Projects
Bioimaging
- Classification of Vibrational Spectroscopy Images. In recent years, several studies have demonstrated the potentioal of infrared spectroscopic imaging for histopathologic recognition of certain diseases such as cancer. Typically, these studies rely on statistical classifiers such as artifical neural networks that are trained on the basis of reference spectra obtained by annotation through a human expert. In this scenarioe, this we work on different aspects and approaches to classify image spectra. This involves the whole computational pipeline from web-based annotation tools to state-of-the-art classifiers and feature selection approaches. In the long run, we also aim to incorporate novel paradigms such as collective classification to improve classification performance.
- Algorithms for tracking cells in microscopy movie sequences. Tracking cells in microscopic movie sequences is an easy, but labor intesive task for a human operator. Doing this with the help of computational methods is still a challenge that we approach by newly developed algorithms based on our novel approaches of morphological cosegmentation and tree-assignments.
- Peak Finding in in vivo fluorescence cytometry data. Currently in an early planning stage, we aim to develop reliable denoising algorithms to quantify fluoresecently labelled cells passing by an in vivo flow cytometer.
- Homology Patterns of non-coding RNAs. Understanding conservation of sequence and structure of non-coding RNAs is key to understand their evolution. We turn our insights in these conservation patterns into improved algorithms for the annotation in genomes.
non-coding RNA
