Iterative reconstruction as a novel method of radiation dose reduction at computed tomography in patients with Crohn's disease

Prior work of our research group, that quantified the alarming levels of radiation dose to patients with Crohn’s disease from medical imaging and the notable shift towards CT imaging making these patients an at risk group, provided context for this work. CT delivers some of the highest doses of ionising radiation in diagnostic radiology. Once a medical imaging examination is deemed justified, there is an onus on the imaging team to endeavour to produce diagnostic quality CT images at the lowest possible radiation dose to that patient. The fundamental limitation with conventional CT raw data reconstruction was the inherent coupling of administered radiation dose with observed image noise – the lower the radiation dose, the noisier the image. The renaissance, rediscovery and refinement of iterative reconstruction removes this limitation allowing either an improvement in image quality without increasing radiation dose or maintenance of image quality at a lower radiation dose compared with traditional image reconstruction. This thesis is fundamentally an exercise in optimisation in clinical CT practice with the objectives of assessment of iterative reconstruction as a method for improvement of image quality in CT, exploration of the associated potential for radiation dose reduction, and development of a new split dose CT protocol with the aim of achieving and validating diagnostic quality submillisiever t CT imaging in patients with Crohn’s disease. In this study, we investigated the interplay of user-selected parameters on radiation dose and image quality in phantoms and cadavers, comparing traditional filtered back projection (FBP) with iterative reconstruction algorithms. This resulted in the development of an optimised, refined and appropriate split dose protocol for CT of the abdomen and pelvis in clinical patients with Crohn’s disease allowing contemporaneous acquisition of both modified and conventional dose CT studies. This novel algorithm was then applied to 50 patients with a suspected acute complication of known Crohn’s disease and the raw data reconstructed with FBP, adaptive statistical iterative reconstruction (ASiR) and model based iterative reconstruction (MBIR). Conventional dose CT images with FBP reconstruction were used as the reference standard with which the modified dose CT images were compared in terms of radiation dose, diagnostic findings and image quality indices. As there are multiple possible user-selected strengths of ASiR available, these were compared in terms of image quality to determine the optimal strength for this modified dose CT protocol. Modified dose CT images with MBIR were also compared with contemporaneous abdominal radiograph, where performed, in terms of diagnostic yield and radiation dose. Finally, attenuation measurements in organs, tissues, etc. with each reconstruction algorithm were compared to assess for preservation of tissue characterisation capabilities. In the phantom and cadaveric models, both forms of iterative reconstruction examined (ASiR and MBIR) were superior to FBP across a wide variety of imaging protocols, with MBIR superior to ASiR in all areas other than reconstruction speed. We established that ASiR appears to work to a target percentage noise reduction whilst MBIR works to a target residual level of absolute noise in the image. Modified dose CT images reconstructed with both ASiR and MBIR were non-inferior to conventional dose CT with FBP in terms of diagnostic findings, despite reduced subjective and objective indices of image quality. Mean dose reductions of 72.9-73.5% were achieved with the modified dose protocol with a mean effective dose of 1.26mSv. MBIR was again demonstrated superior to ASiR in terms of image quality. The overall optimal ASiR strength for the modified dose protocol used in this work is ASiR 80%, as this provides the most favourable balance of peak subjective image quality indices with less objective image noise than the corresponding conventional dose CT images reconstructed with FBP. Despite guidelines to the contrary, abdominal radiographs are still often used in the initial imaging of patients with a suspected complication of Crohn’s disease. We confirmed the superiority of modified dose CT with MBIR over abdominal radiographs at comparable doses in detection of Crohn’s disease and non-Crohn’s disease related findings. Finally, we demonstrated (in phantoms, cadavers and in vivo) that attenuation values do not change significantly across reconstruction algorithms meaning preserved tissue characterisation capabilities with iterative reconstruction. Both adaptive statistical and model based iterative reconstruction algorithms represent feasible methods of facilitating acquisition diagnostic quality CT images of the abdomen and pelvis in patients with Crohn’s disease at markedly reduced radiation doses. Our modified dose CT protocol allows dose savings of up to 73.5% compared with conventional dose CT, meaning submillisievert imaging is possible in many of these patients.

