How to make machine learning work in medicine

 

In this episode, Karsten Borgwardt gives insights into his current research projects. In two examples, Karsten explains how machine learning algorithms are being developed specifically in the medical sector. Learn about the importance of massive amounts of data, and how this data is used to train a machine learning model to predict the probability of a sepsis of individual patients. Understand the various challenges, the importance of big datasets, the importance of validation, and why it’s a huge undertaking to make machine learning in medicine work. Apart from the Sepsis Study hear how Karstens career started in plant genetics and how the field has grown in importance.

Prof. Karsten Borgwardt studied Computer Science at the Ludwig-​Maximilians-Universität in Munich (LMU) and Biology at the University of Oxford. In 2007, he completed his PhD thesis on ''Graph Kernels'' at LMU, followed by a postdoctoral stay at the University of Cambridge. After his time as group leader at the Max Planck Institutes in Tübingen and as Professor at the University of Tübingen, Karsten moved to ETH Zürich, where he is now Full Professor of Data Mining. He is currently also deputy chair of the Department of Biosystems Science and Engineering at ETH Zürich, and co-leads the “Personalised Swiss Sepsis Study”. Additionally, he is the scientific coordinator of a Marie Curie Innovative Training Network on Machine Learning in Medicine. During his career he received numerous awards and recognitions, e.g. being ranked among the "25 Germans who will shape the next 25 years" and among the "Top 40 under 40" in Science from Germany.

 

Read about the antibiotic resistance example in this episode in a recent publication in nature medicine.