In recent years, magnetic resonance imaging has proven to be an important imaging modality for diagnosing and locating pathology. Recent studies have shown that multispectral tissue classification (MTC) may segment pathology from healthy tissues. Several studies have been done to classify brain tissues such as white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF) using MTC with slice thicknesses ranging from 5 to 10 mm (Kohn, 1991; Fletcher, 1993; Kao, 1994). In one of the previous studies (Fletcher, 1993) MTC has been used to classify brain tissues such as WM, GM, CSF, adipose tissue (AD), muscle (MS) and skin and meninges (S&M) with a slice thickness of 5 mm. The chosen slice thickness in the above mentioned studies is not quantified. Therefore a question remains as to what is the optimum slice thickness for MTC of brain tissues. The purpose of this research is to evaluate the ability of MTC to segment the brain tissues as a function of slice thickness using spectral regions such as spin-lattice relaxation time (TO, spin-spin relaxation time (T2), and spin density (p). The slice thicknesses used in the study were 3 mm, 5 mm, and 10 mm. Raw spin-echo images were acquired from a 39 year old volunteer at the level of lateral ventricles through the brain on a General Electric (Milwaukee, WI) 1.5 Tesla Signa imager with quadrature bird cage head coil. Ti, and p images were calculated from a set of seven raw spin-echo images using non-linear least square procedure (Gong, 1992) with varying repetition time (TR) and a constant echo time (TE). Similarly T2 images were calculated from a set of eight raw images using linear least square fit algorithm (Li, 1993) with varying TE and constant TR images. The Ti, T2, and p images were calculated for 3, 5, and 10 mm slice thicknesses. The ability to segment tissues WM, GM, CSF, AD, MS, and S&M as a function of slice thickness, was analyzed using optimization parameters such as false positive ratio (FPR), false negative ratio (FNR), true positive ratio (TPR), unclassified pixel ratio (UPR) and signal-to-noise ratio (S/N). The effect of partial voluming and spatial resolution on tissue classification was also evaluated. The optimum slice thickness for six brain tissue classification was determined.
Library of Congress Subject Headings
Brain--Magnetic resonance imaging; Brain--Diseases--Diagnosis; Diagnostic imaging; Mathematical optimization
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Narendranath, Nagesh, "Optimization of slice thickness in multispectral MRI tissue classification of brain tissues" (1996). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus