Machine Learning and Data Science Group
Data science is one of the fastest growing fields in science. Its applications are numerous and have led to advances in many areas of the scientific spectrum. One of the earliest applications for data science and specific applications of artificial intelligence (AI) such as machine learning was radiology reports. Essentially, the question was whether it was possible to train an AI system with large amounts of radiological image data so that it would learn and subsequently be able to find these patterns in new images. This proved to be no easy task. Nevertheless, in the late 2000s and early 2010s, there were already assistance systems that could help find patterns in images that radiologists might have missed, but which were still far from being image analysts in their own right. This was one of the first useful interactions between humans and AI in the real world. Since then, a lot has happened in this field, and numerous methods, from convolutional neural networks to variational autoencoders, have made their way into the field. Deep learning algorithms in particular have led to major advances in automatic image analysis. The field itself has also grown over the years and has spawned sub-areas such as computer vision and radiomics.
Here at the University Hospital of Cologne, data science in radiology is an important area of research. We don't see it as a replacement for the radiologist's work, but as a support and an increasingly useful clinical tool. This is why we have greatly expanded our research department for data science in recent years. It is now home to four different research groups working on different applications of data science and AI in radiology:
- Our general data science group, which works on new methods for AI in radiology, radiomics, image segmentation and classification
- Our special group for artificial intelligence in neuroradiology
- Our interdisciplinary Group for Applied Imaging Applications of Artificial Intelligence in Medicine, which aims to combine AI imaging methods in different disciplines of the medical field.
