Hidden in plain sight

Tuesday 7 May 2024, 10:30 am
PhD student
J.M. Bokhorst MSc.
prof. dr. J.A.W.M. van der Laak, prof. dr. I. Nagtegaal, dr. F. Ciompi

Tumor budding (TB), the presence of single cells or small clusters of up to four tumor cells at the invasive front of colorectal cancer (CRC), is a proven risk factor for adverse outcomes. Despite their prognostic value, the potential of TBs as prognostic biomarkers is currently underused within clinical practice, partly due to observer variability among pathologists and partly because it is very time-consuming. Therefore, we dedicate this thesis to developing deep learning algorithms for automatically, efficiently, and precisely detecting these tumor buds. Every deep learning model needs good training data. Therefore, we start by collecting a reference standard with the help of 11 field pathology experts that are subsequently used to develop these algorithms. The algorithms developed in this thesis will contribute to the implementation of tumor buds in daily practice and yield further insights into tumor buds and their interaction with the microenvironment.

John-Melle Bokhorst (1991) graduated from Eindhoven University of Technology in 2018. In the same year, he became a Ph.D. student at the computational pathology group and pathology department of the Radboud University Medical Center. He developed AI algorithms for better stratification of colorectal cancer patients. Subsequently, he was commissioned by the department to develop AI algorithms for research on ovarian and breast cancer.