Bayesian computational methods for spatial analysis of images

This thesis introduces a new way of using prior information in a spatial model and develops scalable algorithms for fitting this model to large imaging datasets. These methods are employed for image-guided radiation therapy and satellite based classification of land use and water quality. This study has utilized a pre-computation step to achieve a hundredfold improvement in the elapsed runtime for model fitting. This makes it much more feasible to apply these models to real-world problems, and enables full Bayesian inference for images with a million or more pixels.

The rapid manufacture of uniform composite multicellular-biomaterial micropellets, their assembly into macroscopic organized tissues, and potential applications in cartilage tissue engineering

We and others have published on the rapid manufacture of micropellet tissues, typically formed from 100-500 cells each. The micropellet geometry enhances cellular biological properties, and in many cases the micropellets can subsequently be utilized as building blocks to assemble complex macrotissues. Generally, micropellets are formed from cells alone, however when replicating matrix-rich tissues such as cartilage it would be ideal if matrix or biomaterials supplements could be incorporated directly into the micropellet during the manufacturing process. Herein we describe a method to efficiently incorporate donor cartilage matrix into tissue engineered cartilage micropellets. We lyophilized bovine cartilage matrix, and then shattered it into microscopic pieces having average dimensions < 10 μm diameter; we termed this microscopic donor matrix "cartilage dust (CD)". Using a microwell platform, we show that ~0.83 μg CD can be rapidly and efficiently incorporated into single multicellular aggregates formed from 180 bone marrow mesenchymal stem/stromal cells (MSC) each. The microwell platform enabled the rapid manufacture of thousands of replica composite micropellets, with each micropellet having a material/CD core and a cellular surface. This micropellet organization enabled the rapid bulking up of the micropellet core matrix content, and left an adhesive cellular outer surface. This morphological organization enabled the ready assembly of the composite micropellets into macroscopic tissues. Generically, this is a versatile method that enables the rapid and uniform integration of biomaterials into multicellular micropellets that can then be used as tissue building blocks. In this study, the addition of CD resulted in an approximate 8-fold volume increase in the micropellets, with the donor matrix functioning to contribute to an increase in total cartilage matrix content. Composite micropellets were readily assembled into macroscopic cartilage tissues; the incorporation of CD enhanced tissue size and matrix content, but did not enhance chondrogenic gene expression.

Incorporating visual information for spoken term detection

Spoken term detection (STD) is the task of looking up a spoken term in a large volume of speech segments. In order to provide fast search, speech segments are first indexed into an intermediate representation using speech recognition engines which provide multiple hypotheses for each speech segment. Approximate matching techniques are usually applied at the search stage to compensate the poor performance of automatic speech recognition engines during indexing. Recently, using visual information in addition to audio information has been shown to improve phone recognition performance, particularly in noisy environments. In this paper, we will make use of visual information in the form of lip movements of the speaker in indexing stage and will investigate its effect on STD performance. Particularly, we will investigate if gains in phone recognition accuracy will carry through the approximate matching stage to provide similar gains in the final audio-visual STD system over a traditional audio only approach. We will also investigate the effect of using visual information on STD performance in different noise environments.

Connecting soundscape to landscape: Which acoustic index best describes landscape configuration?

Soundscape assessment has been proposed as a remote ecological monitoring tool for measuring biodiversity, but few studies have examined how soundscape patterns vary with landscape configuration and condition. The goal of our study was to examine a suite of published acoustic indices to determine whether they provide comparable results relative to varying levels of landscape fragmentation and ecological condition in nineteen forest sites in eastern Australia. Our comparison of six acoustic indices according to time of day revealed that two indices, the acoustic complexity and the bioacoustic index, presented a similar pattern that was linked to avian song intensity, but was not related to landscape and biodiversity attributes. The diversity indices, acoustic entropy and acoustic diversity, and the normalized difference soundscape index revealed high nighttime sound, as well as a dawn and dusk chorus. These indices appear to be sensitive to nocturnal biodiversity which is abundant at night in warm, subtropical environments. We argue that there is need to better understand temporal partitioning of the soundscape by specific taxonomic groups, and this should involve integrated research on amphibians, insects and birds during a 24 h cycle. The three indices that best connected the soundscape with landscape characteristics, ecological condition and bird species richness were acoustic entropy, acoustic evenness and the normalized difference soundscape index. This study has demonstrated that remote soundscape assessment can be implemented as an ecological monitoring tool in fragmented Australian forest landscapes. However, further investigation should be dedicated to refining and/or combining existing acoustic indices and also to determine if these indices are appropriate in other landscapes and for other survey purposes.

