Search

This paper describes a multicentre longitudinal cohort study to evaluate the predictive accuracy of a regularised Bayesian neural network model in a prognostic application. The study sample (n = 5442) comprises subjects treated with intraocular melanoma in two different centres in Liverpool and Paris. External validation was carried out by fitting the model to the data from Liverpool set and predicting for the data from Paris. The performance of the model in out-of-sample prediction was assessed statistically for discrimination of outcomes and calibration. It was also evaluated clinically by comparing against the accepted TNM staging system. The model had good discrimination with Harrell's C index > 0.7 up to ten years of follow-up. Calibration results were also good up to ten years using a Hosmer-Lemeshow type analysis (p > 0.05). The paper: 1) deals with the issue of missing data using methods that are well accepted in the literature; 2) proposes a framework for externally validating machine learning models applied to survival analysis; 3) applies accepted methods for dealing with missing data; 4) proposes an alternative staging system based on the model. The new staging system, which takes into account histopathologic information, has several advantages over the existing staging system.

Missing values are common in medical datasets and may be amenable to data imputation when modelling a given data set or validating on an external cohort. This paper discusses model averaging over samples of the imputed distribution and extends this approach to generic non-linear modelling with the partial logistic artificial neural network (PLANN) regularised with automatic relevance determination (ARD). The study then applies the imputation to external validation, considering also predictions made for individual patients. A prognostic index is defined for the non-linear model and validation results show that four statistically significant risk groups identified at 95% level of confidence from the modelling data, from Christie Hospital (n = 931), retain good separation during external validation with data from the BC Cancer Agency (BCCA) (n = 4,083). A satisfactory discrimination and calibration performance was assessed with the time dependent C index (C td) and Hosmer-Lemeshow statistic, respectively, for both, training and validated model.

The authors describe two different approaches to object orientation detection: one method uses first- and second-order moments, while the other uses a multilayer perceptron network trained by back error propagation. A comparison between these methods shows that the neural network is able to generalize the trained orientations for different classes of objects and affords better control in determining the orientation at the pick-up point. Maximum orientation resolution is achieved economically by using a form of coarse coding, in which the output excitations for all orientations are spread out among neighboring output nodes. >

The classification of two-dimensional binary images by artificial neural networks, irrespective of their position, orientation, and size, is investigated using two complementary methods. Third-order networks were used first. Invariance under all three transformations was achieved by grouping triplets of pixels into appropriate equivalent classes. A suitable reduction in the number of weights then resulted in economical networks which exhibit high recognition rates under all transformations simultaneously, together with robustness against local distortions. The performance of these networks was tested on invariant classification of both typed and handwritten numerals. It was found to be superior to that achieved by the more traditional method of preprocessing that image using moments. Zernike moments were used because they are known to be well suited for pattern classification. In both cases, the network was trained by back-error propagation. >

A novel image compression technique using classified vector quantiser and singular value decomposition is proposed for the efficient representation of still images. The proposed method is called improved hybrid classified vector quantisation. The proposed technique was benchmarked with the standard vector quantiser generated using the k-means algorithm, and JPEG-2000. Simulation results indicate that the proposed approach alleviates edge degradation and can reconstruct good visual quality images with higher peak signal-to-noise ratio than the benchmarked techniques.

A new application of dynamic link architectures is proposed for the automatic selection of key features in multiple instances of facial images, from manually tagged features in a single facial image of each person. This method is evaluated for its accuracy in face recognition, determined by the frequency with which the model graph fitted to a single instance of an individual's face achieves the closest match to other images of the same person. In addition to the potential of this method for face recognition, the methodology for feature alignment is a practical enabler to recognise emotion in affective computing. The system was first trained to evaluate typical intrapersonal variations of facial features on a training subset with ten facial images from six individuals in the Manchester Face Database (Lanitis et al., 1993). After the training stage, facial features in an image of a new face were assigned the intrapersonal variations obtained for the corresponding features during the training stage. The saliency measure for each local image feature was then computed within a Bayesian framework and the accuracy of face recognition was evaluated with a further three images each from 24 people, taken from the same data set. A refinement of the saliency framework that used only a subset of local features for face recognition further increased the accuracy of face recognition on the same test database.

The design of high speed permanent magnet machines needs to address conflicting requirements arising from the requirements of high electromechanical efficiency and mechanical strength and smoothness of running. This paper investigates the use of canning material surrounding the rotor, to give it mechanical strength, and investigates its consequences with regard to the electromagnetic performance of the motor at high speeds. Two designs are investigated in detail. The performance of a high-field magnet configuration is assessed experimentally, and compares well with the results predicted from finite-element computations.

