A conversation with Dr Jarrel Seah and Dr Jennifer Tang
Can you please tell us what is the aim of your project is?
The project’s aims are to use deep learning models to:
- Predict the growth and appearances of different types of brain tumours on MRI
- Distinguish between true progression and pseudoprogression
- Identify different molecular subtypes of brain tumours based on imaging appearances
- Discover new imaging phenotypes that correspond to prognosis and molecular subtypes
What goal outcomes are you hopeful for and what would that mean as an impact towards individuals and their families facing this diagnosis?
We hope to improve the management of brain tumours by giving surgeons and oncologists better tools to predict a patient’s clinical course, as well as to identify if current treatment is working or not. Furthermore, if this is successful, we are interested in probing the workings of these deep learning models to discover new imaging phenotypes and markers that these models may have learnt that predict prognosis.
Patients and their families in the future will have access to more information about their personal condition at an earlier stage, and their clinicians will be able to alter or personalise treatment based on the outputs of these deep learning models.
What progress have you made to date?
This is a large project which first and foremost requires a large and varied dataset. We have been working on collaborations between different institutions to acquire datasets under an appropriate ethics and governance framework. Furthermore, we have been setting up the necessary computing infrastructure at Monash University and Alfred Health to store and process this data, with the purchase of specialised deep learning hardware worth around AUD 60,000 so far.
Have you been able to leverage your work, knowledge or funding to attract any partnerships?
This funding has enabled us to attract Peter MacCallum Cancer Centre as a partner to contribute both data as well as relevant research and clinical skills.