Ian R. Greenshields

A Nonlocal Maximum Likelihood Estimation Method for Rician Noise Reduction in MR Images

Postacquisition denoising of magnetic resonance (MR) images is of importance for clinical diagnosis and computerized analysis, such as tissue classification and segmentation. It has been shown that the noise in MR magnitude images follows a Rician distribution, which is signal-dependent when signal-to-noise ratio (SNR) is low. It is particularly difficult to remove the random fluctuations and bias introduced by Rician noise. The objective of this paper is to estimate the noise free signal from MR magnitude images. We model images as random fields and assume that pixels which have similar neighborhoods come from the same distribution. We propose a nonlocal maximum likelihood (NLML) estimation method for Rician noise reduction. Our method yields an optimal estimation result that is more accurate in recovering the true signal from Rician noise than NL means algorithm in the sense of SNR, contrast, and method error. We demonstrate that NLML performs better than the conventional local maximum likelihood (LML) estimation method in preserving and defining sharp tissue boundaries in terms of a well-defined sharpness metric while also having superior performance in method error.

An MRF spatial fuzzy clustering method for fMRI SPMs

Abstract The paper presents a method for spatial fuzzy clustering (SFC) via Markov Random Fields (MRF) for the detection of brain activation regions in Functional Magnetic Resonance Imaging (fMRI) statistical parametric maps (SPMs) to improve the accuracy of the detection of such regions. The fMRI SPM is assumed to be an MRF and we define a fuzzy neighborhood energy function to describe the interaction between neighboring voxels. The final labeling is determined by a joint fuzzy membership. We compare the proposed spatial fuzzy clustering technique with the usual voxel-wise thresholding, traditional fuzzy clustering and Contextual Clustering (CC) [E. Salli, H.J. Aronen, S. Savolainen, A. Korvenoja, A. Visa, Contextual clustering for analysis of functional MRI data, IEEE Transactions on Medical Imaging 20 (2001) 403–414]. Experiments based on synthetic and real fMRI data demonstrate that the clustering performance of our method is significantly better than both simple thresholding and conventional non-spatial fuzzy clustering techniques. Our experiments also show that in relatively high quality SPMs (contrast to noise ratio ( CNR ) > 2.5 ), the performance of SFC and CC is very similar. In the case of the simulated datasets, when the SPMs have poor quality ( CNR 2.5 ), our method outperforms CC in reducing false positives and improving classification accuracy.

Classification Algorithm for Multi-Echo Magnetic Resonance Image Using Gibbs Distributions

This paper describes a new three dimensional image (volumetric image) classification technique which is based on the Markov Random Field (MRF)-Gibbs Random Field (GRF) model together with a stochastic relaxation algorithm. For the classification of Multi-Echo (multispectral) Magnetic Resonance Images (MRI), a Bayesian context decision rule is adopted and an MRF-GRF stochastic model is introduced for the original image. To obtain the maximum a posterior probability (MAP) classification a new multivariate image context-dependent classification based on relaxation and annealing is developed. Conventially, a digital image is considered as a two-dimensional random field defined over rectangular lattice structure and the domain of image classification is the plane. However, in the volumetric image classification, we use volumetric images, i.e., three dimensional image data sets.

A fast wavelet-based Karhunen–Loeve transform

Abstract The paper describes the role of the standard wavelet decomposition in computing a fast Karhunen–Loeve transform. The standard wavelet decomposition (which we show is different from the conventional wavelet transform) leads to a highly sparse and simply structured transformed version of the correlation matrix which can be easily subsetted (with little loss of Frobenius norm). The eigenstructure of this smaller matrix can be efficiently computed using standard algorithms such as QL. Finally, we provide an example of the use of the efficient transform by classifying a 219-channel AVIRIS image with respect to its eigensystem.

Architecture of a distributed imaging system

We describe the architecture of a distributed medical imaging system. Designed to support both code and data mobility, the system attempts to exploit available computational resources within a collaborating group to apportion imaging tasks to the platforms best suited to their completion. The system is designed around a mixed Java-RMI (remote method invocation), CORBA and agent architecture. This paper surveys various features of the system.

Coherent computation of the multispectral maximal directional derivative

Abstract We describe an approach to the computation of the planar or volumetric maximal directional derivative (gradient) of a multispectral or hyperspectral image. We show that the planar multispectral case has an immediate solution. For the non-planar (volumetric or multitemporal) case we demonstrate that an iterative optimization technique (downhill simplex) exploiting image coherency is faster than the conventional eigendecomposition. Finally, we show that the iterative technique, based on matrix norms, has extensions not meaningful in the eigendecomposition method.

De-noising ENMR spectra by wavelet shrinkage

Wavelet shrinkage de-noising is applied to electrophoretic nuclear magnetic resonance (ENMR) data. Both threshold rules for removing noise, namely soft and hard, proposed in Donohos (1993) VisuShrink, are used simultaneously. Soft thresholding is applied to fine levels of wavelet decomposition coefficients and hard thresholding is applied to coarse levels. This implementation is coordinated by visualizing the features presented in ENMR spectra.

Unsupervised classification of multiecho magnetic resonance images of the pediatric brain with implicit spatial and statistical hypotheses validation

We describe an image segmentation method applied to multi-echo MR images which is unsupervised in that the analyst need not specify prototypical tissue signatures to guide the segmentation. It is well known that different tissue types may be distinguished by their signatures in NMR parameter space (spin density and relaxation parameters T1 and T2). Also, normal tissue may be differentiated from abnormal by means of these signatures. Even though pixel intensity is proportional to weighted mixtures of these parameters in real images several researchers feel there is potential for better segmentation results by processing dual-echo images. These images are inherently registered and require no additional time to acquire the image for the second echo. Our segmentation procedure is a multi-step process in which tissue class mean vectors and covariance matrices are first determined by a clustering technique. The goal here is to achieve an intermediate segmentation which may be subject to quantitative validation.© (1993) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

3D shape approximants via spherical wavelet decompositions [clinical pelvimetry]

We describe a simple application of a spherical wavelet to the problem of the shape representation of the bony pelvis. We show a simple shape approximant computed over the sphere S/sup 2/ and demonstrate the use of the Narcowich-Ward (1996) spherical wavelet decomposition of this shape approximant.

Distributed visualization of volumetric medical datasets

Increasing spatial and spectral image resolution, coupled with increased expectations both in visualization output and in the time-responsiveness of that output relative to certain clinical tasks have led to a comcomitant increase in the role of parallel and distributed solutions to the visualization problem. In this paper we examine the distribution of classification and visualization tasks over a heterogenous distributed architecture in the context of one specific visualization problem viz. the directed visualization of small-scale structures from the Visible Human Dataset. Specifically, we examine a Challenger-like architecture in which tasks are dynamically bid onto and transported across the fabric under the control of agents.