We propose a nonlocal diffusion equation (NDE) as a new forward model, which uses the concepts of differential operators under the nonlocal vector calculus. The discretization of the NDE is performed using an effective graph-based numerical method (GNM). We evaluate the proposed forward modelling method on a homogeneous slab where the analytical solution is available. Our experiments show that the results of the NDE (discretized by GNM) is quantitatively comparable to the analytical solution. The proposed method has an identical implementation for geometries in two and three dimensions due to the nature of the graph representation.
We consider L1-regularization of spectrally constrained DOT. Three popular algorithms are investigated: iteratively reweighted least square algorithm (IRLS), alternating directional method of multipliers (ADMM) and fast iterative shrinkage-thresholding algorithm (FISTA). We evaluate different regularizers and algorithms on a 3D simulated multi-spectral example.
Bioluminescence imaging (BLI) is a widely used pre-clinical imaging technique, but there are a number of limitations to its quantitative accuracy. This work uses an animal model to demonstrate some significant limitations of BLI and presents processing methods and algorithms which overcome these limitations, increasing the quantitative accuracy of the technique. The position of the imaging subject and source depth are both shown to affect the measured luminescence intensity. Free Space Modelling is used to eliminate the systematic error due to the camera/subject geometry, removing the dependence of luminescence intensity on animal position. Bioluminescence tomography (BLT) is then used to provide additional information about the depth and intensity of the source. A substantial limitation in the number of sources identified using BLI is also presented. It is shown that when a given source is at a significant depth, it can appear as multiple sources when imaged using BLI, while the use of BLT recovers the true number of sources present.
Plenoptic (light-field) imaging is a technique that allows a simple CCD-based imaging device to acquire both spatially and angularly resolved information about the “light-field” from a scene. It requires a microlens array to be placed between the objective lens and the sensor of the imaging device1 and the images under each microlens (which typically span many pixels) can be computationally post-processed to shift perspective, digital refocus, extend the depth of field, manipulate the aperture synthetically and generate a depth map from a single image. Some of these capabilities are rigid functions that do not depend upon the scene and work by manipulating and combining a well-defined set of pixels in the raw image. However, depth mapping requires specific features in the scene to be identified and registered between consecutive microimages. This process requires that the image has sufficient features for the registration, and in the absence of such features the algorithms become less reliable and incorrect depths are generated. The aim of this study is to investigate the generation of depth-maps from light-field images of scenes with insufficient features for accurate registration, using projected patterns to impose a texture on the scene that provides sufficient landmarks for the registration methods.
A plenoptic imaging system records simultaneously the intensity and the direction of the rays of light. This additional information allows many post processing features such as 3D imaging, synthetic refocusing and potentially evaluation of wavefront aberrations. In this paper the effects of low order aberrations on a simple plenoptic imaging system have been investigated using a wave optics simulations approach.
Decreasing the acquisition time in bioluminescence imaging (BLI) and bioluminescence tomography (BLT) will enable animals to be imaged within the window of stable emission of the bioluminescent source, a higher imaging throughput and minimisation of the time which an animal is anaesthetised. This work investigates, through simulation using a heterogeneous mouse model, two methods of decreasing acquisition time: 1. Imaging at fewer wavelengths (a reduction from five to three); and 2. Increasing the bandwidth of filters used for imaging. The results indicate that both methods are viable ways of decreasing the acquisition time without a loss in quantitative accuracy. Importantly, when choosing imaging wavelengths, the spectral attenuation of tissue and emission spectrum of the source must be considered, in order to choose wavelengths at which a high signal can be achieved. Additionally, when increasing the bandwidth of the filters used for imaging, the bandwidth must be accounted for in the reconstruction algorithm.
The structure of an artificial ligament was examined using Raman microscopy
in combination with novel data analysis. Basis approximation and compressed principal component
analysis are shown to provide efficient compression of confocal Raman microscopy
images, alongside powerful methods for unsupervised analysis. This scheme allows the acceleration
of data mining, such as principal component analysis, as they can be performed on
the compressed data representation, providing a decrease in the factorisation time of a single
image from five minutes to under a second.
Using this workflow the interface region between a chemically engineered ligament construct
and a bone-mimic anchor was examined. Natural ligament contains a striated interface
between the bone and tissue that provides improved mechanical load tolerance, a similar
interface was found in the ligament construct.
Practical imaging constraints restrict the number of wavelengths that can be measured in a single Biolumines- cence Tomography imaging session, but it is unclear which set of measurement wavelengths is optimal, in the sense of providing the most information about the bioluminescent source. Mutual Information was used to integrate knowledge of the type of bioluminescent source likely to be present, the optical properties of tissue and physics of light propagation, and the noise characteristics of the imaging system, in order to quantify the information contained in measurements at different sets of wavelengths. The approach was applied to a two-dimensional sim- ulation of Bioluminescence Tomography imaging of a mouse, and the results indicate that different wavelengths and sets of wavelengths contain different amounts of information. When imaging at a single wavelength, 580nm was found to be optimal, and when imaging at two wavelengths, 570nm and 580nm were found to be optimal. Examination of the dispersion of the posterior distributions for single wavelengths suggests that information regarding the location of the centre of the bioluminescence distribution is relatively independent of wavelength, whilst information regarding the width of the bioluminescence distribution is relatively wavelength specific.
