Multi-resolution deep learning characterizestertiary lymphoid structures in solid tumors

Abstract

Tertiary lymphoid structures (TLSs) are dense accumulations of lym-phocytes in inflamed peripheral tissues, including cancer, and are associated with improved survival and response to immunotherapy in various solid tumors. Histological TLS quantification has been pro-posed as a novel predictive and prognostic biomarker, but lack of standardized methods of TLS characterization hampers assessment of TLS densities across different patients, diseases, and clinical centers. We introduce a novel approach based on HookNet-TLS, a multi-resolution deep learning model, for automated and unbiased TLS quantification and identification of germinal centers in routine hema-toxylin and eosin stained digital pathology slides. We developed a HookNet-TLS model using n=1019 manually annotated TCGA slides from clear cell renal cell carcinoma, muscle-invasive blad-der cancer, and lung squamous cell carcinoma. We show that HookNet-TLS automates TLS quantification with a human-level performance and demonstrates prognostic associations similar to visual assessment. We made HookNet-TLS publicly available to aid the adoption of objective TLS assessment in routine pathology.