The program will be available shortly. Please check back later.
Suaiba Amina Salahuddin, UiT, PhD-student i Visual Intelligence, UiT Machine Learning Group.
This presentation will start with a brief background on my educational history and experience. I will then present some preliminary work on segmentation of medical images leveraging so-called few-shot learning, which in this case is based on learning foreground and prototypes which form the basis for pixel classification. The new idea is to incorporate a self-guidance module in the generation of prototypes based on initial errors, to achieve boosted segmentation with primary and auxiliary prototypes. The proposed framework´s performance on cardiac MR and liver CT segmentation will be discussed. The presentation ends with an outlook on potential future research focus.
Magnus Oterhals Størdal, UiT, PhD-student i Visual Intelligence, UiT Machine Learning Group.
In this presentation I’ll briefly present my educational background. The presentation is divided into two parts. The first is to present my previous work with DR data. We got a DR dataset collected from a general sample of the population in the municipality of Tromsø. The dataset presented multiple challenges such as the information of a single retina being spread across multiple images, noisy data, etc. I will present the proposed multistream CNN network which can fuse image-level features into retina-level features, and still retains a network structure which is compatible with most domain adaptation, and post-hoc explainability methods.
For the second part of the presentation, I’ll be presenting my future research direction in the field of DR screening.
Two Visual Intelligence PhD fellows present their research topics
This seminar is open for members of the consortium. If you want to participate as a guest please sign up.