Malignant Mesothelioma subtyping of tissue images via sampling driven multiple instance prediction

Malignant Mesothelioma is a difficult to diagnose and highly lethal cancer usually associated with asbestos exposure. It can be broadly classified into three subtypes: Epitheliod, Sarcomatoid, and Biphasic. Early diagnosis and identification of the subtype informs treatment and can help improve patient outcome. However, the subtyping of malignant mesothelioma, and specifically the recognition of transitional features from routine histology slides has a high level of inter-observer variability. In this work, we propose the first end-to-end multiple instance learning (MIL) approach for malignant mesothelioma subtyping. This uses an instance-based sampling scheme for training deep convolutional neural networks on this task that allows learning on a wider range of relevant instances compared to max or top-N based MIL approaches. The proposed MIL approach enables identification of malignant mesothelial sub-types of specific tissue regions. From this a continuous characterization of a sample according to predominance of sarcomatoid vs epithelioid regions is possible, thus avoiding the arbitrary and highly subjective categorisation by currently used subtypes. Instance scoring also enables studying tumor heterogeneity and identifying patterns associated with different subtypes. We have evaluated the proposed method on a dataset of 243 tissue micro-array cores with an AUROC of 0.87 ± 0.04 for this task. The dataset and developed methodology is available for the community at: https://github.com/measty/PINS.