A study is presented that demonstrates that bioluminescence tomography can reconstruct accurate 3D images of internal light sources placed at a range of depths within a physical phantom and that it provides more reliable quantitative data than standard bioluminescence imaging. Specifically, it is shown that when imaging sources at depths ranging from 5 to 15mm, estimates of total source strength are stable to within ±11% using tomography whilst values deduced by traditional methods vary 10-fold. Additionally, the tomographic approach correctly localises sources to within 1.5mm error in all cases considered.
We show how a random matrix can be used to reduce the dimensionality of the bioluminescence tomography reconstruction problem. A randomised low-rank approximation for the sensitivity matrix is computed, and we show how this can be used to reconstruct the bioluminescence source distribution on a randomised basis for the mesh nodes. The distribution on the original mesh can be found easily via a simple matrix multiplication. The majority of the computation required can be performed in advance of the reconstruction, and the reconstruction time itself is of the order milliseconds. This could allow for high frame rate real-time reconstructions to be performed.
There exist numerous methods that aim to extract the optical parameters of a tissue by relating reflectance
measurements to a theoretical model of light transport. During the parameter recovery process, assumptions
are often made about the characteristics of the tissue. However, specious assumptions lead to inaccurate or even
incorrect results. We present a method based on the maximum a posteriori estimation technique to recover the
concentrations of the main chromophores present in a biological tissue from reflectance or transmittance measurements.
The method provides correct results even in the presence of significant uncertainty in the underlying
properties of the tissue. A preliminary analysis of the results obtained from simulated skin reflectance spectra
suggests that the proposed MAP based method provides accurate estimates and is robust against a high level
of uncertainty in the tissue's model. The results of phantom data are in agreement with the findings from our
simulations as they emphasise the importance of including prior information about the unknown parameter in
the estimation process.
Steps are presented towards the development of a new bioluminescence tomography (BLT) imaging system for
in vivo small animal studies. A 2-mirror-based multi-view data collection scheme is investigated in conjunction
with multi-spectral imaging, leading to the production of 3D volumetric maps of molecular source distributions
in simulation and in physical phantom studies by way of a finite element model (FEM) based reconstruction
method. A proof of concept is subsequently demonstrated showing a full work flow from data acquisition to 3D
reconstruction. Results suggest that the multi-view mirror-based approach represents a strong improvement over
standard single-view methods, with improvements of up to 58% in source localisation accuracy being observed
for deep sources.
We present preliminary data from an imaging system based on LED illumination for obtaining sequential multispectral
optical images of the human ocular fundus. The system is capable of acquiring images at speeds of up
to 20fps and we have demonstrated that the system is fast enough to allow images to be acquired with minimal
inter-frame movement. Further improvements have been identified that will improve both imaging speed and
image quality. The long-term goal is to use the system in conjunction with novel image analysis algorithms to
extract chromophore concentrations from images of the ocular fundus, with a particular emphasis on age-related
macular degeneration. The system has also found utility in fluorescence microscopy.
A three dimensional (3D) photon transport model has been developed based on the frequency domain simplified
spherical harmonics approximation (SPN) to the Radiative Transport Equation. Based on preliminary Monte Carlo
studies, it is shown that for problems exhibiting strong absorption, the solutions using the 7th order SPN model (N = 7) are
significantly more accurate than those from a standard Diffusion (SP1) based solver. This advance is of particular
interest in the field of bioluminescent imaging where the peak emission of light emitting molecular markers are closer to
the visible range (500 - 650 nm) corresponding to strong absorption due to hemoglobin.
Multi-spectral imaging of the ocular fundus suffers from three main problems: the image must be taken through an aperture (the pupil), meaning that the absolute light intensity at the fundus cannot be known; long acquisition times are not feasible due to patient discomfort; patient movement can lead to loss of image quality. These difficulties have meant that multi-spectral imaging of the fundus has not yet seen wide application. We have developed a new method for optimizing the multi-spectral imaging process which also allows us to derive semi-quantitative information about the structure and properties of the fundus. We acquire images in six visible spectral bands and use these to deduce the concentration and distribution of the known absorbing compounds in the fundus: blood haemoglobins in the retina and choroid, choroidal melanin, RPE melanin and xanthophyll. The optimisation process and parameter recovery uses a Monte Carlo model of the spectral reflectance of the fundus, parameterised by the concentrations of the absorbing compounds. The model is used to compute the accuracy with which the values of the model parameters can be deduced from an image. Filters are selected to minimise the error in the parameter recovery process. Theoretical investigations suggest that parameters can be recovered with RMS errors of less than 10%. When applied to images of normal subjects, the technique was able to successfully deduce the distribution of xanthophyll in the fundus. Further improvement of the model is required to allow the deduction of other model parameters from images.