Available postgraduate research projects

Evidence-based Decision Support

Modelling Communication in Health Organizations Using Multi- Agent Models

(Full time Scholarship available for this project)

In partnership with the Prince of Wales Hospital we are seeking expressions of interest for a full-time scholarship funded by an Australian Research Council Linkage project ’Agent-based methods for communication system design in complex organizations’. Working as part of our Clinical Communication team, and liaising with our Health System Simulation group, the candidate will develop an agent based simulation of communication and team work in a hospital setting. The simulation will be used to support IT system design and integration and will be validated by observation of current work practices.

Prerequisites: The candidate will have a primary background in computer science or information systems, or be able to demonstrate equivalent skills.
Contact:
Prof. Enrico Coiera, e.coiera@unsw.edu.au
A/Prof. Wayne Wobcke, wobcke@cse.unsw.edu.au

Supporting Decision-Making and Discovery in the Biomedical Sciences

(Full time Scholarship available for this project)

A fundamental barrier to research is the ability of researchers and clinicians to 'join the dots' and formulate innovative hypotheses, for example about the cause of a disease or the diagnosis of an ill patient. In this project we will attack the complex challenge of using computer-based methods to support hypothesis formation and machine discovery, a long standing challenge in the machine learning research area. Through the use of novel
computational technologies for machine discovery, the candidate will be asked to develop novel approaches to support hypothesis formation,
reasoning and discovery for clinical bioscientists. The area of application will be negotiable, but is likely to be associated with a major UNSW bioscience research strength such as cancer or infectious diseases.

Prerequisites: The candidate will have a primary background in computer science or information systems, or be able to demonstrate equivalent skills.
Contact: Prof. Enrico Coiera, e.coiera@unsw.edu.au

Biomedical Text Mining

Automatic Document Summarization of Clinical Studies

Today medical practitioners and researchers are confronted with a hugh information overload as the amount of scientific literature and medical knowledge expands at an exponential pace. Clinicians rely on systematic reviewers or human experts to synthesize medical evidence and create summaries and guidelines for information on the latest medical evidence. This project aims to investigate ways to automatically synthesize concise and informative summaries across multiple documents. The student will work on computational algorithms for locating relevant information content from documents, and compressing them to form well formulated summaries of selected topics.

Information Extraction of Clinical Studies

This project involves the development natural language processing algorithms towards the extraction of knowledge documented in research articles in medical science. Randomized controlled trials, and cohort studies are important in providing evidence of effects of treatments for any conditions. The student will design and implement a program for automatically extracting experimental methods, as well as the benefits and harms of medical interventions. The results of this project could be applied to intelligent search engines, knowledge discovery systems and integrated with clinical decision support systems.

Prerequisites:

Prof. Enrico Coiera, e.coiera@unsw.edu.au

Translational Bioinformatics

Antibiotic Resistance Management and Antibiotics Prescription Support

We are working with clinicians from Westmead Hospital to improve the process of antibiotic drug prescription by removing the guesswork from the antibiotic selection process. A key aspect of the research is finding the biological mechanisms that make bacterial infections resistant to antibiotics. These mechanisms typically make resistance genes mobile allowing them to spread between organisms and accumulate resistances over time.

In this project you will use machine learning and grammatical approaches to discover genes and structures in DNA sequences; develop clinical decision support systems that can help with an efficient management of antibiotics delivery, leading to best patient outcomes and reduced hospital costs. You will develop novel algorithms using state-of-the-art data-mining and machine learning techniques, generate computational models from a variety of data sources including DNA structures, published literature and public health databases.

Multiple projects exist in this project at any level between undegraduate honours thesis to PhD.

Prerequisites:

Intelligent High-throughput Genetic Analysis

Since the completion of the Human Genome project in 2003, the surprising low number of genes (~25,000) made it obvious that analysis of the relationships between genes is required. Such analysis methods can be used to explain a diverse arrange of diseases from viral and bacterial infections through Cancer and Diabetes to Alzheimer’s. The amount and detail of genetic data is overwhelming even for computers so a lot of it goes unanalysed. The work in this project aims to improve our understanding of large-scale DNA processes. In this project you will use statistical, machine learning and formal languages to discover new genes, new structures and related molecular mechanisms.

Multiple projects exist in this project at any level between undegraduate honours thesis to PhD.

