Hyperspectral Data Processing: Algorithm Design and Analysis

32

Multispectral Magnetic Resonance Imaging

Due to recent advances in instrument technology, applications of remote sensing are no more confined to geoscience and earth science but have rather expanded to other areas such as medical diagnosis, food safety and inspection, law enforcement, defense, homeland security, and so on. Chapter 31 explores an expansion of hyperspectral imaging (HSI) techniques to multispectral imaging (MSI). This chapter deals with another new expansion of HSI to magnetic resonance (MR) imaging. Specifically, the problem of partial volume estimation (PVE) will be studied here. In the past years, two general approaches, a finite Gaussian mixture (FGM) model-based statistical approach and a fuzzy c-means (FCM)-based structural approach have been implemented in conjunction with Markov random field (MRF) to investigate PVE problems. This chapter develops a third spectral approach, which is completely different from the two above-mentioned approaches. It is a new PVE approach, which is based on linear spectral mixture analysis (LSMA) discussed in Part III of this book and Chapter 31. It assumes that an MR image voxel is linearly mixed with tissues of different types via a linear mixing model from which it can be further unmixed into abundance fractions of these tissues in terms of their partial volumes. To further effectively explore intravoxel spectral information within an MR image, two nonlinear expansions using nonlinear band dimensionality and nonlinear ...