35 Esther Smeets Computer Vision News Computer Vision News Using semi-supervised clustering, she identified two distinct patterns of MCT4-marker. Secondly, from the corresponding PET-scans the PET texture features were extracted. Based on tissue derived MCT4marker patterns, she developed PET radiomic signature that allows to identify subgroups of PDAC patients with high MCT4-marker expression. By applying these tissue-based PET radiomic signatures in an independent patient cohort, she found that high glucose consumption in PDAC tumors associates with worse prognosis. Esther’s findings indicate that tumor glucose consumption is a major determinant of tumor aggressiveness in PDAC. This discovery underscores the potential of biology and AI-driven imaging biomarkers in PDAC subtyping. Computer Vision in Tumor Analysis Esther also used computational pathology techniques to analyze different cells in the tumor, the cancer cells, cells which support the cancer cells, and immune cells which try to attack the cancer cells. These findings revealed new insights into interplay between cancer cells, cancer support cells and the population of immune cells that is commonly targeted for therapy. As a next step, Esther used these computer vision tools to quantitatively analyze immunohistochemistry images to support the development of a new treatment for PDAC patients. Future Prospects: AI-Driven Treatment Strategies This thesis highlights the transformative potential of computer vision and AI in PDAC research. By integrating imaging, computational pathology, and machine learning, the study showed a way for non-invasive, AI-driven diagnostic tools. Demonstrating that interdisciplinary innovation holds the key to overcoming one of oncology’s most formidable challenges. Schematic representation of the image analysis method, in which [18F]FDG-PET is linked to a biological rationale based on immunohistochemistry.
RkJQdWJsaXNoZXIy NTc3NzU=