaging (fMRI) is a robust noninvasive and modern technique for imaging of brain functions. However, the acquired data from fMRI experiences cannot be interpreted readily directly due to several factors. These factors are the signal weakness, abounding noises in the fMRI data, and the challenge of separating significant activations of interest from other kinds. The complexity of the raw fMRI data leads to significant challenges faced with multi operations with data, such as image conversion, read/write, and extract information. To overcome these difficulties and challenges, a suggested workflow for fMRI data analysis is proposed in this thesis. It includes a complete analysis of fMRI data, starting from DICOM (Digital Imaging and Communications in Medicine) conversion, then checking the quality of data at each step, pipeline steps of preprocessing, and ending in proposing the clustering method as a postprocessing analysis of fMRI data. Consequently, in the first part of this thesis, a novel conversion and visualization fMRI (VCfMRI) toolbox package is designed and developed to address many of the problems when visualizing multi-format fMRI data. The presented design is the first significant contribution of the thesis. In this thesis, the VCfMRI tool is extended and updated to include a comprehensive analysis of fMRI data. The developed tool is called CPREPP fMRI, in which the preprocessing pipeline steps are added in addition to statistical analysis as well. The main goal of this thesis is to identify the activation regions of the human brain. Therefore, a novel unsupervised learning approach is proposed that relies on using Enhanced Neural Gas (ENG) algorithm in fMRI data for comparison with Neural Gas (NG) method, which has yet to be utilized for that aim. It is considered a significant contribution of this thesis. Four validation indices are applied to evaluate the performance of the proposed ENG method with fMRI and compare it with a clustering approach (NG algorithm) and model-based data analysis using statistical parametric mapping (SPM). The comparison outcomes on real auditory fMRI data show that ENG outperforms the NG and SPM methods due to its insensitivity to the ordering of input data sequence, various initializations for selecting a set of neurons, and the existence of extreme values (outliers). The ENG technique can tackle all shortcomings of NG application with fMRI data, identify the active area of the human brain effectively, and determine the locations of the cluster center based on the MDL value during the process of network learning.