Biomedical Data Science
Big medical data and machine learning

In the medical field, there is a vast treasure trove of data (such as registries, diagnostic devices, health apps, etc.) waiting to be utilized—of course, with careful consideration of methodological principles and "good statistical practice"!

In the context of big data—but not limited to it—methods from machine learning or artificial intelligence are often employed. In machine learning/AI, modern algorithms generate statistical models based on training data, which are then thoroughly tested. Ideally, these algorithms can identify patterns and relationships in the data. Such methods can be applied to the analysis of demographic and clinical routine data, or for tasks like automated diagnosis from imaging data (e.g., "image classification/image regression") and recognizing structures in those images (e.g., "medical image segmentation"). We utilize a range of modern techniques from deep learning and broader machine learning, such as support vector machines, random forest models, Bayes classifiers, classification trees, and many others. Additionally, we have expertise in developing multivariable prediction models, methods for constructing and evaluating clinical scores, and analyzing clinical routine data.

Here are a few typical examples from the clinics:

• Urology: early detection of prostate cancer
• Dermatology: neural networks predicting the likelihood of melanoma occurrence
• Predicting epileptic seizures or algorithm-supported automated epilepsy diagnostics
• Identifying a “focal onset zone” of an epileptic seizure
• Predicting drug-related complications
• And many more…

The technical possibilities are vast, but our primary focus is on a methodologically sound approach that takes clinical needs into account.