Post-doc and PhD positions for deep and reinforcement learning in medical images

I haven’t written about my own work much on this blog, but since we have some positions available I thought I would talk a bit about some of our research, and a bit about why you might want to come work in Adelaide.

First things first, the post-doc position.

Jobs, jobs, jobs

We were recently funded by the ARC to dive deeper into what we called “precision radiology” in the first paper of my PhD program. The main idea behind this research is that individual variation in health is more complex than we can currently deal with in clinical practice. We typically only diagnose diseases that have symptoms, and by this stage they are usually quite advanced. By looking at the body directly using high resolution medical imaging, we want to detect and quantify the burden of undiagnosed, pre-clinical disease. In doing so, we hope we can tailor treatments to subgroups and even individuals (and in doing so provide the personalised treatment that genomics promised but has struggled to deliver).

A post-doc coming to work with us will be working on this task, exploring the technical challenges specific to dealing with medical images, mostly using deep and reinforcement learning. The challenges are quite varied, ranging from the technical (like how to manage enormous, imbalanced datasets, finding the needles in haystacks), to the statistical (how do we reliably evaluate and validate the performance of these systems), to the clinical (how do we apply these systems to patients). While a CS post-doc would mostly focus on the first two, exposure to all of these questions is a really important part of being able to do medical image analysis.

This is because medical work is very interdisciplinary. To work in the field you need to know a little about a lot of very different disciplines, at least enough to be able to discuss and collaborate with a range of experts from a variety of backgrounds. We work as a team of doctors and med students, epidemiologists and statisticians, computer scientists and engineers. It is a great environment, and certainly one I have found to support the my own growth and skill development.

Our core team includes:

  • Gustavo Carneiro would be the primary supervisor of the post-doc and PhD. He is well known in the field of medical image analysis and he is one of the core organisers of the deep learning workshop at MICCAI.
  • Lyle Palmer is a Professor of genetic epidemiology and a statistician. His CV is frankly ridiculous, and Scholar says his h-index is 84, with over 30,000 citations. Despite all of his accolades and achievements, he is an active, hands-on member of our team and a great resource for when deep learning alchemy runs headlong into the rigorous world of biostats.
  • Andrew Bradley is a Professor of electrical engineering, and the Assistant Dean of Research at the Queensland University of Technology. He is specialised in medical image analysis, and among other things is an expert on the assessment of machine learning models. He is the source of most of my knowledge on ROC curves and other performance metrics. Again, despite the distance involved, he is always available to provide advice, run experiments, or have a chat.
  • Me. I write blogs and do some researchy things 🙂

We also have a PhD spot available, which comes with a 3 year scholarship (we do 3 year PhDs in Australia). Essentially the PhD will be on the topics already discussed, covering deep learning and reinforcement learning applications in medical imaging with a focus on precision radiology tasks. You would be likely to publish papers in conferences like MICCAI, CVPR, ICCV, and so on. It is likely you will also be involved in medical publications in high-end journals. Some of the more theory-minded PhDs with Gustavo have published in venues like NIPS too; he certainly enjoys supervising that sort of work.

Of course, in both the PhD and post-doc roles there will be significant scope to tackle topics that interest you. The areas mentioned are just starting points.

That is all well and good, but any Australian reading this is probably asking one question: why Adelaide?

The festival state

Adelaide is the small capital city in one of Australia’s smallest states, with barely over a million people in the metropolitan region. We have nice beaches, great food and wine, lots of green space, short commutes, and a low cost of living, but it is true that most Aussies don’t think of Adelaide as a go-to destination.

The strange thing is that Adelaide is actually one of the best places in the world to be working in AI, particularly as it relates to vision. We punch so far above our weight is it pretty absurd. This is mostly due to the Australian Centre for Visual Technologies, which was the premier research group in the Computer Science department of Adelaide Uni. This group has over 60 members (including academics, research staff, and students), and according to CSRankings is second in the world in computer vision over the last decade (by the metrics they use, which are mostly based on the number of publications at top venues).

csrankings cv

ACVT researchers also regularly ranked in the upper portion of various computer vision competitions, including ImageNet and VQA.


For a small town, Uni Adelaide is a great place to work on computer vision problems.

It is only getting better too. Last year the South Australian government gave the head of the ACVT over 7 million dollars to create the Australian Institute for Machine Learning, which replaces the ACVT at Uni Adelaide and will be the centre of a new technology and innovation agenda. Obviously the institute has high end GPU clusters and so on, for all your training needs.

Furthermore, Adelaide is incredibly well suited to medical AI work. We have centralised data governance, particularly in medical imaging, which makes access and ethics approvals much easier. We have one of the most mature data linkage systems in the world, meaning clinical data, government data, even genomic data is often just a request away. We are currently working towards including image data in this linkage system, which would make us one of the best integrated imaging biobanks anywhere. Our own team has a lot of images to work with on our current projects, and more at our fingertips as needed.

The government has just spent an enormous amount of money to make Adelaide a world leader in medical research too. We have a new flagship hospital (the second most expensive hospital in the world), a shiny research centre, and a new medical and health sciences campus. Lots of scope for collaboration with clinicians from a wide range of disciplines.

adelaide medical precinct

So, essentially, it is a really cool place to be. We don’t have the name recognition of some of the famous US and European centres, but we have a great team, great infrastructure, great medical resources, and a great environment to work in.

So do check out the ad if you might be interested in the post-doc position, or this ad for the PhD spot.

Anyway, that is enough self-interested blather, hopefully my next blog post isn’t far off 🙂



4 thoughts on “Post-doc and PhD positions for deep and reinforcement learning in medical images

  1. Brilliant, exciting stuff! Having moved to Adelaide 10 years ago, I can attest it’s a wonderful city to work and play. And seems to get better every year. ‘Goldilocks’ size for innovation and collaboration (and potentially setting up biobanks…).


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s