Skip to main content

Research Highlights

Research undertaken by the Curtin Institute for Computation (CIC) has been utilised in a number of industries across Australia. Below are short snippets of research highlights, some more in depth information is provided for the following case studies of research undertaken by CIC researchers and our collaborators:

Classification of webpages

The Multimodal Analysis Group within the Faculty of Humanities has been working with the CIC specialists to demonstrate the potential of a context-based mixed methods approach. They have been analysing how violent extremist groups use images and text to legitimise their views, incite violence and influence potential recruits and supporters in online propaganda materials. Additionally, the group is tracking how images from these materials are re-used for different purposes across media sites. The objective is to identify online communications that aim to legitimise, support, reject or resist terrorism in order to inform counter-terrorism initiatives designed to combat and prevent terrorist activities.

This collaboration involved developing an app to visualise the distribution of image and article types across the ISIS propaganda magazines Dabiq and Rumiyah, and reverse image searches identifying webpages on which these images have reappeared. Furthermore, machine learning was applied to detect the language of these webpages and to classify the category and tonality of English webpages (see figure below). Next steps will involve applying theories from multimodal analysis in further automating the analysis of violent extremist images, text, context and their relations.

An interactive visualisation of the webpage classifications is available at

image showing blood leakage

Working over at the Curtin Health Innovation Research Institute (CHIRI), our group has a specialised interest in exploring the relationship between blood brain barrier (BBB) leakage and cognitive decline. To do this, we collect images in mouse models using interventions that impair, restore, and/or maintain BBB function. The images are generated by fluorescence staining of brain slices for the protein immunoglobulin G (IgG). This enables visualization of brain capillary structures and leakage of IgG out of the capillary under a microscope. In our images, capillaries present as distinct branch like structures with a relatively well defined trajectory, while BBB-leakage presents as diffuse fluorescence with little definable form (see figure below).

Human derived quantitation of these images for BBB-leakage is time consuming, taking weeks for a researcher to classify the data obtained within a single study. Moreover, while consistency in classification amongst multiple people is relatively high, differences between raters and images is an unavoidable occurrence due to natural human fluctuations in attention and strategy. Our intention was therefore to collaborate with the CIC specialists to explore and ultimately implement a machine learning (deep learning) model for automated image processing and quantitation of BBB leakage in our immunofluorescent data. So far, we have a highly promising model that can classify 1,000s of images in seconds and the quantitations correlate strongly with human aided quantitation. This speed can be compared to the human time to complete this task of weeks to months. Our next steps are to stress test the model with further validation on different experiments to hopefully improve the quantitation ability of the model.

Image of computational fluid dynamics example used by BP's oil refinery in Kwinana, Western Australia

Researchers in the institute have used advanced computer modelling to improve the performance of large process plants. Working closely with BP’s oil refinery in Kwinana, the research team developed advanced models that were used to revamp the refinery systems to attain the optimal mix of steam, catalyst and hydrocarbons inside a fluid catalytic cracking unit. The changes have saved the refinery hundreds of thousands of dollars in steam usage and hydrocarbon recovery.

Computer generated image of the ITER fusion reactor

Institute researchers are undertaking computationally expensive atomic scattering simulations that are being used in the design of the ITER fusion reactor. These calculations are necessary to  understand the electron density and energy losses in the fusion plasma.

Image of a computer generated report on Value-at-Risk (VaR) forecasts

Researchers in the institute investigated the information content in the Value-at-Risk (VaR) forecasts reported by the Authorised Deposit Institutes (ADI) within Australia. The research was commissioned by the Australian Prudential Regulatory Authority (APRA) and required extensive computation for both information identification and extraction. The results have serious implications on risk management practice in Australia and its impacts are currently being reviewed by APRA.

Image of a computational fluid dynamics simulation CETO5 wave energy power unit

Computational fluid dynamics simulations undertaken by institutional researchers, have been used in the design and validation of devices for two WA based wave energy companies: Carnegie Wave Energy and Bombora Wave Power.

Read more about the devices being used by Carnegie Wave Energy.