Showing papers by "Simon R. Arridge published in 2001"

Time resolved optical tomography of the human forearm

Abstract: A 32-channel time-resolved optical imaging instrument has been developed principally to study functional parameters of the new-born infant brain. As a prelude to studies on infants, the device and image reconstruction methodology have been evaluated on the adult human forearm. Cross-sectional images were generated using time-resolved measurements of transmitted light at two wavelengths. All data were acquired using a fully automated computer-controlled protocol. Images representing the internal scattering and absorbing properties of the arm are presented, as well as images that reveal physiological changes during a simple finger flexion exercise. The results presented in this paper represent the first simultaneous tomographic reconstruction of the internal scattering and absorbing properties of a clinical subject using purely temporal data, with additional co-registered difference images showing repeatable absorption changes at two wavelengths in response to exercise.

Three-dimensional time-resolved optical tomography of a conical breast phantom

Abstract: A 32-channel time-resolved imaging device for medical optical tomography has been employed to evaluate a scheme for imaging the human female breast. The fully automated instrument and the reconstruction procedure have been tested on a conical phantom with tissue-equivalent optical properties. The imaging protocol has been designed to obviate compression of the breast and the need for coupling fluids. Images are generated from experimental data with an iterative reconstruction algorithm that employs a three-dimensional (3D) finite-element diffusion-based forward model. Embedded regions with twice the background optical properties are revealed in separate 3D absorption and scattering images of the phantom. The implications for 3D time-resolved optical tomography of the breast are discussed.

Simultaneous imaging and optode calibration with diffuse optical tomography

Abstract: In order for diffuse optical tomography to realize its potential of obtaining quantitative images of spatially varying optical properties within random media, several potential experimental systematic errors must be overcome One of these errors is the calibration of the emitter strength and detector efficiency/gain While in principle these parameters can be determined accurately prior to an imaging experiment, slight fluctuations will cause significant image artifacts For this reason, it is necessary to consider including their calibration as part of the inverse problem for image reconstruction In this paper, we show that this can be done successfully in a linear reconstruction model with simulated continuous-wave data The technique is general for frequency and time domain data

A Regularization Scheme for Diffusion Tensor Magnetic Resonance Images

Abstract: A method for regularizing diffusion tensor magnetic resonance images (DT-MRI) is presented. The scheme is divided into two main parts: a restoration of the principal diffusion direction, and a regularization of the 3 eigenvalue maps. The former make use of recent variational methods for restoring direction maps, while the latter makes use of the strong structural information embedded in the diffusion tensor image to drive a non-linear anisotropic diffusion process. The whole process is illustrated on synthetic and real data, and possible improvements are discussed.

Analysis of MTR histograms in multiple sclerosis using principal components and multiple discriminant analysis.

Abstract: Magnetization transfer ratio (MTR) histograms have the potential to characterize subtle diffuse changes in multiple sclerosis (MS) and other white matter disease. A new method is described which gives improved correlation with the Expanded Disability Status Scale (EDSS). Classification of individual subjects into normal and MS subgroups is shown. Principal component analysis (PCA) and multiple discriminant analysis (MDA) are shown to give results superior to methods of MTR histogram analysis using traditional features such as peak height and peak location. Scatterplots confirm the improved separation between groups achieved using the MDA score. The histogram analysis provides a comparison of two classification approaches, based on PCA and MDA, to recognize differences between normal controls and the four different subgroups of MS disease (and all MS patients). Multiple linear regression of these PCs vs. EDSS established an MR-based measure of disease. Using a central 60-mm slab of brain tissue, the success rate of binary classification between control and MS subgroups using MDA was 75-95%, depending on which two groups were being compared. Multiple regression analysis of EDSS with the first three PCs as independent variables was significant (r = 0.83 for secondary progressive MS, and r = 0.80 for all MS patients).

Effect of roughness in nondiffusive regions within diffusive media

Abstract: Recently it has been shown that clear regions within diffusive media can be accurately modeled within the diffusion approximation by means of a novel boundary condition [J. Opt. Soc. Am. A17, 1671 (2000)] or by an approximation to it [Phys. Med. Biol.41, 767 (1996); Med. Phys.27, 252 (2000)]. This can be directly applied to the study of light propagation in brain tissue, in which there exist clear regions, and in particular in the cerebrospinal fluid (CSF) layer under the skull. In this work we present the effect that roughness in the boundary of nondiffusive regions has on the measured average intensity, since, in practice, the CSF layer is quite rough. The same conclusions can be extended to any diffusive medium that encloses rough nondiffusive regions. We will demonstrate with numerical calculations that the roughness statistics of the interfaces (although not their actual profiles) must be known a priori to correctly predict the shape of the average intensity. We show that as the roughness increases, the effect of the nondiffusive region diminishes until it disappears, thus yielding data similar to those of a fully diffusive region. We also present a numerical study of the diffuse light scattered in the presence of both diffusive and nondiffusive regions and the interaction between the two, showing that when the nondiffusive region is rough, the spatial-intensity distribution produced by the two regions can be very similar.

Differential imaging in heterogeneous media: limitations of linearization assumptions in optical tomography

Abstract: Image reconstruction and data collection in optical tomography can be achieved in a number of different ways. This paper explores the limitations of using assumptions of linearity, particularly in the case where image data is acquired before and after a change in optical properties within an object with heterogeneous optical properties. The effects of using a 2 dimensional (2D) reconstruction scheme for changes in 3D measurements are also demonstrated. Problems are a direct result of the inherent non-linearity of optical tomographic image reconstruction. We show how these assumptions affect images of changes in absorption in the presence of a) heterogeneous background scatter, and b) heterogeneous background absorption using both simulations and time-resolved experimental data. Comparisons of results using non-linear and linear image reconstruction techniques are included throughout. The origin and dependence of the error are investigated. Methods to improve results by using estimates of background structure from baseline images are shown to improve quantitation and object localization in simple images. The potential significance of this error is discussed in relation to successful, reliable clinical imaging of the neonatal brain.

