A Review of Artificial Intelligence in Precise Assessment of Programmed Cell Death-ligand 1 and Tumor-infiltrating Lymphocytes in Non-Small Cell Lung Cancer.

Abstract

Advances in immunotherapy have increased the need for stratified predictive biomarkers in patients with non-small cell lung cancer. However, precise evaluation of tumor tissue-based immune biomarkers, such as programmed cell death-ligand 1 (PD-L1) and the characteristics of tumor infiltrating lymphocytes (TILs), is a challenge in clinical practice. In recent years, the digitization of whole-slide images of tissue has accelerated the implementation of artificial intelligence (AI) approaches in tumor pathology and provided an opportunity to use AI tools to improve the interpretation of immune biomarkers. This review describes the current challenges in the assessment of PD-L1 scoring and TILs and demonstrates the role of AI in helping pathologists integrate PD-L1 and biomarkers of the tumor immune microenvironment. Computer-aided PD-L1 scoring is highly consistent with pathologists and reduces the variation among interobservers, providing a promising diagnostic tool in pathology clinics. In addition, applications of image analysis algorithms, in combination with multiplex staining, enable in-depth quantitative and spatial analysis of the broader tumor microenvironment. Upon combining digital pathology and AI, an automatic analysis system of PD-L1 and TILs, which was established using a set of digital staining images and deep learning algorithms, might be an effective way to overcome the challenges in the precise assessment of immune biomarkers.