Fire design rules for load bearing cold-formed steel frame walls exposed to realistic design fire curves

Light gauge Steel Frame (LSF) walls are extensively used in the building industry due to the many advantages they provide over other wall systems. Although LSF walls have been used widely, fire design of LSF walls is based on approximate prescriptive methods based on limited fire tests. Also these fire tests were conducted using the standard fire curve [1] and the applicability of available design rules to realistic design fire curves has not been verified. This paper investigates the accuracy of existing fire design rules in the current cold-formed steel standards and the modifications proposed by previous researchers. Of these the recently developed design rules by Gunalan and Mahendran [2] based on Eurocode 3 Part 1.3 [3] and AS/NZS 4600 [4] for standard fire exposure [1] were investigated in detail to determine their applicability to predict the axial compression strengths and fire resistance ratings of LSF walls exposed to realistic design fire curves. This paper also presents the fire performance results of LSF walls exposed to a range of realistic fire curves obtained using a finite element analysis based parametric study. The results from the parametric study were used to develop a simplified design method based on the critical hot flange temperature to predict the fire resistance ratings of LSF walls exposed to realistic fire curves. Finally, the stud failure times (fire resistance rating) obtained from the fire design rules and the simplified design method were compared with parametric study results for LSF walls lined with single and double plasterboards, and externally insulated with rock fibres under realistic fire curves.

Generalised features for bird vocalisation retrieval in acoustic recordings

Bioacoustic monitoring has become a significant research topic for species diversity conservation. Due to the development of sensing techniques, acoustic sensors are widely deployed in the field to record animal sounds over a large spatial and temporal scale. With large volumes of collected audio data, it is essential to develop semi-automatic or automatic techniques to analyse the data. This can help ecologists make decisions on how to protect and promote the species diversity. This paper presents generic features to characterize a range of bird species for vocalisation retrieval. In the implementation, audio recordings are first converted to spectrograms using short-time Fourier transform, then a ridge detection method is applied to the spectrogram for detecting points of interest. Based on the detected points, a new region representation are explored for describing various bird vocalisations and a local descriptor including temporal entropy, frequency bin entropy and histogram of counts of four ridge directions is calculated for each sub-region. To speed up the retrieval process, indexing is carried out and the retrieved results are ranked according to similarity scores. The experiment results show that our proposed feature set can achieve 0.71 in term of retrieval success rate which outperforms spectral ridge features alone (0.55) and Mel frequency cepstral coefficients (0.36).

Bayesian parametric bootstrap for models with intractable likelihoods

In this paper it is demonstrated how the Bayesian parametric bootstrap can be adapted to models with intractable likelihoods. The approach is most appealing when the semi-automatic approximate Bayesian computation (ABC) summary statistics are selected. After a pilot run of ABC, the likelihood-free parametric bootstrap approach requires very few model simulations to produce an approximate posterior, which can be a useful approximation in its own right. An alternative is to use this approximation as a proposal distribution in ABC algorithms to make them more efficient. In this paper, the parametric bootstrap approximation is used to form the initial importance distribution for the sequential Monte Carlo and the ABC importance and rejection sampling algorithms. The new approach is illustrated through a simulation study of the univariate g-and- k quantile distribution, and is used to infer parameter values of a stochastic model describing expanding melanoma cell colonies.

Scalable Bayesian Computation for intractable likelihoods in image analysis

The inverse temperature hyperparameter of the hidden Potts model governs the strength of spatial cohesion and therefore has a substantial influence over the resulting model fit. The difficulty arises from the dependence of an intractable normalising constant on the value of the inverse temperature, thus there is no closed form solution for sampling from the distribution directly. We review three computational approaches for addressing this issue, namely pseudolikelihood, path sampling, and the approximate exchange algorithm. We compare the accuracy and scalability of these methods using a simulation study.

