Generalized fuzzy clustering for segmentation of multi-spectral magnetic resonance images.

An integrated approach for multi-spectral segmentation of MR images is presented. This method is based on the fuzzy c-means (FCM) and includes bias field correction and contextual constraints over spatial intensity distribution and accounts for the non-spherical cluster's shape in the feature space. The bias field is modeled as a linear combination of smooth polynomial basis functions for fast computation in the clustering iterations. Regularization terms for the neighborhood continuity of intensity are added into the FCM cost functions. To reduce the computational complexity, the contextual regularizations are separated from the clustering iterations. Since the feature space is not isotropic, distance measure adopted in Gustafson-Kessel (G-K) algorithm is used instead of the Euclidean distance, to account for the non-spherical shape of the clusters in the feature space. These algorithms are quantitatively evaluated on MR brain images using the similarity measures.

Improved cerebellar tissue classification on magnetic resonance images of brain.

PURPOSE: To develop and implement a method for improved cerebellar tissue classification on the MRI of brain by automatically isolating the cerebellum prior to segmentation. MATERIALS AND METHODS: Dual fast spin echo (FSE) and fluid attenuation inversion recovery (FLAIR) images were acquired on 18 normal volunteers on a 3 T Philips scanner. The cerebellum was isolated from the rest of the brain using a symmetric inverse consistent nonlinear registration of individual brain with the parcellated template. The cerebellum was then separated by masking the anatomical image with individual FLAIR images. Tissues in both the cerebellum and rest of the brain were separately classified using hidden Markov random field (HMRF), a parametric method, and then combined to obtain tissue classification of the whole brain. The proposed method for tissue classification on real MR brain images was evaluated subjectively by two experts. The segmentation results on Brainweb images with varying noise and intensity nonuniformity levels were quantitatively compared with the ground truth by computing the Dice similarity indices. RESULTS: The proposed method significantly improved the cerebellar tissue classification on all normal volunteers included in this study without compromising the classification in remaining part of the brain. The average similarity indices for gray matter (GM) and white matter (WM) in the cerebellum are 89.81 (+/-2.34) and 93.04 (+/-2.41), demonstrating excellent performance of the proposed methodology. CONCLUSION: The proposed method significantly improved tissue classification in the cerebellum. The GM was overestimated when segmentation was performed on the whole brain as a single object.

{\it In vivo\/} magnetic resonance studies of experimental liver disease: Carbon tetrachloride hepatotoxicity and alcohol-induced fatty liver in rat

Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS) were used to non-invasively determine if cirrhosis induced by carbon tetrachloride (CCl$\sb4$) and phospholipase-D (PLD) could be distinguished from fatty infiltration in rat. MRS localization and water suppression methods were developed, implemented and evaluated in terms of their application to in vivo proton NMR studies of experimental liver disease. MRS studies were also performed to quantitate fatty infiltration resulting from carbon tetrachloride (CCl$\sb4$) or alcohol (ethanol) administration and the MRS results were confirmed using biochemical total lipid analysis and histology. $\rm T\sb1$ weighted MR images acquired weekly, 48 hours post administration, demonstrated only a slight increase in overall liver intensity with CCl$\sb4$ or alcohol administration, which is consistent with previously reported results. The MR images were able to detect nodules resulting from CCl$\sb4$+PLD induced cirrhosis as hypointense regions, also consistent with previous reports. Localized in vivo water and lipid proton $\rm T\sb1$ relaxation time measurements were performed and demonstrated no statistically significant trends for either agent. In vivo proton spectra were also acquired using stimulated echo techniques to quantitatively follow the changes in liver lipid content. The changes in liver lipid content observed using MRS were verified by total lipid analysis using the Folch technique and histology. The in vivo $\rm T\sb1$ and lipid quantification data str inconsistent with the previous hypothesis that the changes in $\rm T\sb1$ weighted images were the result of increased "free" water content and, therefore, increased water $\rm T\sb1$ relaxation times. These data indicate that the long term changes are more likely the result of changes in lipid content. The data are also shown to agree with the accepted hypothesis that the time course and mechanism of fatty infiltration are different for CCl$\sb4$ and alcohol. The hypothesis that the lipids resulting from either protocol are from the same lipid fraction(s), presumably triglycerides, is also supported. And lastly, on the basis of MR images and quantitative MRS lipid information, it was shown that cirrhosis could be distinguished from fatty infiltration. ^

Automatic quantitation of multiple sclerosis lesions on MR images

A two-pronged approach for the automatic quantitation of multiple sclerosis (MS) lesions on magnetic resonance (MR) images has been developed. This method includes the design and use of a pulse sequence for improved lesion-to-tissue contrast (LTC) and seeks to identify and minimize the sources of false lesion classifications in segmented images. The new pulse sequence, referred to as AFFIRMATIVE (Attenuation of Fluid by Fast Inversion Recovery with MAgnetization Transfer Imaging with Variable Echoes), improves the LTC, relative to spin-echo images, by combining Fluid-Attenuated Inversion Recovery (FLAIR) and Magnetization Transfer Contrast (MTC). In addition to acquiring fast FLAIR/MTC images, the AFFIRMATIVE sequence simultaneously acquires fast spin-echo (FSE) images for spatial registration of images, which is necessary for accurate lesion quantitation. Flow has been found to be a primary source of false lesion classifications. Therefore, an imaging protocol and reconstruction methods are developed to generate "flow images" which depict both coherent (vascular) and incoherent (CSF) flow. An automatic technique is designed for the removal of extra-meningeal tissues, since these are known to be sources of false lesion classifications. A retrospective, three-dimensional (3D) registration algorithm is implemented to correct for patient movement which may have occurred between AFFIRMATIVE and flow imaging scans. Following application of these pre-processing steps, images are segmented into white matter, gray matter, cerebrospinal fluid, and MS lesions based on AFFIRMATIVE and flow images using an automatic algorithm. All algorithms are seamlessly integrated into a single MR image analysis software package. Lesion quantitation has been performed on images from 15 patient volunteers. The total processing time is less than two hours per patient on a SPARCstation 20. The automated nature of this approach should provide an objective means of monitoring the progression, stabilization, and/or regression of MS lesions in large-scale, multi-center clinical trials. ^