PORTFOLIO / CLOSED-LOOP OPTICAL TECHNOLOGY FOR REAL-TIME POSE DETECTION
Closed-loop optical technology for real-time pose detection
| TYPE | Doctoral Research, UCL |
| YEAR | 2020-2024 |
| ROLE | Research Ideation Design Development Analysis Writing |
| TOOLS | Python Illustrator SolidWorks |
| PUBLICATION | Precision cutaneous stimulation in freely moving mice |
| CODE | Github repository→ |
CONTEXT
Somatosensation — our sense of touch, pressure, temperature, and pain — is essential for how we move through and experience the world. It's what lets you feel the warmth of the sun on your skin, the texture of a rough surface, or the slight shift in balance as you walk. In preclinical research, studying how somatosensation shapes behaviour is notoriously tricky, especially because it is hard to deliver touch-like sensations to animals that are moving freely.
As part of my PhD, I tackled this challenge by designing and building a new technology that could capture how mice respond to somatosensory input while they explored their environment naturally.
System build rendered in SolidWorks
This case study highlights selected data visualisations I designed to explore how behavioural patterns unfold as mice explore, quantifying exploration, and demonstrating the technology's ability to capture it.
Reconstructions of the mouse 'skeleton' when the mouse is turning.
Data Analysis Pipeline
Movement was recorded on video. Using a neural network trained to detect different parts of a mouse's body, I extracted the positions (coordinates) of each body part frame by frame. This transformed the video into 'keypoint' data that I could plot.
How To Read The Visualisation
I visualised the footprints of a mouse as it explored an arena using 2D histograms of paw coordinates. These footprints represent moments during walking when the mouse's paws are stationary and touching the ground, a key part of the walking (gait) cycle. While the histogram on the right captures the continuous motion of the mouse's tail in comparison to the footprints of the paws.
Visualisation of the movement of the left and right paw alternating in 2D space while the mouse walks.
I was interested in studying natural exploratory behaviours. I designed a maze environment, which leveraged the natural behaviours of mice as burrowing animals, inspired by the tunnel systems they create underground in the wild to shelter from predators.
How To Read The Visualisation
Water rewards were delivered through ports on the left or right handside of the maze. This visualisation shows a mouse's movement trajectory over an entire session in the maze from a bottom-up view. On the right, a single trial highlights the mouse running from the entry point of the maze to a reward port.
Transition matrices for two groups of mice show the mice predominately alternate between drinking water from the left and right reward ports (n = 4 for the group on the left and n = 3 for the group on the right).