Emerging artificial intelligence (AI) techniques based on machine learning, such as radiomics and deep learning, hold the promise to turn medical images (such as X-ray, CT, MRI, PET, Ultrasound) into objective and quantitative biomarkers, optimally guiding diagnosis and treatment. In the field of medical image analysis, machine learning has become the main workhorse in automating any image processing task, ranging from low-level operations like segmentation to high-level interpretations such as diagnosis, subtyping, and prediction.
In this seminar, Stefan Klein will show several examples of the aforementioned, illustrating how medical image analysis can play an important role in personalised medicine. He will explain the underlying technical principles, and highlight some of the infrastructural challenges encountered in this type of research.
Machine learning is a “data-hungry” approach: the computer needs vast amounts of clinically representative example data to learn models that generalise well to new data. Therefore, efficient workflows for data gathering, anonymisation, harmonisation, storage, sharing, annotation, processing and integration are necessary to accelerate personalised medicine research.
About the speaker
Stefan Klein is Associate Professor in Medical Image Analysis and is affiliated with the Biomedical Imaging Group Rotterdam (BIGR), Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam.
In 2002, Stefan received his MSc degree from the faculty of Mechanical Engineering at the University of Twente, Enschede. In 2008 he obtained his PhD degree at the Image Sciences Institute, UMC Utrecht/NL, for his research on optimisation methods for medical image registration.
He was co-principal developer of a widely used open-source software package for medical image registration, called Elastix (article cited >2500x), co-organiser of the CADDementia, TADPOLE, and KNOAP2020 grand challenges, general chair of the WBIR2018 conference, and serves as Associate Editor for the IEEE Transactions on Medical Imaging.
His current research interests include image reconstruction, radiomics, machine learning, and disease progression modelling. He is also active in setting up infrastructures for research and has for instance initiated a national research archive for medical imaging data, currently used by numerous multi-centre imaging studies in the Netherlands. The scientific publications he contributed to can be found here.