In my work, I investigate algorithms aiding the decision process of radiologists for various diseases. Here, I mostly focus on deep learning-based approaches on medical imagery and face various problems occuring in real world scenarios such as low amounts of annotions and/or label noise. Thus, I am especially interested in weakly and semi-supervised approaches for medical image analysis, which manage to thrive in these scenarios.
Practical Course - CV:HCI
I currently supervise the practical course "Computer Vision for Human-Computer Interaction". Here, groups of three students each try to tackle various problems solvable by computer vision ranging from standard methods to SOTA deep learning algorithms.
If you are interested in the different types of problems we investigated in this course, please check the following:
- Research Assistant/PhD Student at Computer Vision for Human-Computer Interaction Lab, Karlsruhe Institute of Technology, since June 2019
- M.Sc in Computer Science at Karlsruhe Institute of Technology, 2016-2019
- B.Sc. in Computer Science at University of Stuttgart, 2013 - 2016
Open Master/Bachelor Thesis Topics
While I currently do not have any particular thesis topic in mind, if you are passionate about deep learning with applications in medical imaging and have a particular topic in mind - please send me an email with a few sentences about yourself and your idea.
Pulmonary embolism describes a blockage of an artery in the lungs affecting about 430,000 people each year in Europe. Due to the high mortality rate of pulmonary embolism accurate detection is crucial. In this collaboration with the university clinic Heidelberg, we are directly working with practicing radiologists on the integration of image-based AI algorithms to aid the process of detecting pulmonary embolims. For this task, we explore methods to visually aid radiologist by either improving image quality leading to easier manual detection by radiologists or the highlighting regions of high confidence of deep learning based algorithm.
Pose2Drone: A Skeleton-Pose-based Framework for Human-Drone Interaction. Zdravko Marinov, Stanka Vasileva, Qing Wang, Constantin Seibold, Jiaming Zhang, Rainer Stiefelhagen. 2021 29th European Signal Processing Conference (EUSIPCO). IEEE, 2021.[pdf]
Every annotation counts: Multi-label deep supervision for medical image segmentation. Simon Reiß, Constantin Seibold, Alexander Freytag, Erik Rodner, Rainer Stiefelhagen. Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR). 2021.[pdf]
Prediction of low-keV monochromatic images from polyenergetic CT scans for improved automatic detection of pulmonary embolism. Constantin Seibold, Matthias A. Fink,Charlotte Goos, Hans-Ulrich Kauczor, Heinz-Peter Schlemmer, Rainer Stiefelhagen and Jens Kleesiek. IEEE 18th International Symposium on Biomedical Imaging (ISBI). IEEE, 2021. [pdf]
Self-Guided Multiple Instance Learning for Weakly Supervised Thoracic Disease Classification and Localization in Chest Radiographs. Constantin Seibold, Jens Kleesiek, Heinz-Peter Schlemmer and Rainer Stiefelhagen. Proceedings of the Asian Conference on Computer Vision (ACCV). Oktober 2020 [pdf][code]
- A reporting and analysis framework for structured evaluation of COVID-19 clinical and imaging data. Salg, G.A., Ganten, MK., Bucher, A.M. et al., npj Digit. Med. 4, 69 (2021) [pdf]