A thorough investigation of the interdependent behavior of prestressed system

Due to the existing of many prestressed members in the structural system, the interdependent behavior of all prestressed members is the main concern in the analysis of the pretension process. A thorough investigation of this mutual effect is essential for an effective, reliable, and optimal analysis. Focus on this aspect, this paper presents an investigation of the interdependent behavior of all prestressed members in the whole structural system based on influence matrix (IFM). Four different types of IFM are introduced. Two different solving methods are brought forth to analyze the pretension process. The direct solving method solves for the accurate solution, whereas the iterative solving method repeatedly amends to achieve an approximate solution. A numerical example is then conducted. The result shows that various kinds of complicated batched and repeated tensioning schemes can be analyzed reliably, effectively, and completely based on IFM.

Online hidden Markov model parameter estimation and minimax robust quickest change detection in uncertain stochastic processes

Stochastic (or random) processes are inherent to numerous fields of human endeavour including engineering, science, and business and finance. This thesis presents multiple novel methods for quickly detecting and estimating uncertainties in several important classes of stochastic processes. The significance of these novel methods is demonstrated by employing them to detect aircraft manoeuvres in video signals in the important application of autonomous mid-air collision avoidance.

The role of environmental factors in the spatial distribution of Japanese encephalitis in mainland China

Japanese encephalitis (JE) is the most common cause of viral encephalitis and an important public health concern in the Asia-Pacific region, particularly in China where 50% of global cases are notified. To explore the association between environmental factors and human JE cases and identify the high risk areas for JE transmission in China, we used annual notified data on JE cases at the center of administrative township and environmental variables with a pixel resolution of 1 km×1 km from 2005 to 2011 to construct models using ecological niche modeling (ENM) approaches based on maximum entropy. These models were then validated by overlaying reported human JE case localities from 2006 to 2012 onto each prediction map. ENMs had good discriminatory ability with the area under the curve (AUC) of the receiver operating curve (ROC) of 0.82-0.91, and low extrinsic omission rate of 5.44-7.42%. Resulting maps showed JE being presented extensively throughout southwestern and central China, with local spatial variations in probability influenced by minimum temperatures, human population density, mean temperatures, and elevation, with contribution of 17.94%-38.37%, 15.47%-21.82%, 3.86%-21.22%, and 12.05%-16.02%, respectively. Approximately 60% of JE cases occurred in predicted high risk areas, which covered less than 6% of areas in mainland China. Our findings will help inform optimal geographical allocation of the limited resources available for JE prevention and control in China, find hidden high-risk areas, and increase the effectiveness of public health interventions against JE transmission.

Multiple primary cancers associated with Merkel cell carcinoma in Queensland, Australia, 1982-2011

The relatively high incidence of Merkel cell carcinoma (MCC) in Queensland provides a valuable opportunity to examine links with other cancers. A retrospective cohort study was performed using data from the Queensland Cancer Registry. Standardized incidence ratios (SIRs) were used to approximate the relative risk of being diagnosed with another primary cancer either following or prior to MCC. Patients with an eligible first primary MCC (n=787) had more than double the expected number of subsequent primary cancers (SIR=2.19, 95% confidence interval (CI)=1.84–2.60; P<0.001). Conversely, people who were initially diagnosed with cancers other than MCC were about two and a half times more likely to have a subsequent primary MCC (n=244) compared with the general population (SIR=2.69, 95% CI=2.36–3.05; P<0.001). Significantly increased bi-directional relative risks were found for melanoma, lip cancer, head and neck cancer, lung cancer, myelodysplastic diseases, and cancer with unknown primary site. In addition, risks were elevated for female breast cancer and kidney cancer following a first primary MCC, and for subsequent MCCs following first primary colorectal cancer, prostate cancer, non-Hodgkin lymphoma, or lymphoid leukemia. These results suggest that several shared pathways are likely for MCC and other cancers, including immunosuppression, UV radiation, and genetics.

