Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 7: Poster 1
BP 7.31: Poster
Monday, September 5, 2022, 18:00–20:00, P1
Image segmentation of irradiated tumour spheroids by Fully Convolutional Networks — •Matthias Streller1, Sona Michlikova2, Leoni A. Kunz-Schughart2, Steffen Lange1, and Anja Voss-Boehme1 — 1University of Applied Sciences Dresden — 2OncoRay, National Center for Radiation Research in Oncology
Multicellular tumour spheroids are an established in-vitro model to quantify the effectiveness of cancer therapies. Spheroids are treated with radiotherapy and their therapeutic response over time is most frequently monitored via microscopic imaging. For analysis, it is necessary to segment the spheroids in these images, to extract their characteristics like the average diameter or circularity. While several image analysis algorithms have been developed for the automatic segmentation of spheroid images, they focus on more or less compact and circular spheroids with clearly distinguishable outer rim throughout growth. In contrast, treated spheroids are usually obscured by debris of dead cells and might be partly detached and destroyed. We train and optimize two Fully Convolutional Networks, in particular UNet and HRNet, to create an automatic segmentation which covers both cases, spheroids with and without therapy. While we successfully demonstrate the automatic segmentation for one spheroid type, we plan to extent the segmentation to other spheroid models.