- Nicolas Rebach, University Hospital Cologne

Title: Advancing Immune Checkpoint Inhibition Therapy through Multimodal Learning Methods

Abstract:

The introduction of immune checkpoint inhibitors (ICI) in 2011 has advanced cancer immunotherapy. The therapy marked a shift in cancer treatment options by reactivating the host’s immune system to target cancer cells, leading to improvements in overall survival and tumor progression-free survival in patients. While ICI rapidly became a mainstay of treatment for many cancers, it is still unknown which patients specifically benefit from therapy and which are at risk of immune-related adverse events affecting healthy organs. Therefore, ICI therapy is accompanied by novel opportunities and unique challenges that must be considered during the treatment period. This makes it crucial to identify patients who are the most likely to benefit from it. In our ongoing research, we aim to expand and deepen the understanding of ICI in an interdisciplinary and multimodal fashion. Multimodal data was collected by the University Hospital Cologne and offers a wide range of research possibilities from exploratory data analysis to the application of Machine Learning (ML) and Deep Learning models to analyze how the disease expresses itself in different modalities. By integrating multimodal data, e.g., through fusion or graph-based methods, we aim to identify patterns that are not necessarily specific to individual cancer types. Currently, we are working on the data integration of different modalities and endpoints. Patient therapy response patterns have been analyzed using the iRECIST criteria to form labels for ML tasks. Furthermore, initial ML experiments have been conducted, aiming to classify patients’response to the treatment based on imaging, body composition, and laboratory markers.

Short Bio:

My name is Nicolas, and I am a first-year PhD student at the Institute for Diagnostic and Interventional Radiology in the working group of Liliana Caldeira at the University Hospital Cologne. My research focuses on multimodal learning methods in the field of immunotherapy. During my bachelor studies at TH Cologne in business information systems, I became increasingly interested in data science, especially how data-driven methods can be applied in the healthcare domain. Following my bachelor’s studies, I pursued a Master’s degree in Data and Information Science at TH Cologne, during which I completed an internship in my current working group and gained initial experience in medical AI. Together, we worked on classifying KRAS gene mutations using radiomics features extracted from CT images. Afterwards, I wrote my Master’s thesis on applying Large Language Models to structure radiology reports. In my initial PhD research, I have mainly focused on supervised ML techniques for therapy response prediction. Additionally, I am actively collaborating on an industry project with TH Cologne to optimize predictive maintenance in energy cable diagnostics.