We address the question of designing optimum discrete sets of points to represent numerically a continuous group manifold. We consider subsets which are extensions of the regular discrete subgroups. Applications to Monte Carlo simulation of SU(2) and SU(3) gauge theory are discussed.

Methods for deriving a static baryon-baryon (nucleon-nucleon as well as nucleon-delta and transition) potential from the Skyrme model are briefly reported. Some results obtained with these methods are also presented and discussed.

We report on a numerical calculation of topological charge densities in non-abelian gauge theory with gauge groups SU(2) and SU(3). The group manifold is represented by a discrete subset thereof which lies outside its finite subgroups. The results shed light on the usefulness of these representations in Monte Carlo evaluations of non-abelian lattice gauge theory.

The elucidation of biological networks regulating the metabolic basis of disease is critical for understanding disease progression and in identifying therapeutic targets. In molecular biology, this process often starts by clustering expression profiles which are candidates for disease phenotypes. However, each cluster may comprise several overlapping processes that are active in the cluster. This paper outlines empirical results using methods for blind source separation to map the pathways of biomarkers driving independent, hidden processes that underpin the clusters. The method is applied to a protein expression data set measured in tissue from breast cancer patients (n=1,076).

Smooth nonlinear topographic maps of the data distribution to guide a Grand Tour visualisation.Prioritisation of data linear views that are most consistent with data structure in the maps.Useful visualisations that cannot be obtained by other more classical approaches. Dimensionality reduction is required to produce visualisations of high dimensional data. In this framework, one of the most straightforward approaches to visualising high dimensional data is based on reducing complexity and applying linear projections while tumbling the projection axes in a defined sequence which generates a Grand Tour of the data. We propose using smooth nonlinear topographic maps of the data distribution to guide the Grand Tour, increasing the effectiveness of this approach by prioritising the linear views of the data that are most consistent with global data structure in these maps. A further consequence of this approach is to enable direct visualisation of the topographic map onto projective spaces that discern structure in the data. The experimental results on standard databases reported in this paper, using self-organising maps and generative topographic mapping, illustrate the practical value of the proposed approach. The main novelty of our proposal is the definition of a systematic way to guide the search of data views in the grand tour, selecting and prioritizing some of them, based on nonlinear manifold models.

There are currently few commercially available clinical products with neural network components-in fact only one has gained FDA approval, and none have yet been certified by the MDA. A major step in the certification process is proving the integrity of the neural network component, which cannot be done by reference to substantially equivalent certified products because they do not exist. Consequently, the case for the use of the new technology has to be made explicitly, by recourse to best practice in its design and evaluation. This paper discusses the use neural networks in medical applications, with reference to a case study from magnetic resonance spectroscopy (MRS).

PROBLEM SETTING: Support vector machines (SVMs) are very popular tools for classification, regression and other problems. Due to the large choice of kernels they can be applied with, a large variety of data can be analysed using these tools. Machine learning thanks its popularity to the good performance of the resulting models. However, interpreting the models is far from obvious, especially when non-linear kernels are used. Hence, the methods are used as black boxes. As a consequence, the use of SVMs is less supported in areas where interpretability is important and where people are held responsible for the decisions made by models. OBJECTIVE: In this work, we investigate whether SVMs using linear, polynomial and RBF kernels can be explained such that interpretations for model-based decisions can be provided. We further indicate when SVMs can be explained and in which situations interpretation of SVMs is (hitherto) not possible. Here, explainability is defined as the ability to produce the final decision based on a sum of contributions which depend on one single or at most two input variables. RESULTS: Our experiments on simulated and real-life data show that explainability of an SVM depends on the chosen parameter values (degree of polynomial kernel, width of RBF kernel and regularization constant). When several combinations of parameter values yield the same cross-validation performance, combinations with a lower polynomial degree or a larger kernel width have a higher chance of being explainable. CONCLUSIONS: This work summarizes SVM classifiers obtained with linear, polynomial and RBF kernels in a single plot. Linear and polynomial kernels up to the second degree are represented exactly. For other kernels an indication of the reliability of the approximation is presented. The complete methodology is available as an R package and two apps and a movie are provided to illustrate the possibilities offered by the method. ispartof: PLoS One vol:11 issue:10 pages:1-33 ispartof: location:United States status: published

We introduce a probabilistic algorithm for binary classification based on the SVM through the application of the ANOVA decomposition for multivariate functions to express the logit of the Platt estimate of the posterior probability as a non-redundant sum of functions of fewer variables (partial responses) followed by feature selection with the Lasso. The partial response SVM (prSVM) is compared with previous interpretable models of the SVM. Its accuracy and stability are demonstrated with real-world data sets.

