My Research Stay at Visual Intelligence: Thea Brüsch

By: Petter Bjørklund, Communication Advisor at SFI Visual Intelligence

Thea Brüsch is a PhD student at the Section for Cognitive Systems at the Technical University of Denmark. She is working with biomedical time series data from different perspectives, focusing mostly on self-supervised learning as a tool for learning powerful representations from unlabelled biomedical time series. Brüsch visited Visual Intelligence in Tromsø from January to June 2024.

What's your name and current affiliation?

My name is Thea Brüsch. I am a PhD student at the Section for Cognitive Systems at the Technical University of Denmark (DTU). I have been visiting the Machine Learning Group at UiT from January to June 2024.

Can you tell us about your PhD project?

During my PhD, I have been working with biomedical time series data from different perspectives. Most progress in deep learning has been made in computer vision and natural language processing with less emphasis on time series. However, time series data is important for many medical tasks, and assisted decision making could alleviate the burden from many doctors. In the beginning of my PhD, I have focused mostly on self-supervised learning as a tool for learning powerful representations from unlabelled biomedical time series. Recently, I have shifted my focus towards explainability for time series.

Why did you choose Tromsø and Visual Intelligence for your research stay?

In the spring of 2023, I was looking for a place to do my external stay. My supervisor already knew Robert Jenssen and suggested that Tromsø might be a good place. I was intrigued by the idea of living in Norway, since I have always loved the nature here. I also had a really good impression of the group based on their published work. I therefore came for a short visit in September 2023 to ensure that our interests aligned. During my visit, I met with a lot of people from the group and brainstormed ideas for projects. I wanted to shift my research focus to explainability, and it became clear that both Robert and Kristoffer had a lot of expertise in this areas and that our interests overlapped.

What has been the outcome of your research stay?

During my stay, I have had the opportunity to work with professor Robert Jenssen and associate professor Kristoffer Wickstrøm on masking based methods for explainability. This has resulted in two completed manuscripts, one with Kristoffer in the lead, and one with myself as first author, as well as a lot of ideas for extensions. I also became involved in a project with Suaiba and Magnus on visualizations using t-SNE methods. All have been very fruitful collaborations that I hope to continue building on.

How was life in Tromsø?

I came to Tromsø during the dark time in January and got to experience the northern lights. The spring was filled with a lot of skiing – both cross country and skitouring. And finally, summer time offered hiking under the midnight sun. The entire group has been extremely welcoming towards me and invited me to all sorts of social activities – something that I am extremely grateful for. I am leaving Tromsø with a lot of new friends and great collaborators. All in all, my stay has been everything that I wished for and much more!