What to do when K-means clustering fails:a simple yet principled alternative algorithm

The K-means algorithm is one of the most popular clustering algorithms in current use as it is relatively fast yet simple to understand and deploy in practice. Nevertheless, its use entails certain restrictive assumptions about the data, the negative consequences of which are not always immediately apparent, as we demonstrate. While more flexible algorithms have been developed, their widespread use has been hindered by their computational and technical complexity. Motivated by these considerations, we present a flexible alternative to K-means that relaxes most of the assumptions, whilst remaining almost as fast and simple. This novel algorithm which we call MAP-DP (maximum a-posteriori Dirichlet process mixtures), is statistically rigorous as it is based on nonparametric Bayesian Dirichlet process mixture modeling. This approach allows us to overcome most of the limitations imposed by K-means. The number of clusters K is estimated from the data instead of being fixed a-priori as in K-means. In addition, while K-means is restricted to continuous data, the MAP-DP framework can be applied to many kinds of data, for example, binary, count or ordinal data. Also, it can efficiently separate outliers from the data. This additional flexibility does not incur a significant computational overhead compared to K-means with MAP-DP convergence typically achieved in the order of seconds for many practical problems. Finally, in contrast to K-means, since the algorithm is based on an underlying statistical model, the MAP-DP framework can deal with missing data and enables model testing such as cross validation in a principled way. We demonstrate the simplicity and effectiveness of this algorithm on the health informatics problem of clinical sub-typing in a cluster of diseases known as parkinsonism.

The development of an algorithm to track blood pressure accurately based on the pulse transit time method

Oscillometric blood pressure (BP) monitors are currently used to diagnose hypertension both in home and clinical settings. These monitors take BP measurements once every 15 minutes over a 24 hour period and provide a reliable and accurate system that is minimally invasive. Although intermittent cuff measurements have proven to be a good indicator of BP, a continuous BP monitor is highly desirable for the diagnosis of hypertension and other cardiac diseases. However, no such devices currently exist. A novel algorithm has been developed based on the Pulse Transit Time (PTT) method, which would allow non-invasive and continuous BP measurement. PTT is defined as the time it takes the BP wave to propagate from the heart to a specified point on the body. After an initial BP measurement, PTT algorithms can track BP over short periods of time, known as calibration intervals. After this time has elapsed, a new BP measurement is required to recalibrate the algorithm. Using the PhysioNet database as a basis, the new algorithm was developed and tested using 15 patients, each tested 3 times over a period of 30 minutes. The predicted BP of the algorithm was compared to the arterial BP of each patient. It has been established that this new algorithm is capable of tracking BP over 12 minutes without the need for recalibration, using the BHS standard, a 100% improvement over what has been previously identified. The algorithm was incorporated into a new system based on its requirements and was tested using three volunteers. The results mirrored those previously observed, providing accurate BP measurements when a 12 minute calibration interval was used. This new system provides a significant improvement to the existing method allowing BP to be monitored continuously and non-invasively, on a beat-to-beat basis over 24 hours, adding major clinical and diagnostic value.

An algorithm for finding the minimum cost of storing and regenerating datasets in multiple clouds

The proliferation of cloud computing allows users to flexibly store, re-compute or transfer large generated datasets with multiple cloud service providers. However, due to the pay-As-you-go model, the total cost of using cloud services depends on the consumption of storage, computation and bandwidth resources which are three key factors for the cost of IaaS-based cloud resources. In order to reduce the total cost for data, given cloud service providers with different pricing models on their resources, users can flexibly choose a cloud service to store a generated dataset, or delete it and choose a cloud service to regenerate it whenever reused. However, finding the minimum cost is a complicated yet unsolved problem. In this paper, we propose a novel algorithm that can calculate the minimum cost for storing and regenerating datasets in clouds, i.e. whether datasets should be stored or deleted, and furthermore where to store or to regenerate whenever they are reused. This minimum cost also achieves the best trade-off among computation, storage and bandwidth costs in multiple clouds. Comprehensive analysis and rigid theorems guarantee the theoretical soundness of the paper, and general (random) simulations conducted with popular cloud service providers' pricing models demonstrate the excellent performance of our approach.

Gateways placement in backbone wireless mesh networks

This paper presents a novel algorithm for the gateway placement problem in Backbone Wireless Mesh Networks (BWMNs). Different from existing algorithms, the new algorithm incrementally identifies gateways and assigns mesh routers to identified gateways. The new algorithm can guarantee to find a feasible gateway placement satisfying Quality-of-Service (QoS) constraints, including delay constraint, relay load constraint and gateway capacity constraint. Experimental results show that its performance is as good as that of the best of existing algorithms for the gateway placement problem. But, the new algorithm can be used for BWMNs that do not form one connected component, and it is easy to implement and use.

