Weakly-supervised learning for image-based classification of primary melanomas into genomic immune subgroups

Lucy Godson, Navid Alemi, Jeremie Nsengimana, Graham Cook, Emily L Clarke, Darren Treanor, D Timothy Bishop, Julia A Newton-Bishop, Ali Gooya

Keywords: digital pathology, deep learning, multiple instance learning, attention, melanoma

TL;DR: We implement different multiple instance learning approaches to classify H&E stained primary melanoma images into immune subgroups determined through consensus clustering of the corresponding tumour transcriptomes.

Abstract: Determining early-stage prognostic markers and stratifying patients for effective treatment are two key challenges for improving outcomes for melanoma patients. Previous studies have used tumour transcriptome data to stratify patients into immune subgroups, which were associated with differential melanoma specific survival and potential treatment strategies. However, acquiring transcriptome data is a time-consuming and costly process. Moreover, it is not routinely used in the current clinical workflow. Here we attempt to overcome this by developing deep learning models to classify gigapixel H\&E stained pathology slides, which are well established in clinical workflows, into these immune subgroups. Previous subtyping approaches have employed supervised learning which requires fully annotated data, or have only examined single genetic mutations in melanoma patients. We leverage a multiple-instance learning approach, which only requires slide-level labels and uses an attention mechanism to highlight regions of high importance to the classification. Moreover, we show that pathology-specific self-supervised models generate better representations compared to pathology-agnostic models for improving our model performance, achieving a mean AUC of 0.76 for classifying histopathology images as high or low immune subgroups. We anticipate that this method may allow us to find new biomarkers of high importance and could act as a tool for clinicians to infer the immune landscape of tumours and stratify patients, without needing to carry out additional expensive genetic tests.

Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: validation/application paper

Primary Subject Area: Application: Histopathology

Secondary Subject Area: Integration of Imaging and Clinical Data

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.

Code And Data:

Author's GitHub repository: https://github.com/lucyOCg/Weakly-supervised-learning-for-histopathology-image-based-classification
CLAM GitHub repository: https://github.com/mahmoodlab/CLAM
Self supervised contrastive learning for digital histopathology GitHub repository: https://github.com/ozanciga/self-supervised-histopathology/releases/tag/tenpercent

The imaging data is from Virtual Pathology at the University of Leeds (https://www.virtualpathology.leeds.ac.uk/). Ethical approval for the Leeds Melanoma does not currently extend to widespread image sharing, however, to request access to the raw data please contact the Virtual Pathology team.