Harry Dempsey
Harry Dempsey
Harry Dempsey

Harry Dempsey

‘Creating a turnkey solution to detect behaviour’

University of Melbourne, Vic
Florey Department of Neuroscience and Mental Health
Awarded 2023
Co-funded by the ‘Rob Henry Memorial’ PhD Scholarship

“Greater understanding of the neurons responsible for behaviours in mental health disorders may eventually lead to the development of better pharmacological therapies, which target these neurons.”

General Health PhD Scholarship

Researcher Profile

My name is Harry Dempsey and I grew up in Melbourne. In primary school, I loved making video games. In secondary school, I played the clarinet and piano and competed in lawn bowls and fencing competitions. During this time, I also fell in love with science which motivated me to study science, so I enrolled in a double degree of Biomedical Science and Physics at Monash University. I knew I wanted to become a researcher, but I was not sure which scientific field to pursue.

Over time, I became fascinated by artificial intelligence (AI) and machine learning. I noticed that problems, previously thought to be unsolvable, were cracked in the span of just a few years using AI. I love coding in the same way I enjoyed making video games when I was in primary school, so I decided to pursue a PhD where I will use machine learning to crack problems in mental health research.

Project Summary

In the study of almost all mental health conditions (PTSD, OCD, autism, etc.), researchers use cameras to record the behaviour of rodent models of these conditions. For example, to measure anxiety levels in these rodents, researchers may place them in an open field and find the time spent near the safety of a wall. Alternatively, to measure obsessive behaviour, researchers may place them in a cage with marbles and count how many marbles they bury. While these statistics are useful, these videos contain a wealth of information about rodent behavioural patterns. However, manually scoring behaviours is both time consuming and vulnerable to individual scoring biases. For this reason, this information is regularly discarded. If these data could be acquired simply and routinely, it would greatly enhance research into rodent models of mental health conditions.

I aim to create a web-based tool that imports videos of rodents and exports a summary of their behaviour over time. I will leverage state of the art deep learning and photo-realistic simulation tools. This will allow the detection of walking, sitting, rearing, grooming and may also detect subtypes of these behaviours. Importantly, I aim to make this tool accessible without coding expertise. I will also apply existing tools to find patterns of behaviour related to cravings for alcohol and analyse the social interactions between rodent mothers and infants following specific neuronal modifications.

Supervisors: Professor Andrew Lawrence and A/Professor David Dowe.

Support Us