My group specialize in bioinformatics, applied statistics, and deep neural networks focusing on applications and research within clinical medicine and forensic science.
We work extensively on high-dimensional data analysis especially within metabolomics (including LC-MS, DESI imaging, Maldi-TOF). We aim to understand complex biological systems and disease mechanisms and apply ML techniques on top.
We develop novel machine learning techniques for medical diagnostics and forensic applications ranging from automated tissue removal from 3D scans to molecular fingerprint imaging but also presymptomatic disease diagnosis. As a side-effect we also develop new experimental designs for different data types to decrease noise or alternatively be able to identify noise.
One of our strengths are working in together with external partners to deliver research results that are implemented in the "clinic". This requires extreme focus on analysis workflows to ensure that results/methods are reproducible and that our methods will work on new samples/lab machinery. We continuously analyze and evaluate results to identify possible data leakage when doing machine learning. This helps to avoid chasing results that later turn out to be results of overfitting or simple biases.
I collaborate with both the Department of Forensic Medicine here at AU, but also groups at Sheffield, Linköping, and Nice (and I'm open to more collaborations).
Besides research, I teach MSc courses in High Performance Computing (computer cluster use), Data Science, and Machine Learning.
