Interpretable AI-based multispectral 3D-analysis of cell interrelations in cancer microenvironment for patient-specific therapy

Immunotherapy is a life-saving option for advanced cancer patients. However, only a few patients develop a durable response.

Despite great efforts to explain response variability and optimize patient selection, current diagnostic tools cannot sufficiently guide clinical practice.

There is evidence that the structural organization and interrelations of different cell types in the microenvironment reflect states of immune reactions, but the characterization of these properties is still insufficient due to conceptual and methodological deficiencies of the current analysis approaches.Based on our interdisciplinary alliance of computational and clinical competency, we will go beyond the norm of histopathology and enable analysis of cell interrelations in their natural 3D environments, considerably extending the information that is currently available from 2D tissue sections.

Towards this goal, we will bring the latest multiplex tissue profiling techniques to 3D space.

To harvest from the created information-rich 3D multispectral data, we will develop advanced graph-based deep learning methods to capture the structural diversity of the cancer microenvironment.

By providing interpretability of the AI methods, we will ensure clinical applicability and enable medical experts to improve patient-specific cancer therapy.

Interpretable AI models will broaden the understanding of complex behaviors by mapping the regulatory processes which characterize biological phenomena of tumorigenesis.