Nico Karssemeijer is professor of Computer-Aided Diagnosis. He studied Physics at Delft University of Technology and graduated at the Radboud University Nijmegen, department of Medical Physics. In 1989 he joined the Department of Radiology of the Radboud University Nijmegen Medical Center, where he formed a research group in computer aided detection (CAD). His professorship is in the Faculty of Science of the Radboud University in the section Intelligent Systems of the Institute for Computing and Information Sciences iCIS. He is Associate Editor of IEEE Transactions on Medical Imaging, and member of the Editorial Boards of Physics in Medicine and Biology and Medical Image Analysis. In 2012 and 2013 he will be symposium chair of SPIE Medical Imaging, while previously he chaired IWDM98 and IPMI 2007. Nico Karssemeijer was closely involved in the development of the R2 ImageChecker, the most widely used CAD system to date, and is co-founder of Matakina, Ltd. (Wellington, New Zealand), a company that develops technology for quantitative mammography.
Talk: “Computer aided detection in medical screening”
Fred Hamprecht develops and applies machine learning methods to solve challenging problems in bioimage analysis. He is particularly interested in active or weakly supervised learning. His group puts great emphasis on the development of user-friendly software (such as the http://ilastik.org framework) that can be used by experimentalists.
Fred defended his PhD at the Swiss Federal Institute of Technology (ETH) in 2001. Since then, he has been a Professor at the University of Heidelberg, where he has co-founded the Heidelberg Collaboratory for Image Processing. This institute is jointly funded by multiple companies and the German Excellence Initiative, and is considered a role model for “industry on campus” research partnerships. Fred is also a Visitor to the HHMI Janelia Farm Research Campus, and still thinks of science as the greatest job on earth.
Talk: “Learning in bioimage analysis”
Bioimage analysis offers a number of intriguing and relevantopen problems to which machine learning and computer vision can make a contribution. I will present two grand challenges, and current work towards solving these. Firstly, I will introduce the partitioning problem that is encountered in the quest for the connectome, that is, the wiring diagram of the brain. Our work shows that structured learning using a structured loss function can, when complemented with multicut consistency constraints, yield improved partitionings that are better than the current state of the art [ECCV 2012 and current work]. Secondly, I will show how tracking is used to try and follow each and every cell in a developing organism. On the technical side, I will lay out how the tracking of an unknown number of divisible targets can be formulated as a structured learning problem, and how learning from only partial annotations can beachieved [NIPS 2011, ICML 2012, ECCV 2012]. Finally, usable software is required to bridge the gap between computer vision conferences and work in a wetlab. I will demonstrate the utility and feasibility of interactive machine learning on vast data sets using the http://ilastik.org framework.