RESEARCH INTERESTS
Decision-theoretic artificial intelligence; temporal probabilistic reasoning; machine learning; adaptive computing; biomedical informatics.
RESEARCH GROUP ACTIVITIES
Medical Computing Laboratory, School of Computing, National University of Singapore
RESEARCH SUMMARY
I am directing and working on an international, multi-disciplinary research agenda on dynamic decision modeling, machine learning in large biomedical databases, automated model construction from multiple knowledge sources, and time-critical reasoning.
The main emphasis is on developing new computational methodologies and information processing frameworks to formulate and analyze complex decision problems that are changing over time, situated in uncertain settings, and under resource constraints.
The projects usually involve real-life biomedical problems in diagnostic test and treatment management, practice guidelines generation, outcomes prediction, and cost-effectiveness analysis; some other domains in homeland security, e-commerce, and finance, etc., are involved as well.
Our recent research foci also include combining genotypic and phenotypic information to support clinical decision making, Bayesian approaches to construct gene, protein, and metabolite pathways and networks in systems biology, and in general bridging the knowledge representation and inference gap between bioinformatics and medical informatics problems.
RECENTLY COMPLETED RESEARCH WORK
Our most recent large project on Intelligent Prognostic Analysis in Medicine, has been successfully completed in 2006:
The Intelligent Prognostic Analysis in Medicine project was a 4-year, $1.67 million project funded by the Biomedical Research Council of A*Star and the Ministry of Education in Singapore. Prognostic analysis is a critical part of evidence-based medicine that emphasizes the effective use of information to improve quality, reduce variation, and manage resources in health care procedures. Prognostic analysis illuminates the natural, as well as the expected, post-intervention course and outcome of natural processes. This work aimed to develop a set of advanced computing and decision engineering techniques to support effective prognostic analysis in medicine; the resulting techniques were incorporated into a set of prototype applications that automate clinical practice guideline and outcome model generation in significant and time-critical health care domains. The basic research issues were motivated by the application challenges; the results were in turn tested in the application prototypes.
The research activities have led to the following areas of continuing work:
Basic Research Continuing research work on new representation, reasoning, and learning techniques are on-going that aim to support modeling and analysis of important phenomena in complex domains that contain one or more of the challenges in: a) Information at multiple level of details, from multiple, heterogeneous sources, targeted at multiple, distributed users; b) Adaptations to temporal and environmental changes; c) Deterministic and uncertain information; d) Multi-modal information, e.g., 3-D magnetic resonance images.
Translational Research – Clinical Implementation Some recent research results and prototypes on outcome analyses and clinical practice guideline development are being implemented in an open-source trial framework in the actual clinical settings at the National University Hospital, the Singapore National Asthma Program, and Gleneagles Hospital; the framework aims to support best practices in prospective and acute care in respiratory medicine. Funding is from from governmental development grants. The trial aims to bring together the research team, an engineering team, and a clinical team to identify and resolve any potential technical and organizational issues pertaining to the successful deployment of the technologies. This is a novel trial implementation model, especially in Singapore, aiming to facilitate and speed-up the technology transfer process. If successful, the research results and insights could be productized and/or commercialized in a 3-5 year time-frame, instead of the normal lag-time for biomedical informatics technology transfer of 10-15 years.