Prerequisites:

Contact: Dr. Guy Tsafnat, guyt@unsw.edu.au

Public Health Biosurveillance

Unraveling contact network structures from genotyping data

Contact network models simulate the spread of infectious diseases considering specific patterns of contacts between individuals or groups
within a community. These epidemiological models have been proved to be more accurate and of higher utility to public health than the more traditional random-mixing models [Keeling 2005, Meyers et al 2006]. Collecting the required information to define a network model is however a daunting task and often idealized networks are used. This project will aim to discover contact network patterns within a community using newly available molecular genotype information from bacteria with epidemic potential.

References:
1. Keeling M., 2005. The implications of network structure for epidemic dynamics. Theoretical Population Biology, 67, 1-8.
2. Meyers L.A., Newman M.E.J., and Pourbohloul B., 2006. Predicting epidemics on directed contact networks. Journal of Theoretical
Biology, 240, 400-418.

Prerequisites: Degree in Mathematics, Physics, Computer Sciences or related field.
Note: This project will take place in collaboration with the University of Sussex, UK
Contact: Dr Blanca Gallego, b.gallego@unsw.edu.au

Imaging Informatics

Medical Image Understanding

Medical imaging systems are constantly improving in image quality, which results a growing number of images that have to be inspected for diagnosis. Computational aids are required to filter the large number of images and to focus the radiologist’s attention on diagnostically
interesting events. Medical image understanding approach combines image processing and machine learning for detection and recognition of
disease patterns in images. It also uses knowledge of the imaged anatomy to guide the image interpretation.

Current project is on Detecting and recognizing disseises of the lungs on High Resolution CT images. However the medical image understanding
approach can be applied to any imaging modality and any body part.

References:
1. T. Zrimec, C. Sammut, ’Medical Image Understanding System’, Engineering applications of Artificial Intelligence,10(1) 31-39, 1997.
2. Zrimec T, Busayarat S, ’System for Computer Aided Detection of Diseases Patterns on High Resolution CT images of the Lungs', 20th
IEEE International Symposium on Computer-Based Medical Systems, CBMS 2007, Maribor, SLO, June 2007; 41-46.

Prerequisites: Computer science, Biomedical Engineering, Electrical Engineering
Contact: Tatjana Zrimec, tatjana@cse.unsw.edu.au

Clinical Systems Safety Engineering

Automation bias in clinical decision errors

One type of error associated with the use of decision support in complex clinical tasks (e.g. prescribing medications) is automation bias where people using an automated decision aid act as the aid directs them to, irrespective of the correctness of the suggested action. For example, individuals may be less vigilant in checking drug orders that are generated by a computer because they assume the computer will have already done the work, or continue with a dangerous drug order because the computer did not alert them that the order was unsafe. Thus use of decision support could lead to errors of omission where individuals miss important data because the system does not prompt them to notice them, or to errors of commission where individuals do what the decision aid tells or allows them to do, even when this contradicts their training and other available data. This two-part project will firstly examine the impact of automation bias on medication errors with ethnographic methods and simulation using cognitive architectures. In the second part we will look at designing novel computer interfaces for prescribing systems that are safer and more tolerant of automation bias. 

Pre-requisites: Degree in Computer Science, Engineering or Psychology

Using a systems approach to modelling accidents in health

An accident model is a representation of the processes which give rise to failures within a system. Leveson's systemic approach to accident modelling considers safety as a control problem. In this model accidents result from inadequate control and enforcement of safety constraints on a system because (i) hazards are not known, (ii) control action is not adequate or the wrong action is performed, (iii) inconsistencies between process models used by the automation or human (mental models) and the actual process, (iv) missing or inadequate feedback. The approach views accidents as consequences of socio-technical interactions among all levels of the organizational hierarchy that violate system safety constraints. Systemic accident models that account for failures in highly automated industrial systems are not directly transferable to healthcare where automation largely supports humans in making complex decisions. However they provide a framework to examine complex relationships at multiple levels of the process, context and task interactions that contribute to clinical error. This project will investigate the application of systemic accident models to examine the safety of clinical software.  

Pre-requisites: Degree in Computer Science or Engineering 

Contact: Dr. Farah Magrabi f.magrabi@unsw.edu.au

Current & past postgraduate research projects

1. The impact of information-searching on health-related decision making
2. Translating bacterial molecular epidemiology into information to improve infectious disease risk assessment and control
3. Understanding the impact of hospital context on the spread and control of Methicillin-resistant Staphylococcus aureus (MRSA) through broad-scale systems simulation
4. A framework to extract, interpret and structure relevant information from clinical free-texts
5. Decision by design - decision support for antibiotic prescribing in critical care
6. 3D visualization tool for viewing data from heterogeneous sources - application of geo-spatial mapping of infectious disease outbreaks
7. Development of a computer aided detection and diagnosis of diffuse lung diseases