Tze-Yun Leong

Dynamic decision analysis in medicine: a data-driven approach

Dynamic decision analysis concerns decision problems in which both time and uncertainty are explicitly considered. Two major challenges in dynamic decision analysis are on proper formulation of a model for the problem and effective elicitation of the numerous time-dependent conditional probabilities for the model. Based on a new, general dynamic decision modeling framework called DynaMoL (Dynamic decision Modeling Language), we propose a data-driven approach to addressing these issues. Our approach uses available problem data from large medical databases, guides the decision modeling at a proper level of abstraction and establishes a Bayesian learning method for automatic extraction of the probabilistic parameters. We demonstrate the theoretical implications and practical promises of this new approach to dynamic decision analysis in medicine through a comprehensive case study in the optimal follow-up of patients after curative colorectal cancer surgery.

Hemorrhage slices detection in brain CT images

Multi-slice computer tomography (CT) scans are widely used in todaypsilas diagnosis of head traumas. It is effective to disclose the bleeding and fractures. In this paper, we present an automated detection of CT scan slices which contain hemorrhages. Our method is robust towards various rotation, displacement and motion blur. Detection of these pathological slices will be useful for further diagnosis and retrieval.

Text Mining in Radiology Reports

Medical text mining has gained increasing interest in recent years. Radiology reports contain rich information describing radiologistpsilas observations on the patientpsilas medical conditions in the associated medical images. However, as most reports are in free text format, the valuable information contained in those reports cannot be easily accessed and used, unless proper text mining has been applied. In this paper, we propose a text mining system to extract and use the information in radiology reports. The system consists of three main modules: a medical finding extractor, a report and image retriever, and a text-assisted image feature extractor. In evaluation, the overall precision and recall for medical finding extraction are 95.5% and 87.9% respectively, and for all modifiers of the medical findings 88.2% and 82.8% respectively. The overall result of report and image retrieval module and text-assisted image feature extraction module is satisfactory to radiologists.

Characterization of medical time series using fuzzy similarity-based fractal dimensions

This paper attempts to characterize medical time series using fractal dimensions. Existing fractal dimensions like box, information and correlation dimensions characterize the time series by measuring the rate at which the distribution of the time series changes when the length (or radius) of the box (or hypersphere) is changed. However, the measured dimensions significantly vary when the box (or hypersphere) position is changed slightly. It happens because the data points just outside the box (or hypersphere) are not accounted for, and all the data points inside the box or hypersphere are treated equally. To overcome these problems, the hypersphere is converted to a Gaussian, and thus the hard boundary becomes soft. The Gaussian represents the fuzzy similarity between the neighbors and the point around which the Gaussian is constructed. This concept of similarity is exploited to propose a fuzzy similarity-based fractal dimension. The proposed dimension aims to capture the regularity of the time series in terms of how the fuzzy similarity scales up/down when the resolution of the time series is decreased/increased. Experiments on intensive care unit (ICU) data sets show that the proposed dimension characterizes the time series better than the correlation dimension.

Multiple perspective dynamic decision making

Abstract Decision making often involves deliberations in different perspectives. Distinct perspectives or views support knowledge acquisition and representation suitable for different types or stages of inference in the same discourse. This work presents a general paradigm for multiple perspective decision making over time and under uncertainty. Based on a unifying task definition and a common vocabulary for the relevant decision problems, this new paradigm balances the trade-off between model transparency and solution efficiency in current decision frameworks. The new paradigm motivates the design of DynaMoL (Dynamic decision Modeling Language), a general language for modeling and solving dynamic decision problems. The DynaMoL framework differentiates inferential and representational support for the modeling task from the solution or computation task. The dynamic decision grammar defines an extensible decision ontology and supports complex problem specification with multiple interfaces. The graphical presentation convention governs parameter visualization in multiple perspectives. The mathematical representation as semi-Markov decision process facilitates formal model analysis and admits multiple solution methods. A set of general translation techniques is devised to manage the different perspectives and representations of the decision parameters and constraints. DynaMoL has been evaluated on a prototype implementation, via some comprehensive case studies in medicine. The results demonstrate practical promise of the framework.

