Dr. Darko Stern will give his trial lecture in connection with the evaluation for an Associate Professor position of Image Analysis and Pattern Recognition / Machine Learning in the DSB group.
Title
Machine Learning in Medical Image Analysis: Integration of Spatial Configuration into Deep Neural Network
Abstract
In many medical image analysis applications, only a limited amount of annotated training data is available, which makes training of deep neural networks challenging. In my talk, I will discuss an architecture that learns to split the anatomical landmark localization task into two simpler sub-problems, reducing the need for large training datasets. Adapting the approach to the segmentation task, we won the MICCAI challenge for Multi-Modality Whole Heart Segmentation in 2017 as well as the MICCAI challenge for Large Scale Vertebrae Segmentation in 2019 and 2020. I will conclude my talk with a discussion on the uncertainty estimation of prediction by revisiting the landmark localization task.
Short biography
Darko Stern is a senior researcher at the Medical University of Graz and an adjunct lecturer at the Technical University of Graz, Austria. He obtained his Ph.D. degree in 2012 from the University of Ljubljana, Slovenia. Darko was a Marie Curie Fellowships researcher at the Technical University of Graz in the period 2013-2015. His research is on applied machine learning in the field of medical image analysis with a specific interest in datasets with limited annotations. Darko has published more than 50 peer-review papers in top-rated journals and conferences in the field. He was a runner-up for the best paper award of the Medical Image Analysis special issue of MICCAI 2018. He is awarded reviewer for several major journals and conferences in the field of medical image analysis and a winner of several MICCAI competitions.
Context
The candidate is invited to give a lecture to be suitable to be given in one of the courses IN5400 Machine Learning for Image Analysis or IN5520 Digital Image Analysis at master level.
Zoom connection
Please navigate to
to access the Zoom login information (requires login)