Discrimination of single photon emission computed tomography brain scans using Fourier series image model

A model for brain SPECT imaging studies from subsample discrete Fourier series image models that can be used for further statistical discrimination is presented. Slice image data is represented as columns modeled by Fourier series. The approach used in this thesis uses the intensity values of the image voxel. Prediction error minimization method is analyzed versus the ordinary least square method applied to modeling brain SPECT image slices. Different column profiles were tested and the PEM model was shown to fit very well the original clinical data.;A common practice in medical diagnosis and patient management is the comparison of two acquisition data. Two statistical discrimination parametric tests were applied: equality of means and equality of covariance matrices.;The data of this study shows promising results in modeling brain SPECT images using prediction error minimization method for estimating model coefficients.