Celebrating the defence
Stine Hansen defended her PhD thesis “Leveraging Supervoxels for Medical Image Volume Segmentation With Limited Supervision” on Dec. 16th at UiT The Arctic University of Norway.
The focus of Stine Hansens thesis is on the development of new ML algorithms for medical image segmentation with limited supervision. In particular, completely unsupervised solutions to lung tumor segmentation and solutions that only require a few labeled samples to perform organ segmentation (few-shot learning) are proposed. To exploit the data-specific opportunities of medical images, automatically generated supervoxels are exploited in various ways to solve key challenges related to the task.
.jpg)
Evaluation Committee
Supervisors
The majority of existing methods for machine learning-based medical image segmentation are supervised models that require large amounts of fully annotated images. These types of datasets are typically not available in the medical domain and are difficult and expensive to generate. A wide-spread use of machine learning based models for medical image segmentation therefore requires the development of data-efficient algorithms that only require limited supervision. To address these challenges, this thesis presents new machine learning methodology for unsupervised lung tumor segmentation and few-shot learning based organ segmentation. When working in the limited supervision paradigm, exploiting the available information in the data is key. The methodology developed in this thesis leverages automatically generated supervoxels in various ways to exploit the structural information in the images. The work on unsupervised tumor segmentation explores the opportunity of performing clustering on a population-level in order to provide the algorithm with as much information as possible. To facilitate this population-level across-patient clustering, supervoxel representations are exploited to reduce the number of samples, and thereby the computational cost. In the work on few-shot learning-based organ segmentation, supervoxels are used to generate pseudo-labels for self-supervised training. Further, to obtain a model that is robust to the typically large and inhomogeneous background class, a novel anomaly detection-inspired classifier is proposed to ease the modelling of the background. To encourage the resulting segmentation maps to respect edges defined in the input space, a supervoxel-informed feature refinement module is proposed to refine the embedded feature vectors during inference. Finally, to improve trustworthiness, an architecture-agnostic mechanism to estimate model uncertainty in few-shot segmentation is developed. Results demonstrate that supervoxels are versatile tools for leveraging structural information in medical data when training segmentation models with limited supervision.
María Castro Fernández is a PhD Candidate at the Research Institute of Applied Microelectronics at Universidad de Las Palmas de Gran Canaria, in Spain. She visited Visual Intelligence in Tromsø from January to June 2024.
AI tools wield enormous potential within the health sector. How do we ensure that such tools are safely implemented and used within this field? The event will take place on August 14th, 11:45-12:45 at Bankgården, Strandgaten 2-4.
Thea Brüsch is a PhD student at the Section for Cognitive Systems at the Technical University of Denmark (DTU). She visited the UiT Machine Learning Group from January to June 2024.
Visual Intelligence happily welcomes Lars Uebbing, who recently started as a doctoral research fellow at Visual Intelligence in Tromsø.
Det tar for lang tid å analysere bilder for å finne ut om en pasient har kreft i leveren. Nå har forskerne laget et system som gjør jobben mer effektivt (Norwegian news story).
We are happy to announce that a Visual Intelligence paper "The 3-billion fossil question: How to automate classification of microfossils" was featured in the scientific newspaper EurekAlert!
In the episode, Østmo was included in an Abels tårn expert panel to answer different questions about AI and deep learning.
Visual Intelligence organized a special curricular course on Generative AI at UiT The Arctic University of Norway this spring. The course covered Large Language Models, Generative AI for images, and Generative AI for audio.
Our sponsorship packages provide a number of benefits, such as registration and networking benefits for company employees, company and brand visibility, and more.
Associate professor and principal investigator Kristoffer Wickstrøm recently talked about a collaborative Visual Intelligence project between UiT and UNN on NRK Radio. The project aims to develop AI algorithms to detect liver cancer on CT images.
We congratulate Tollef Jahren for successfully defending his PhD thesis "Deep learning for detecting valvular events and suppressing reverberations in cardiac ultrasound" for the degree Philosophiae Doctor at Department of Informatics, University of Oslo.
.jpg)
Noen få tastetrykk er alt som skal til for å finne falske nakenbilder og videoer av norske kjendiser. – Ekstremt bekymringsfullt, sier KI-ekspert
Visual Intelligence congratulates Daniel Johansen Trosten for receiving the Dissertation Award of 2023 from Norwegian Artificial Intelligence Society (NAIS)!
Centre director Robert Jenssen represented Visual Intelligence at the marine conference Blått kompass.
The Visual Intelligence Graduate School (VIGS) recently gathered our Tromsø based researchers at Árdna - The Sami Cultural Center for a social event about exchange stays.
We congratulate professor Michael Kampffmeyer for receiving the "Early career investigator" award at the NORA Annual Awards!
Til tross for gode karakterer, søker svært få jenter seg til studier innen matematikk, naturfag og teknologi. En studie viser at det kan handle om lav selvtillit.
We invite submissions presenting new and original research on all aspects of Deep Learning for the Northern Lights Deep Learning Conference (NLDL) 2025.
I am eager to absorb knowledge from these experts and contribute to our collective understanding, says Filippo Bascetta, who is one of two new VI interns.
Listen to VI co-director Line Eikvil and research director Anders Løland (Norsk Regnesentral) talk about developments before and after the breakthrough of deep learning in the NR podcast, Sannsynligvis VIKTIG (Norwegian dialogue).