Poster Abstract : A sensor network for compression and streaming of GPS trajectory data

We present the design and deployment results for PosNet - a large-scale, long-duration sensor network that gathers summary position and status information from mobile nodes. The mobile nodes have a fixed-sized memory buffer to which position data is added at a constant rate, and from which data is downloaded at a non-constant rate. We have developed a novel algorithm that performs online summarization of position data within the buffer, where the algorithm naturally accommodates data input and output rate mismatch, and also provides a delay-tolerant approach to data transport. The algorithm has been extensively tested in a large-scale long-duration cattle monitoring and control application.

Activity modelling in crowded environments : a soft decision approach

Probabilistic topic models have recently been used for activity analysis in video processing, due to their strong capacity to model both local activities and interactions in crowded scenes. In those applications, a video sequence is divided into a collection of uniform non-overlaping video clips, and the high dimensional continuous inputs are quantized into a bag of discrete visual words. The hard division of video clips, and hard assignment of visual words leads to problems when an activity is split over multiple clips, or the most appropriate visual word for quantization is unclear. In this paper, we propose a novel algorithm, which makes use of a soft histogram technique to compensate for the loss of information in the quantization process; and a soft cut technique in the temporal domain to overcome problems caused by separating an activity into two video clips. In the detection process, we also apply a soft decision strategy to detect unusual events.We show that the proposed soft decision approach outperforms its hard decision counterpart in both local and global activity modelling.

Towards condition-invariant sequence-based route recognition

In this paper we present a novel algorithm for localization during navigation that performs matching over local image sequences. Instead of calculating the single location most likely to correspond to a current visual scene, the approach finds candidate matching locations within every section (subroute) of all learned routes. Through this approach, we reduce the demands upon the image processing front-end, requiring it to only be able to correctly pick the best matching image from within a short local image sequence, rather than globally. We applied this algorithm to a challenging downhill mountainbiking visual dataset where there was significant perceptual or environment change between repeated traverses of the environment, and compared performance to applying the feature-based algorithm FAB-MAP. The results demonstrate the potential for localization using visual sequences, even when there are no visual features that can be reliably detected.

Feature-based visual odometry and featureless place recognition for SLAM in 2.5D environments

In this paper we present a novel algorithm for localization during navigation that performs matching over local image sequences. Instead of calculating the single location most likely to correspond to a current visual scene, the approach finds candidate matching locations within every section (subroute) of all learned routes. Through this approach, we reduce the demands upon the image processing front-end, requiring it to only be able to correctly pick the best matching image from within a short local image sequence, rather than globally. We applied this algorithm to a challenging downhill mountain biking visual dataset where there was significant perceptual or environment change between repeated traverses of the environment, and compared performance to applying the feature-based algorithm FAB-MAP. The results demonstrate the potential for localization using visual sequences, even when there are no visual features that can be reliably detected.

Mathematical modelling of tumour-induced angiogenesis

The growth of solid tumours beyond a critical size is dependent upon angiogenesis, the formation of new blood vessels from an existing vasculature. Tumours may remain dormant at microscopic sizes for some years before switching to a mode in which growth of a supportive vasculature is initiated. The new blood vessels supply nutrients, oxygen, and access to routes by which tumour cells may travel to other sites within the host (metastasize). In recent decades an abundance of biological research has focused on tumour-induced angiogenesis in the hope that treatments targeted at the vasculature may result in a stabilisation or regression of the disease: a tantalizing prospect. The complex and fascinating process of angiogenesis has also attracted the interest of researchers in the field of mathematical biology, a discipline that is, for mathematics, relatively new. The challenge in mathematical biology is to produce a model that captures the essential elements and critical dependencies of a biological system. Such a model may ultimately be used as a predictive tool. In this thesis we examine a number of aspects of tumour-induced angiogenesis, focusing on growth of the neovasculature external to the tumour. Firstly we present a one-dimensional continuum model of tumour-induced angiogenesis in which elements of the immune system or other tumour-cytotoxins are delivered via the newly formed vessels. This model, based on observations from experiments by Judah Folkman et al., is able to show regression of the tumour for some parameter regimes. The modelling highlights a number of interesting aspects of the process that may be characterised further in the laboratory. The next model we present examines the initiation positions of blood vessel sprouts on an existing vessel, in a two-dimensional domain. This model hypothesises that a simple feedback inhibition mechanism may be used to describe the spacing of these sprouts with the inhibitor being produced by breakdown of the existing vessel's basement membrane. Finally, we have developed a stochastic model of blood vessel growth and anastomosis in three dimensions. The model has been implemented in C++, includes an openGL interface, and uses a novel algorithm for calculating proximity of the line segments representing a growing vessel. This choice of programming language and graphics interface allows for near-simultaneous calculation and visualisation of blood vessel networks using a contemporary personal computer. In addition the visualised results may be transformed interactively, and drop-down menus facilitate changes in the parameter values. Visualisation of results is of vital importance in the communication of mathematical information to a wide audience, and we aim to incorporate this philosophy in the thesis. As biological research further uncovers the intriguing processes involved in tumourinduced angiogenesis, we conclude with a comment from mathematical biologist Jim Murray, Mathematical biology is : : : the most exciting modern application of mathematics.