Gynaecologists and obstetricians visually interpret cardiotocography (CTG) traces using the International Federation of Gynaecology and Obstetrics (FIGO) guidelines to assess the wellbeing of the foetus during antenatal care. This approach has raised concerns among professionals with regards to inter- and intra-variability where clinical diagnosis only has a 30\% positive predictive value when classifying pathological outcomes. Machine learning models, trained with FIGO and other user derived features extracted from CTG traces, have been shown to increase positive predictive capacity and minimise variability. This is only possible however when class distributions are equal which is rarely the case in clinical trials where case-control observations are heavily skewed in favour of normal outcomes. Classes can be balanced using either synthetic data derived from resampled case training data or by decreasing the number of control instances. However, this either introduces bias or removes valuable information. Concerns have also been raised regarding machine learning studies and their reliance on manually handcrafted features. While this has led to some interesting results, deriving an optimal set of features is considered to be an art as well as a science and is often an empirical and time consuming process. In this paper, we address both of these issues and propose a novel CTG analysis methodology that a) splits CTG time-series signals into n-size windows with equal class distributions, and b) automatically extracts features from time-series windows using a one dimensional convolutional neural network (1DCNN) and multilayer perceptron (MLP) ensemble. Collectively, the proposed approach normally distributes classes and removes the need to handcrafted features from CTG traces., Comment: 11 Pages, 12 Figures (excluding profile pictures), accepted for publication in IEEE Transactions in Emerging Topics in Computational Intelligence

The need for multivariate analysis of magnetic resonance spectroscopy (MRS) data was recognized about 20 years ago, when it became evident that spectral patterns were characteristic of some diseases. Despite this, there is no generally accepted methodology for performing pattern recognition (PR) analysis of MRS data sets. Here, the data acquisition and processing requirements for performing successful PR as applied to human MRS studies are introduced, and the main techniques for feature selection, extraction, and classification are described. These include methods of dimensionality reduction such as principal component analysis (PCA), independent component analysis (ICA), non-negative matrix factorization (NMF), and feature selection. Supervised methods such as linear discriminant analysis (LDA), logistic regression (LogR), and nonlinear classification are discussed separately from unsupervised and semisupervised classification techniques, including k –means clustering. Methods for testing and metrics for gauging the performance of PR models (sensitivity and specificity, the ‘Confusion Matrix’, ‘k –fold cross-validation’, ‘Leave One Out’, ‘Bootstrapping’, the ‘Receiver Operating Characteristic curve’, and balanced error and accuracy rates) are briefly described. This article ends with a summary of the main lessons learned from PR applied to MRS to date.

Several approaches have been proposed for the analysis of DNA microarray datasets, focusing on the performance and robustness of the final feature subsets. The novelty of this paper arises in the use of q-values to pre-filter the features of a DNA microarray dataset identifying the most significant ones and including this information into a genetic algorithm for further feature selection. This method is applied to a lung cancer microarray dataset resulting in similar performance rates and greater robustness in terms of selected features (on average a 36.21% of robustness improvement) when compared to results of the standard algorithm.

Summarization: Acute Myeloid Leukaemia (AML) is a haematologic malignancy. Despite of improvements in prognosis it is still a severe disease and only a minority of patients are cured. Recently, insights into specific cytogenetics and biological aberration involved in the leukaemogenesis process are the target of research. LAM-99P GIMEMA Protocol identifies treatment guidelines for AML in adults. A methodological approach is proposed, compounding different modelling techniques, to assess the prognostic relevance of clinical as well as biological markers with regard to clinical endpoint, such as the achievement of Complete Remission in the short-term and long-term survival taking into account competing risks. Παρουσιάστηκε στο: International Journal of Biomedical Engineering and Technology

BackgroundThe relationship between well-being and mental ill health is complex; people may experience very low levels of well-being even in the absence of overt mental health problems.AimsThis study tested the hypothesis that anxiety, depression and well-being have different causal determinants and psychological mediating mechanisms.MethodThe influence of causal and mediating factors on anxiety, depression and well-being were investigated in a cross-sectional online questionnaire survey hosted on a UK national broadcasting website.ResultsMultivariate conditional independence analysis of data from 27 397 participants revealed different association pathways for the two constructs. Anxiety and depression were associated with negative life events mediated by rumination; low levels of subjective well-being were associated with material deprivation and social isolation, mediated by adaptive coping style.ConclusionsOur findings support the ‘two continua’ model of the relationship between psychological well-being and mental health problems, with implications for both treatment and prevention.

