Survival Prediction and Tumor-Immune Microenvironment in Non-Small Cell Lung Cancer – Love Nordling

The immune microenvironment plays an important role in the tumorigenesis of non-small cell lung cancer (NSCLC). Accumulating evidence suggests that immune cell infiltration is strongly associated with therapy response and patient survival. However, most studies evaluate one or a few markers, disregarding cell interactions. Furthermore, their often descriptive nature questions the clinical value of such immune features. Using multiplex immunofluorescence imaging, the tumor-immune microenvironment can be visualized with precise cell locations and phenotyping of multiple immune cell types. We aim to predict survival in Non-small cell lung cancer (NSCLC) patients using the rich information found in multiplex immunofluorescence images of tumor tissue.

The aim of the talk is to discuss the results and challenges of analyzing information-rich multiplex image data with a limited number of samples and with noisy labels. Classical cell density and handcrafted feature analysis are compared with patch-based multiple instance learning and graph neural networks.