The following abstract is drawn from a recently published paper in the National Library of Medicine. We invite you to read the full paper and join the conversation, become a member of the Pathology News community to share your thoughts, ask questions, and engage with others around this work.

Authors: Yasmine Makhlouf¹, Perry Maxwell¹, Paul O’Reilly², Alice Geaney², Jacqueline A James^1,3, Manuel Salto-Tellez^1,2,3,4

Abstract

The assessment of PD-L1 in lung cancer using the Tumour Proportion Score (TPS) is one of the cornerstones of immune-oncology, but it is open to inter- and intra-pathologist variation, particularly around the clinical thresholds of less than 1%and ⩾ 50%, which correspond to analytical thresholds less than 5% and between 40%-60%. In this paper we describe the development of a deep learn- ing (DL) tool to assist TPS calculation. To confirm ground truth values around the clinical thresholds, we used a validated multiplex immunfluorescence panel including PD-L1, CD68 and cytokeratin. Practically, the DL tool is designed to assist in highlighting cases about these thresholds around the 1% and 50% levels for manual review, and allowing a direct interpretation of inbetween scores. Us- ing such an assisted system, we highlight the potential use of such DL tools in providing a route to future clinical quantitation of tissue-based biomarkers.