Poster
Deep-Learning–Based Prediction of c-MET Status From Digitized H&E-Stained Non-small Cell Lung Cancer Tissue Samples
WCLC 2022
Study Background
The human-interpretable features (HIF) generated by the convolutional neural networks (CNN) models showed that the composition of the tumor microenvironment (TME) is significantly different in tissue, including lymphocyte density, when c-MET is overexpressed.
Multivariate regression (MR) based on TME features and graph neural networks (GNN) models was able to identify patients with tumors that overexpressed c-MET; however, the performance of the GNN model was stronger.
These results indicate there is a potential to develop and validate an H&E-based screening tool for patient selection for c-MET targeting therapies, thereby increasing efficiency and access.
Multivariate regression (MR) based on TME features and graph neural networks (GNN) models was able to identify patients with tumors that overexpressed c-MET; however, the performance of the GNN model was stronger.
These results indicate there is a potential to develop and validate an H&E-based screening tool for patient selection for c-MET targeting therapies, thereby increasing efficiency and access.
Authors
Rajan et al.