Radiooncology / Radiobiology

HypoxRM - Imaging and Multiomics integration for tumour hypoxia characterization

This research project aims to enhance our understanding and improve therapeutic strategies for squamous cell carcinoma of the head and neck (HNSCC), a malignant tumour arising from various sites in the oral cavity and upper respiratory tract and, despite many treatment options, is still associated with a poor survival prognosis.

Radiation therapy remains a cornerstone of HNSCC treatment. However, therapeutic resistance frequently develops, partly due to hypoxia - regions of the tumour with insufficient oxygen supply. Strategies to sensitize hypoxic cancer cells to radiation, such as enhancing oxygen availability during treatment, have been explored. Interestingly, "reoxygenation" can also occur following radiation exposure, yet this process may paradoxically compromise treatment efficacy. This underscores the critical role of transiently hypoxic tumour regions in influencing patient responses to therapy.

It is now well-recognized that hypoxia is a highly dynamic phenomenon affecting virtually all cell types within the tumour microenvironment, presenting significant challenges for its accurate modelling in laboratory settings. Our primary objective is to bridge the gap between findings from in vitro and in vivo preclinical studies and the complex reality of hypoxia in cancer patients before and during treatment.

To achieve this, we are adopting innovative methodologies to assess oxygenation at the cellular level. Leveraging biobank resources (comprising collections of biological samples) and state-of-the-art imaging technologies, we aim to investigate individual cells and their microenvironment. Through this approach, we seek to elucidate how fluctuating oxygen levels influence cancer cell behaviour and treatment outcomes. Ultimately, our research aspires to pave the way for more effective therapeutic interventions for HNSCC in the presence of hypoxia.

Contact: Maria José Besso

Survival analysis of HNSCC through Histopathology Slides and AI

Together with our collaboration partner in Dresden (Division of Clinical Artificial Intelligence, TU Dresden, Prof. Jakob Kather), we are analysing which potential markers can be extracted from histology slides, particularly hematoxylin and eosin (H&E) stains. For this, foundation models are utilized as rich feature extractors of histopathological characteristics, in combination with attention-based models to weight areas of the images that are associated with survival. These models are either trained on specific molecular characteristics (e.g., omics modalities) or on available clinical data (e.g., follow-up information). With H&E stains being widely available in clinical practice, this method offers a cost-efficient way to identify prognostic features.

Contact: Verena Bitto

Propagator Methods for Survival Analysis

In this project, inspired by the concepts and ideas of quantum mechanics, we aim to model the survival of patients and their tumour growth with a patient-near stochastic process. We imagine time-dependent patient data as positional states in an abstract space and try to describe a transition function (propagator) between these states; the parameters of this function can be understood as “natural laws of cancer growth”. In this way, we aim to characterise both deterministic and seemingly random aspects of disease progression and develop a novel method to extract information from survival data, in particular by incorporating temporal evolution. This represents a crucial step for scientific understanding and subsequent individualised clinical treatment.

Contact: Julian Schlecker

FAIRification of preclinical and clinical data

Our work focuses on standardizing and formalizing data collected from preclinical and clinical studies to enhance knowledge discovery. We began by standardizing digitized data hosted on the RadPlanBio platform, structuring it according to the SEND standard for nonclinical datasets. The data is then formalized using the Ontology for Preclinical Trials in Radiation Oncology (PTRO). Building on this foundation, we are using Ontop to develop a Virtual Knowledge Graph (VKG) based on the PTRO ontology. This VKG allows users to access and query heterogeneous data sources as a unified graph, without physically moving or duplicating the data. These tools provide a structured and transparent representation of data, ensuring clarity, consistency, and reproducibility across studies. Looking ahead, we are extending these FAIRification efforts to clinical data, paving the way for greater integration, accessibility, and usability across the research continuum.

Contact: Olga Giraldo Pasmin