Sie befinden sich hier
AI-based Models for multi-omic data analysis and automated evaluation
In the analysis of few-cell regions in tissue samples, spatially resolved imaging has proven to be a suitable analysis method. At the same time, purely statistical analysis of single high-dimensional data has reached its limits, making the use of machine learning techniques promising. So far, more classical learning methods, such as Support Vector Machines (SVM), have been used.
In this project, advanced learning techniques will be explored for both image description and detection of relevant image regions in high-dimensional data. In particular, deep learning techniques will be applied. These methods have helped related areas, such as the fusion of high-dimensional data sources or object detection, to achieve a breakthrough. At the same time, these deep learning methods are highly susceptible to measurement inaccuracies and require a large amount of annotated (labeled) data. Therefore, this project aims to characterize and model measurement uncertainties so that the neural networks used can generalize through data augmentation. At the same time, novel methods will be implemented to compensate for the small amount of labeled data.
- Rishav; Schuster, R.; Battrawy, R.; Wasenmüller, O.; Stricker, D., ResFPN: Residual Skip Connections in Multi-Resolution Feature Pyramid Networks for Accurate Dense Pixel Matching. IEEE International Conference on Pattern Recognition (ICPR), 2020.
- Rabe, J. H., A Sammour, D., Schulz, S., Munteanu, B., Ott, M., Ochs, K., ... & Hopf, C., Fourier transform infrared microscopy enables guidance of automated mass spectrometry imaging to predefined tissue morphologies. Scientific reports, 2018.
- Thomas, S. A., Jin, Y., Bunch, J., & Gilmore, I. S, Enhancing classification of mass spectrometry imaging data with deep neural networks. IEEE symposium series on computational intelligence (SSCI), 2017
- machine learning, deep learning
- data augmentation
- high-dimensional data analysis
- Very good master's degree in engineering or natural sciences or comparable
- Extensive experience in programming, especially under Python
- Theoretical knowledge in the area of Artificial Intelligence and Deep Learning as well as practical experience with common libraries in this area (e.g. PyTorch, TensorFlow)