Smart pathology slide scanner for diagnosis and patient-specific treatment recommendation in oncology

Our long-term vision is to develop a smart digital pathology slide scanner that can (semi)automatically generate a patient-specific tumour digital twin and simulate various drug treatments on such twin. The digital twin will be generated by combining slide-level multi-omics profiling of patients’ biopsies with other standard clinical data and will be hardware-embedded within a slide scanner.This foundational architecture involves five pillars: experimental mapping of tumor communication, reconstruction of Biomedical Knowledge Graphs (KG) based on experimental data, identification of structural and temporal properties of communication networks, development of a mathematical framework for modeling treatment effects on a tumor, and the design of a novel hardware acceleration architecture for a medical digital twin. As a model system, we will use pancreatic cancer, combining in vitro experiments with clinical data. We will analyze how tumor microenvironment composition affects cellular crosstalk and drug efficacy by developing, growing, and treating a series of tumor organoids with different TME compositions, and performing a detailed molecular analysis of samples. Clinical validation will involve matching organoid and patient biopsy structures, ensuring relevance and applicability of our findings. This project is the initial step towards a platform assisting doctors in improving the diagnosis and assessment of drug treatment efficacy for individual cancer patients so that each patient gets the best possible drug, with the best possible treatment regimen.