Translation initiation sites prediction with mixture Gaussian models in human cDNA sequences

Translation initiation sites (TISs) are important signals in cDNA sequences. Many research efforts have tried to predict TISs in cDNA sequences. In this paper, we propose to use mixture Gaussian models for TIS prediction. Using both local features and some features generated from global measures, the proposed method predicts TISs with a sensitivity of 98 percent and a specificity of 93.6 percent. Our method outperforms many other existing methods in sensitivity while keeping specificity high. We attribute the improvement in sensitivity to the nature of the global features and the mixture Gaussian models.

Feature selection for the prediction of translation initiation sites

Translation initiation sites (TISs) are important signals in cDNA sequences. In many previous attempts to predict TISs in cDNA sequences, three major factors affect the prediction performance: the nature of the cDNA sequence sets, the relevant features selected, and the classification methods used. In this paper, we examine different approaches to select and integrate relevant features for TIS prediction. The top selected significant features include the features from the position weight matrix and the propensity matrix, the number of nucleotide C in the sequence downstream ATG, the number of downstream stop codons, the number of upstream ATGs, and the number of some amino acids, such as amino acids A and D. With the numerical data generated from these features, different classification methods, including decision tree, naïve Bayes, and support vector machine, were applied to three independent sequence sets. The identified significant features were found to be biologically meaningful, while the experiments showed promising results.

Methods in Year 50: Preserving the Past and Preparing for the Future

BACKGROUND Founded in 1962 and, therefore, the oldest international journal in medical informatics, Methods of Information in Medicine will publish its 50th volume in 2011. At the start of the journals sixth decade, a discussion on the journals profile seems appropriate. OBJECTIVES To report on the new opportunities for online access to Methods publications as well as on the recent strategic decisions regarding the journals aims and editorial policies. METHODS Describing and analyzing the journals aims and scope. Reflecting on recent publications and on the journals development during the last decade. RESULTS From 2011 forward all articles of Methods from 1962 until the present can be accessed online. Methods of Information in Medicine stresses the basic methodology and scientific fundamentals of processing data, information and knowledge in medicine and health care. Although the journals major focus is on publications in medical informatics, it has never been restricted to publications only in this discipline. For example, articles in medical biometry, in or close to biomedical engineering, and, later, articles in bioinformatics continue to be a part of this journal. CONCLUSIONS There is a continuous and, as it seems, ever growing overlap in the research methodology and application areas of the mentioned disciplines. As there is a continuing and even growing need for such a publication forum, Methods of Information in Medicine will keep its broad scope. As an organizational consequence, the journals number of associate editors has increased accordingly.

Cost-effectiveness analysis of colorectal cancer screening strategies in Singapore: a dynamic decision analytic approach.

A dynamic decision analytic framework using local statistics and experts opinions is put to study the cost-effectiveness of colorectal cancer screening strategies in Singapore. It is demonstrated that any of the screening strategies, if implemented, would increase the life expectancy of the population of 50 to 70 years old. The model also determined the normal life expectancy of this population to be 76.32 years. Overall, Guaiac Fecal Occult Blood Test (FOBT) is most cost effective at SGD162.11 per life year saved per person. Our approach allowed us to model problem parameters that change over time and study the utility measures like cost and life expectancy for specific age within the range of 50- 69 through to 70 years old.

A model driven approach to imbalanced data sampling in medical decision making.

Classification is an important medical decision support function that can be seriously affected by disproportionate class distribution in the training data. In medical decision making, the rate of misclassification and the cost of misclassifying a minority (positive) class as a majority (negative) class are especially high. In this paper, we propose a new model-driven sampling approach to balancing data samples. Most existing data sampling methods produce new data points based on local, deterministic information. Our approach extends the idea of generative sampling to produce new data points based on an induced probabilistic graphical model. We present the motivation and the design of the proposed algorithm, and compare it with two representative imbalanced data sampling approaches on four medical data sets varying in size, imbalance ratio, and dimension. The empirical study helped identify the challenges in imbalanced data problems in medicine, and highlighted the strengths and limitations of the relevant sampling approaches. Performance of the model driven approach is shown to be comparable with existing approaches; potential improvements could be achieved by incorporating domain knowledge.