Distribution state estimation based on particle swarm and doubly loop mutant optimization (DLM-PSO)

This paper presents a novel algorithm based on particle swarm optimization (PSO) to estimate the states of electric distribution networks. In order to improve the performance, accuracy, convergence speed, and eliminate the stagnation effect of original PSO, a secondary PSO loop and mutation algorithm as well as stretching function is proposed. For accounting uncertainties of loads in distribution networks, pseudo-measurements is modeled as loads with the realistic errors. Simulation results on 6-bus radial and 34-bus IEEE test distribution networks show that the distribution state estimation based on proposed DLM-PSO presents lower estimation error and standard deviation in comparison with algorithms such as WLS, GA, HBMO, and original PSO.

Enhanced stochastic mobility prediction with multi-output Gaussian processes

Outdoor robots such as planetary rovers must be able to navigate safely and reliably in order to successfully perform missions in remote or hostile environments. Mobility prediction is critical to achieving this goal due to the inherent control uncertainty faced by robots traversing natural terrain. We propose a novel algorithm for stochastic mobility prediction based on multi-output Gaussian process regression. Our algorithm considers the correlation between heading and distance uncertainty and provides a predictive model that can easily be exploited by motion planning algorithms. We evaluate our method experimentally and report results from over 30 trials in a Mars-analogue environment that demonstrate the effectiveness of our method and illustrate the importance of mobility prediction in navigating challenging terrain.

An efficient and robust system for multi-person event detection in real world indoor surveillance scenes

Due to the popularity of security cameras in public places, it is of interest to design an intelligent system that can efficiently detect events automatically. This paper proposes a novel algorithm for multi-person event detection. To ensure greater than real-time performance, features are extracted directly from compressed MPEG video. A novel histogram-based feature descriptor that captures the angles between extracted particle trajectories is proposed, which allows us to capture motion patterns of multi-person events in the video. To alleviate the need for fine-grained annotation, we propose the use of Labelled Latent Dirichlet Allocation, a “weakly supervised” method that allows the use of coarse temporal annotations which are much simpler to obtain. This novel system is able to run at approximately ten times real-time, while preserving state-of-theart detection performance for multi-person events on a 100-hour real-world surveillance dataset (TRECVid SED).

Transdimensional sequential Monte Carlo using variational Bayes — SMCVB

A new transdimensional Sequential Monte Carlo (SMC) algorithm called SM- CVB is proposed. In an SMC approach, a weighted sample of particles is generated from a sequence of probability distributions which ‘converge’ to the target distribution of interest, in this case a Bayesian posterior distri- bution. The approach is based on the use of variational Bayes to propose new particles at each iteration of the SMCVB algorithm in order to target the posterior more efficiently. The variational-Bayes-generated proposals are not limited to a fixed dimension. This means that the weighted particle sets that arise can have varying dimensions thereby allowing us the option to also estimate an appropriate dimension for the model. This novel algorithm is outlined within the context of finite mixture model estimation. This pro- vides a less computationally demanding alternative to using reversible jump Markov chain Monte Carlo kernels within an SMC approach. We illustrate these ideas in a simulated data analysis and in applications.

Recommendation of Multimedia Objects Using Metadata and Link Analysis

We investigate methods for recommending multimedia items suitable for an online multimedia sharing community and introduce a novel algorithm called UserRank for ranking multimedia items based on link analysis. We also take the initiative of applying EigenRumor from the domain of blogosphere to multimedia. Furthermore, we present a strategy for making personalized recommendation that combines UserRank with collaborative filtering. We evaluate our method with an informal user study and show that results obtained are promising.