Advancing oncological PET imaging using machine learning
Algorithms with the capability to fuse information from different image modalities, such as PET, CT, and MRI, is of key importance in effective analysis of medical images. The UiT Machine Learning Group is heading one of the work packages in the 180° North project, which is tasked with developing novel methodology for extracting and combining information from different medical image modalities.
Paper published at European Conference on Computer Vison 202022nd September 2020
Our recent work entitled ”SEN: A Novel Feature Normalization Dissimilarity Measure for Prototypical Few-Shot Learning Networks” was accepted at the ECCV'20. The ECCV is a premier conference on machine learning where only the highest quality work is presented. The recent paper introduces a new dissimilarity measure designed for the task of few-shot classification, which aims at identifying unseen classes form just a handful of labeled examples.
Center for Research-based Innovation
The UiT Machine Learning Group will be heading a Center for Research-based Innovation, entitled Visual intelligence. Such a center is aimed at ensuring close collaboration between research and industry. The center will focus on developing cutting-edge technology for extracting information from images from many different domains such as medical and remote sensing imagery. The interested reader can find more information here, here and here.
Best Master's Thesis for 2019
Daniel Johansen Trosten, UiT Machine Learning Group, has been awarded best master's thesi at UiT for the year 2019. In his thesis, entitled "Deep Image Clustering with Tensor Kernels and Unsupervised Companion Objectives", Trosten developed new methodology for unsupervised machine learning for image data. The thesis has resulted in several publications. The interested reader can find more information here.
National Collaboration for Machine Learning in Cancer Research
Machine learning is an important component of a national research collaboration ("Kystsamarbeidet"), where the UiT Machine Learning Group plays a key role. Recently the group have been working on new methodology for combining different modalities in medical imagining. The interested reader can find more information here and here.
Detecting Skin Cancer Using Hyperspectral Images
Skin cancer is one of the most common types of cancer, and in some parts of the world, it is becoming increasingly more common. Most skin cancers are not life-threatening, but one particular type called malignant melanoma is known to be quite deadly. The UiT Machine Learning Group have recently explored the state-of-the-art in the application of hyperspectral imaging for melanoma detection. The interested reader can read more here.
Leading in Health and AI
The UiT Machine Learning Group continues its fruitful collaboration with the University Hospital of North Norway, and has recently received funding for two research projects focused on the development of understandable decision support systems for medical application. The interested reader can find more information here and here.
Fake News Detection
The UiT Machine Learning Group has been awarded a Facebook research grant to develop algorithms that can automatically detect "fake news" and logical fallacy in online discourse. Postdoctoral researcher Jonas Nordhaug Myhre will, in collaboration with researchers from the University of Cambridge of and the University of Bergen, use natural language processing and machine learning techniques to develop such detectors. The interested reader can find more information here or here.
International Best Paper Award
Karl Øyvind Mikalsen, UiT Machine Learning Group, has been awarded best article within the field “computerized clinical decision support". The award was issued by the International Medical Informatics Society for the paper "Using anchors from free text in electronic health records to diagnose postoperative delirium". Interested readers can find the full article here.
Mikalsen's article also garnered attention from Norwegian media, which highlighted its potential for identifying patients with a higher risk of experiencing postoperative complications. The full article can be found here.
Machine Learning and PET Imaging
A new collaboration between the PET research groups in Tromsø, Bergen and Trondheim, named "Kystsamarbeidet", has been granted 80 MNOK from Tromsø Forskningsstiftelse and Trond Mohn stiftelse. The UiT Machine Learning Group is an integral part of the project as developers of automatic analysis tools for the PET imagery. Interested readers can find the the full article here.
Northern Lights Deep Learning Workshop 2019
The UiT Machine Learning will be hosting the Northern Lights Deep Learning Workshop from the 10-11 January. Top researchers from across the globe gather to present their work and share ideas. Interested reader can find more information at the official webpage of the workshop or in this article here (Note: Only available in Norwegian).
Automatic Power Line Surveillance
Power companies face significant challenges with regards to monitoring the condition of their electricity grid, equipment and power lines. The Norwegian company Esmart Systems has initiated an innovative project on automatic surveillance of power lines through images and videos captured by drones equipped with cameras. Towards the goal of automatic surevillance, the UiT Machine Learning Group has been engaged to provide guidance with regards to the development of the computer vision model that will underpin the monitoring system. Interested readers can read more about our groups collaboration with Esmart Systems here. (Note: Only available in Norwegian)
Change Detection in Satellite Images
Change detection in heterogeneous multitemporal satellite images is an emerging topic in remote sensing. The UiT Machine Learning Group is leading a project on advancing the methodology for change detection in heterogeneous remote sensing images, with a particular focus on deep learning. Interested readers can read more about our project on change detection in heterogeneous remote sensing images here. (Note: Only available in Norwegian)
Credit Risk Assessment
Lending is the principal driver of bank revenues in retail banking, where banks must assess whether to grant a loan at the moment of application. Therefore, banks focus on different aspects to improve this credit assessment. Santander Consumer Bank has solicited support from the UiT Machine Learning Group in order to explore the potential for novel, automatic credit risk assessment through machine learning, and in particular, deep learning methodology. Interested readers can read more about our groups collaboration with Santander here. (Note: Only available in Norwegian)
UiT Machine Learning Group
The UiT Machine Learning Group combines basic research in machine learning with cutting edge applications for industrial applications. The group has experience and expertise in a number of machine learning research areas, in particular kernel methods and deep learning, and close connections with renowned universities around the globe. Interested reader can read more about the activities of the UiT Machine Learning Group here. (Note: Only available in Norwegian)