| layout | group |
|---|---|
group |
mml |
Medical Machine Learning (MML) The research group Medical Machine Learning works on developing and deploying cutting-edge machine learning methods with the goal of making a meaningful difference for patients, doctors, and hospital staff. To this end, and in collaboration with a digital-forward clinic administration, we continue to build on a SMART hospital information structure that provides access to real-world medical data. Strong funding and state-of-the-art equipment support this effort. One focus of our group lies in the exploration of unsupervised learning paradigms for recognition of oncologically relevant patterns in large and complex data.
In medical research a common approach is from bench to beside – use insights gained in laboratory experiments for new ways to treat patients. In an analogous approach, we aim at bringing our algorithms to the point-of-care, i.e. translating them into clinical practice.
Affiliations
The TIO research group is part of the Cancer Research Center Cologne Essen (CCCE). In addition, Prof. Kleesiek is PI in the German Cancer Consortium (DKTK) and at the Helmholtz Information & Data Science School for Health. There is a close cooperation with the German Cancer Center (DKFZ) for developing the Joint Imaging Platform (JIP) for distributed data analysis and federated learning.
Selection of Current PhD Projects
Digital Tumor Signatures
In this project we seek for the identification of digital signatures / biomarkers that allow for the quantification of oncological disease progression. Machine learning methods (supervised and unsupervised approaches) will be used to extract characteristics from medical data, which in turn can be related to outcomes and measurable target variables, comprising laboratory parameters, biopsy results, but also endpoints such as Progression Free Survival (PFS) or Overall Survival (OS). The aim is to establish a prediction model that can be prospectively evaluated for added value in diagnostics and therapy.
Holistic Integration
This project aims at developing methods for integrating multi-modal medical data. Next to imaging data from radiology and pathology, this includes all kinds of structured and unstructured additional sources, e.g. laboratory values, gene profiles and clinical reports. For instance, contents can be extracted from text data using NLP techniques. We conjecture that multi-modal models will outperform currently available approaches. Integration can be used for diagnostic and therapeutic predictions as well as for the selection (filtering) of data. One use case is the virtually augmented tumor board, which uses algorithms that support interdisciplinary decision making for personalized medicine.
Opportunities
We currently offer several open research positions at various levels, including PhD and postdoc positions as well as a junior group leader position. We are also constantly on the look-out for interested and motivated student assistants, bachelor and master students.
We are seeking enthusiastic, self-motivated, and curious personalities, who strive to create meaningful impact with their work and enjoy working in a team of like-minded persons. You should have training in computer science, physics or in related fields and a solid foundation in machine learning.