Neurological information retrieval in brain imaging based on nonlinear system identification

The development of biomedical sciences and engineering has now evolved to a modern era of system level quantitative biomedicine. The biomedical imaging techniques, such as computer tomography, magnetic resonance imaging and positron emission tomography, have been widely used to obtain physiological information of patients. Among these imaging modalities, PET imaging is capable of retrieving functional information from image sequences because tracer kinetic modeling can be used to aid the understanding of images. In this paper we investigate the brain neurotransmitter imaging with nonlinear compartmental model. We propose a fast and robust method to quantify such model from PET data using Volterra expansion of nonlinear systems. A reliable and robust quantitative biomarker based on Volterra kernels is introduced to identify the receptor concentrations. The method has been tested on simulation studies and the results show that the receptor concentration can be accurately and robustly identified that has potential applications in clinical practices.