Mixture models of nucleotide sequence evolution that account for heterogeneity in the substitution process across sites and across lineages

Molecular phylogenetic studies of homologous sequences of nucleotides often assume that the underlying evolutionary process was globally stationary, reversible, and homogeneous (SRH), and that a model of evolution with one or more site-specific and time-reversible rate matrices (e.g., the GTR rate matrix) is enough to accurately model the evolution of data over the whole tree. However, an increasing body of data suggests that evolution under these conditions is an exception, rather than the norm. To address this issue, several non-SRH models of molecular evolution have been proposed, but they either ignore heterogeneity in the substitution process across sites (HAS) or assume it can be modeled accurately using the distribution. As an alternative to these models of evolution, we introduce a family of mixture models that approximate HAS without the assumption of an underlying predefined statistical distribution. This family of mixture models is combined with non-SRH models of evolution that account for heterogeneity in the substitution process across lineages (HAL). We also present two algorithms for searching model space and identifying an optimal model of evolution that is less likely to over- or underparameterize the data. The performance of the two new algorithms was evaluated using alignments of nucleotides with 10 000 sites simulated under complex non-SRH conditions on a 25-tipped tree. The algorithms were found to be very successful, identifying the correct HAL model with a 75% success rate (the average success rate for assigning rate matrices to the tree's 48 edges was 99.25%) and, for the correct HAL model, identifying the correct HAS model with a 98% success rate. Finally, parameter estimates obtained under the correct HAL-HAS model were found to be accurate and precise. The merits of our new algorithms were illustrated with an analysis of 42 337 second codon sites extracted from a concatenation of 106 alignments of orthologous genes encoded by the nuclear genomes of Saccharomyces cerevisiae, S. paradoxus, S. mikatae, S. kudriavzevii, S. castellii, S. kluyveri, S. bayanus, and Candida albicans. Our results show that second codon sites in the ancestral genome of these species contained 49.1% invariable sites, 39.6% variable sites belonging to one rate category (V1), and 11.3% variable sites belonging to a second rate category (V2). The ancestral nucleotide content was found to differ markedly across these three sets of sites, and the evolutionary processes operating at the variable sites were found to be non-SRH and best modeled by a combination of eight edge-specific rate matrices (four for V1 and four for V2). The number of substitutions per site at the variable sites also differed markedly, with sites belonging to V1 evolving slower than those belonging to V2 along the lineages separating the seven species of Saccharomyces. Finally, sites belonging to V1 appeared to have ceased evolving along the lineages separating S. cerevisiae, S. paradoxus, S. mikatae, S. kudriavzevii, and S. bayanus, implying that they might have become so selectively constrained that they could be considered invariable sites in these species.

Plasmon nanofocusing with negative refraction in a high-index dielectric wedge

Here, we describe a metal-insulator-insulator nanofocusing structure formed by a high-permittivity dielectric wedge on a metal substrate. The structure is shown to produce nanofocusing of surface plasmon polaritons (SPPs) in the direction opposite to the taper of the wedge, including a range of nanoplasmonic effects such as nanofocusing of SPPs with negative refraction, formation of plasmonic caustics within a nanoscale distance from the wedge tip, mutual transformation of SPP modes, and significant local field enhancements in the adiabatic and strongly nonadiabatic regimes. A combination of approximate analytical and rigorous numerical approaches is used to analyze the strength and position of caustics in the structure. In particular, it is demonstrated that strong SPP localization within spatial regions as small as a few tens of nanometers near the caustic is achievable in the considered structures. Contrary to other nanofocusing configurations, efficient nanofocusing is shown to occur in the strongly nonadiabatic regime with taper angles of the dielectric wedge as large as ∼40° and within uniquely short distances (as small as a few dozens of nanometers) from the tip of the wedge. Physical interpretations of the obtained results are also presented and discussed.

Properties of the correlated electronic states of pyrene and hydrogenated pyrene

Approximate calculations are reported on pyrene within the PPP model Hamiltonian using a novel restricted CI scheme which employs both molecular orbital and valence bond techniques. Also reported are detailed full CI results of the PPP model on 2,7-dihydropyrene obtained using the valence bond method. Spectral studies, charge and spin density calculations in ground and excited states, and ring current calculations in the ground state of the molecules are presented. In pyrene, the calculated excitation energies are in good agreement with experiment. The closed structure pi-conjugated molecule pyrene appears to show smaller distortions from the ground state geometry compared with the open structure pi-conjugated molecule 2,7-dihydropyrene. The ground state equilibrium structure of 2,7-dihydropyrene can be viewed as two hexatriene molecules connected by a vinyl crosslink, as is evident from bond order and ring current calculations. This is consistent with the only Kekule resonant structure possible for this molecule.