Celebrating the defence
Image:
Photo: Changkyu Choi

Celebrating the defence

Excellent PhD defense in the Visual Intelligence research centre

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.

Excellent PhD defense in the Visual Intelligence research centre

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.

Dr. Stine Hansen (Center) with family, committee and supervisor. Photo: Changkyu Choi

Evaluation Committee

  • Professor Alexandros Iosifidis, Department of Electrical and Computer Engineering, Aarhus University, Danmark (1. Opponent)
  • Professor Kjersti Engan, Institutt for data og teknologi, Universitetet i Stavanger (2. Opponent)
  • Associate professor Shujian Yu, IFT, UiT (internal member)

Supervisors

  • Professor Robert Jenssen, IFT, UiT (main supervisor)
  • Associate professor Stian Normann Anfinsen, IFT, UiT (co-supervisor)

Summary of Stine Hansen's thesis work:

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.

Full Thesis Here

1

Latest news

Visual Intelligence Annual Report 2024

April 2, 2025

The fifth Visual Intelligence annual report, showcasing the centre's activities, results, staff, funding and publications for 2024, is now available on our web pages.

uit.no: UiT er vertskap for landsdekkende KI-konferanse

April 1, 2025

I juni møtes det norske KI-miljøet i Tromsø for å presentere ny forskning og diskutere nye retninger innen feltet. KI-forskere inviteres til å delta og vise fram forskningen sin under konferansen.

Successful Industry Pitch Day at UiT

March 20, 2025

Visual Intelligence and the Digital Innovation Lab invited industry professionals to present ideas for master's projects to computer science and machine learning students at UiT The Arctic University of Norway.

Dagens Næringsliv: Norges eldste fagmiljø innen KI

March 18, 2025

Kunstig intelligens (KI) endrer måten vi løser komplekse problemer på. Ved UiT Norges arktiske universitet leder professor Robert Jenssen Visual Intelligence, et senter for forskningsdrevet innovasjon som utvikler neste generasjons KI-metoder.

Visual Intelligence at the UiT Open Day in Tromsø

March 13, 2025

Visual Intelligence researchers had the great pleasure of talking to high school students about the AI study programme at UiT The Arctic University of Norway during the UiT Open Day‍‍.

Visual Intelligence represented at CuttingEdgeAI seminar

March 11, 2025

Director Robert Jenssen represented Visual Intelligence at the CuttingEdgeAI seminar "KI anno 2023: I offentlighetens interesse?" at the University of Bergen on March 7th.

forskning.no: Derfor fungerer KI dårligere på kvinner

March 8, 2025

Det hender at kunstig intelligens behandler menn og kvinner ulikt. Hvordan skjer dette? KI-forsker Elisabeth Wetzer forklarer hva som ligger bak skjevhetene i teknologien (Popular science story in forskning.no and sciencenorway.no)

Visual Intelligence at TEKdagen 2025

February 11, 2025

We had the pleasure of talking to students about the exciting career opportunities at Visual Intelligence and UiT The Arctic University of Norway during TEKdagen 2025

School visit from Breivang upper secondary school

February 5, 2025

Last week, we welcomed students from Breivang upper secondary school to a full-day practical session on AI and programming at UiT The Arctic University of Norway

Successful course on collaborative coding and reproducible research

January 30, 2025

Visual Intelligence and UiT researchers organized a special curricular course on collaborative coding and reproducible research at UiT The Arctic University of Norway.

Insightful talks on deep learning-based sea ice forecasting at UiT

January 16, 2025

Andrew McDonald, a PhD student from University of Cambridge and the British Antarctic Survey, presented his work on sea ice forecasting with diffusion models, as well as the downstream applications of those forecasts‍, at Visual Intelligence in Tromsø.

A very successful Northern Lights Deep Learning Conference 2025!

January 14, 2025

Deep learning researchers from 27 different countries congregated at the 8th edition of the NLDL conference. - I have seen several interesting posters and talks that have opened my mind about different paths to explore in my own research, says NLDL 2025 attendee Tiago Ramos.

NLDL 2025 featured on NRK: the Norwegian national news

January 8, 2025

NRK, the national broadcasting company in Norway, spoke to Visual Intelligence researchers Robert Jenssen, Suaiba Amina Salahuddin, Michael Kampffmeyer and Elisabeth Wetzer about this year's Northern Lights Deep Learning Conference in Tromsø, Norway.

uit.no: Unge professor Kampffmeyer

January 2, 2025

In record time, principal investigator Michael Kampffmeyer became a professor in machine learning at the young age of 32. Read more about his academic journey and work in the latest UiT Researcher Portrait

Happy Holidays from Visual Intelligence!

December 24, 2024

2024 has been a year full of exciting events and accomplishments, and we look forward to continuing our journey on researching the next generation of deep learning methodology in 2025!