Computer Vision News - February 2023

40 Congrats, Doctor Stine! Most current methods for machine learning-based medical image segmentation are supervisedmodels that require large amounts of fully annotated images. However, obtaining such datasets in the medical domain can be difficult and expensive due to the specialized expertise and resources required to generate them. 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. When working in the limited supervision paradigm, exploiting the available information is key. During my PhD, I focused on addressing this challenge by developing new machine Stine Hansen recently finished her PhD with the UiT Machine Learning Group and SFI Visual Intelligence at UiT The Arctic University of Norway. Her research aimed at developing machine learning-based models for data-efficient medical image volume segmentation that only require limited supervision. She now works as a researcher at UiT, where she will continue her research on machine learning and medical image analysis. Congrats, Doctor Stine! Figure 1: Anomaly detection-inspired network for few-shot medical image segmentation.