Algorithms in the Adaptive Resonance Theory (ART) family adapt to structural changes in data as new information presents, making it an exciting candidate for dynamic online clustering of big health data. Its use however has largely been restricted to the signal processing field. In this paper we introduce an refinement of the ART2-A method within an adapted separation and concordance (SeCo) framework which has been shown to identify stable and reproducible solutions from repeated initialisations that also provides evidence for an appropriate number of initial clusters that best calibrates the algorithm with the data presented. The results show stable, reproducible solutions for a mix of real-world heath related datasets and well known benchmark datasets, selecting solutions which better represent the underlying structure of the data than using a single measure of separation. The scalability of the method and it's facility for dynamic online clustering makes it suitable for finding structure in big data.

Objectives Early identification of patients with blood stream infection (BSI), especially bacteraemia, is important as prompt treatment improves outcome. The initial stages of severe infection may be characterised by increased numbers of neutrophils in the peripheral blood and depression of the lymphocyte count (LC). The neutrophil to LC ratio (NLCR) has previously been compared with conventional tests, such as C-reactive protein (CRP) and white cell count (WCC), and has been proposed as a useful marker in the timely diagnosis of bacteraemia.\ud \ud Methods Data on consecutive adult patients presenting to the emergency department with pyrexial illness during the study period, November 2009 to October 2010, were analysed. The main outcome measure was positive blood cultures (bacteraemia). Sensitivity, specificity, positive and negative predictive values and likelihood ratios were determined for NLCR, CRP, WCC, neutrophil count and LC.\ud \ud Results 1954 patients met the inclusion criteria. Blood cultures were positive in 270 patients, hence the prevalence of bacteraemia was 13.8%. With the exception of WCC, there were significant differences in the mean value for each marker between bacteraemic and non-bacteraemic patients (p

In this paper we examine the use of population segmentation modelling for targeting fire prevention to the needs of the community. A population segmentation approach based upon socio-economic characteristics data was developed to provide a deeper understanding of the fire risks associated with different social groups by a partnership consisting of a UK fire and rescue service, a National Health Service trust, a local council, and a police force. This approach supported more targeted and co-ordinated community fire prevention measures by the agencies involved. This approach was used to target those most at risk, and improve intra-agency co-ordination and collaboration between the agencies involved. The modelling enabled differences in terms of the risk of fire related injuries and fatalities between the population segments to be examined. Overall, the research examines how and why population segmentation was undertaken by the fire and rescue service studied, and how this was implemented and used operationally to support fire prevention activities. The project was funded by the UK Department of Communities and Local Government.

A major challenge in delivering reliable and trustworthy computational intelligence for practical applications in clinical medicine is interpretability. This aspect of machine learning is a major distinguishing factor compared with traditional statistical models for the stratification of patients, which typically use rules or a risk score identified by logistic regression.\ud We show how functions of one and two variables can be extracted from pre-trained machine learning models using anchored Analysis of Variance (ANOVA) decompositions. This enables complex interaction terms to be filtered out by aggressive regularisation using the Least Absolute Shrinkage and Selection Operator (LASSO) resulting in a sparse model with comparable or even better performance than the original pre-trained black-box. \ud Besides being theoretically well-founded, the decomposition of a black-box multivariate probabilistic binary classifier into a General Additive Model (GAM) comprising a linear combination of non-linear functions of one or two variables provides full interpretability. In effect this extends logistic regression into non-linear modelling without the need for manual intervention by way of variable transformations, using the pre-trained model as a seed.\ud The application of the proposed methodology to existing machine learning models is demonstrated using the Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), Random Forests (RF) and Gradient Boosting Machines (GBM), to model a data frame from a well-known benchmark dataset available from Physionet, the Medical Information Mart for Intensive Care (MIMIC-III). Both the classification performance and plausibility of clinical interpretation compare favourably with other state-of-the-art sparse models namely Sparse Additive Models (SAM) and the Explainable Boosting Machine (EBM).

