MesoNet
A biomarker of outcome prediction in Mesothelioma
Predicting mesothelioma survival from pathology images
Context
Malignant mesothelioma patients are diagnosed and classified by pathologists via tissue biopsy. Only three FDA-approved drugs exist for mesothelioma with varying response. Novel biomarkers are needed to improve clinical care and offer possible targets for new drugs.
Methods
Owkin built MesoNet, a machine learning model that predicts the overall survival of malignant mesothelioma patients from digital pathology images better than existing subtype classification used by pathologists.
Results
MesoNet highlighted new biomarkers predictive of prognosis in collaboration with expert mesothelioma pathologists: stroma, tumor cell localization, inflammation, cellular diversity & vascularization.
Impact
MesoNet can be used in clinical routine to aid patient prognosis, providing important input for patient management.
The novel morphological biomarkers identified in the tumour microenvironment could lead to the discovery of new druggable targets.
Published in Nature medicine Courtiol et al. 2019