Methods for the Inverse Problem in Optical Tomography

Abstract: In this paper we give a brief overview of the image reconstruction problem in Optical Tomography together with some examples of simulation reconstructions using a numerical optimisation scheme. We discuss first a Diffraction Tomography approach, wherein it is assumed that the measurement is of a scattered wave representing the difference in fields between an unknown and a known state. Both the Born and Rytov approximations are presented and lead to a linear reconstruction problem. Secondly we discuss image reconstruction as an Optimization problem, wherein we develop a model capable of predicting the total field and minimize a least-squares error functional. We introduce the Finite Element Method as a tool for calculating photon density fields in general complex geometries. By means of this method we simulate a number of images and their reconstructions using both a Born and Rytov approximation. The Rytov approximation appears superior in all cases.

Radiosity diffusion model in 3D

Abstract: We present the Radiosity-Diffusion model in three dimensions(3D), as an extension to previous work in 2D. It is a method for handling non-scattering spaces in optically participating media. We present the extension of the model to 3D including an extension to the model to cope with increased complexity of the 3D domain. We show that in 3D more careful consideration must be given to the issues of meshing and visibility to model the transport of light within reasonable computational bounds. We demonstrate the model to be comparable to Monte-Carlo simulations for selected geometries, and show preliminary results of comparisons to measured time-resolved data acquired on resin phantoms.© (2001) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Fast 3D image reconstruction in optical tomography using a coarse-gain parallelization strategy

Abstract: Absolute dual-parameter image reconstruction in optical tomography (OT) is a nonlinear and ill-posed problem, requiring a model-based iterative approach and an accurate and fully three-dimensional light transport forward model These factors make OT a computationally expensive problem, resulting in reconstruction times that range from several minutes to hours, which is not acceptable in most clinical applications In order to reduce reconstruction times we propose a coarse-grain parallel implementation of the inverse algorithm using a shared-memory threaded approach Unlike most conventional parallelization strategies which operate on the level of the linear algebra subsystem our approach uses a parallelization at the application level, thereby leaving the underlying linear matrix solution routines untouched This allows to exploit the inherently parallel structure of the problem, while at the same time utilizing fast direct or efficiently preconditioned iterative serial linear solvers, which in most cases can not be efficiently parallelized We compare the performance of the coarse-grain parallel approach with a low-level parallelization of the linear conjugate gradient solver and show that the proposed method achieves significant performance improvements, thereby bringing reconstruction times within a clinically acceptable range

Quantification of Spinal Cord Atrophy via an Active Surface Model.

Abstract: Introduction Spinal cord atrophy is of interest for multiple sclerosis (MS) studies because it is sensitive to disability in MS patients and can be detected even in early disease. We aim to provide measurements of atrophy along the cord by extracting the cord surface in a continuous parameterised form, computing its medial axis, and defining crosssections orthogonal to this medial axis. Measurements of those crosssections area along the nerve are then available to locate and quantify potential atrophy. Methods Inversion recovery gradient echo images were acquired in the sagittal plane with 256x256x60, 0.98x0.98x1mm3. Images are pre-processed using a non-linear edge-preserving diffusion filter, to reduce noise and enhance edges [1]. A B-Spline active surface is then embedded in the image. The model is parametric and defined as presented in [2]. A mesh of control points define the entire surface. An energy is associated with the surface, whose minimum corresponds to the desired location [2,3] (the boundaries of the spinal cord). This energy is a sum of a gradient term that attracts the surface to structure boundaries, and an internal term that guarantees the surface smoothness. The surface is iteratively optimised by sequential optimisation of the set of control points, namely the greedy algorithm [4]. Because this optimisation scheme converges to local minima of the energy function, it is important to provide a good initialisation of the surface. Our model is interactively initialised as follows: through a graphic interface, the user choses a set of landmarks in the center of axial sections along the cervical cord. The first landmark is set at the foramen magnum, the last landmark at the base of the 7th cervical vertebra, and two landmarks per vertebra are set. Each landmark is the center of a set of control points located on a circular section of fixed diameter (11mm). This defines the initial surface as a pseudo-cylinder of constant radius that roughly approximates the cord shape. After optimisation, a parameterisation S(r,s) of the cord surface is obtained, with r an axial parameter and s a longitudinal parameter, both in the interval [0;1]. A cord medial axis A is then defined by the following:

Optical tomography of a 3D multilayered head model

Abstract: The head is inherently a heterogenous 3 dimensional (3D) domain and any useful image reconstruction algorithm in optical tomography should be based on the true shape and dimensions of the domain to be imaged. However due to the computational complexity of the problem, most reconstruction algorithms have been based on either 2 dimensional (2D) domains or simple 3D homogenous models. In this work we use a complex multi-layered model of the head to generate data in the presence of known changes of absorption within the brain. In order to evaluate the reconstruction of images from 3D data, we use either a 2D mesh based on the outline of a single measurement plane, or a 3D mesh based on the forward model itself. We show that the use of a 2D model for the reconstruction of images from a complex heterogenous 3D model is not successful for absolute imaging of internal optical property. In 3D, image reconstruction using meantime data only from a 3D domain is not successful when reconstructing for absorption only while keeping background scatter constant and homogenous. Image reconstruction in 3D is greatly improved when a-priori structural knowledge is used. Reconstruction of internal absorption only, from difference data (data after a change in absorption minus data before a change) works both in 2D and 3D, however in 2D, any depth information available from the data is lost.