A fundamental bimodality of galaxies in the local Universe is apparent in many of the features used to describe them. Multiple sub-populations exist within this framework, each representing galaxies following distinct evolutionary pathways. Accurately identifying and characterising these sub-populations requires that a large number of galaxy features be analysed simultaneously. Future galaxy surveys such as LSST and Euclid will yield data volumes for which traditional approaches to galaxy classification will become unfeasible. To address this, we apply a robust $k$-means unsupervised clustering method to feature data derived from a sample of 7338 local-Universe galaxies selected from the Galaxy And Mass Assembly (GAMA) survey. This allows us to partition our sample into $k$ clusters without the need for training on pre-labelled data, facilitating a full census of our high dimensionality feature space and guarding against stochastic effects. We find that the local galaxy population natively splits into $2$, $3$, $5$ and a maximum of $6$ sub-populations, with each corresponding to a distinct ongoing evolutionary mechanism. Notably, the impact of the local environment appears strongly linked with the evolution of low-mass ($M_{*} < 10^{10}$ M$_{\odot}$) galaxies, with more massive systems appearing to evolve more passively from the blue cloud onto the red sequence. With a typical run time of $\sim3$ minutes per value of $k$ for our galaxy sample, we show how $k$-means unsupervised clustering is an ideal tool for future analysis of large extragalactic datasets, being scalable, adaptable, and providing crucial insight into the fundamental properties of the local galaxy population., 27 pages, 28 figures, 8 tables. Accepted for publication in the Monthly Notices of the Royal Astronomical Society

Type 2 Diabetes condition is a multifactorial disorder involves the convergence of genetics, environment, diet and lifestyle risk factors. This paper investigates genetic and conventional (clinical, sociodemographic) risk factors and their predictive power in classifying Type 2 Diabetes. Six statistically significant Single Nucleotide Polymorphisms (SNPs) associated with Type 2 Diabetes are derived by conducting logistic association analysis. The derived SNPs in addition to conventional risk factors are used to model supervised machine learning algorithms to classify cases and controls in genome wide association studies (GWAS). Models are trained using genetic variable analysis, genetic and conventional variable analysis, and conventional variable analysis. The results demonstrate of the three models, higher predictive capacity is evident when genetic and conventional predictors are combined. Using a Random Forest classifier, the Area Under the Curve=73.96%, Sensitivity=68.42 %, and Specificity=78.67%.

Background Magnetic resonance spectroscopy provides metabolic information about living tissues in a non-invasive way. However, there are only few multi-centre clinical studies, mostly performed on a single scanner model or data format, as there is no flexible way of documenting and exchanging processed magnetic resonance spectroscopy data in digital format. This is because the DICOM standard for spectroscopy deals with unprocessed data. This paper proposes a plugin tool developed for jMRUI, namely jMRUI2XML, to tackle the latter limitation. jMRUI is a software tool for magnetic resonance spectroscopy data processing that is widely used in the magnetic resonance spectroscopy community and has evolved into a plugin platform allowing for implementation of novel features. Results jMRUI2XML is a Java solution that facilitates common preprocessing of magnetic resonance spectroscopy data across multiple scanners. Its main characteristics are: 1) it automates magnetic resonance spectroscopy preprocessing, and 2) it can be a platform for outputting exchangeable magnetic resonance spectroscopy data. The plugin works with any kind of data that can be opened by jMRUI and outputs in extensible markup language format. Data processing templates can be generated and saved for later use. The output format opens the way for easy data sharing- due to the documentation of the preprocessing parameters and the intrinsic anonymization - for example for performing pattern recognition analysis on multicentre/multi-manufacturer magnetic resonance spectroscopy data. Conclusions jMRUI2XML provides a self-contained and self-descriptive format accounting for the most relevant information needed for exchanging magnetic resonance spectroscopy data in digital form, as well as for automating its processing. This allows for tracking the procedures the data has undergone, which makes the proposed tool especially useful when performing pattern recognition analysis. Moreover, this work constitutes a first proposal for a minimum amount of information that should accompany any magnetic resonance processed spectrum, towards the goal of achieving better transferability of magnetic resonance spectroscopy studies. Electronic supplementary material The online version of this article (doi:10.1186/s12859-015-0796-5) contains supplementary material, which is available to authorized users.

Purpose – The purpose of the research project was to examine the process of developing a data sharing framework between different public sector organisations. Design/methodology/approach – A two-year case study of a data sharing project between a UK fire and rescue service, local council, NHS primary care trust and a police force was undertaken. Findings – It is important to carefully determine the requirements for data sharing, to establish data sharing agreements, to have secure arrangements for data sharing, and to ensure compliance with data protection legislation. Research limitations/implications – Data sharing between public sector organisations can operate effectively if appropriate care is taken when creating data sharing agreements between partner organisations. Practical implications – Data sharing can assist in reducing duplication of effort between public sector organisations and can reduce costs and enable more co-ordinated provision of public services. Originality/value – The detailed analysis of a data sharing case study identified the need for a systematic data sharing framework. Such a framework is proposed and illustrated with practical examples of specification, implementation and evaluation.

We construct a regular 38 880 element subset of SU(3) which is suitable for Monte Carlo simulation of SU(3) lattice gauge theory in the region of coupling β