Development of an Automatic Measurement Method for CD8 and PD-1 Positive T Cells Using Image Analysis Software

Abstract

Background/Aim: With the progress in cancer immunotherapy using immune checkpoint blockade (ICB) therapy, histological observations of tumor-infiltrating lymphocyte (TIL) status are needed to evaluate the antitumor effect of ICB using imaging analysis software. Materials and Methods: Formalin-fixed paraffin-embedded sections obtained from colorectal cancer and gastric cancer patients with more than 500 single nucleotide variants were stained with anti-CD8 and anti-PD-1 antibodies. Based on our own algorithm and imaging analysis software, an automatic TIL measurement method was established and compared to the manual counting methods. Results: In the CD8 + T cell number measurement, there was a good correlation (r=0.738 by Pearson test) between the manual and automated counting methods. However, in the PD-1 + T cell measurement, there was a large difference in TIL numbers in both groups. After adjustment of the parameter settings, the correlation between the manual and automated methods in the PD-1 + T cell measurements improved (r=0.668 by Pearson test). Conclusion: An imaging software-based automatic measurement could be a simple and useful tool for evaluating the therapeutic effect of cancer immunotherapies in terms of TIL status.Since the clinical applications of immune checkpoint blockade (ICB) therapy have had great success in various cancer patients, the number of clinical trials focused on novel therapies, including the combination of ICB with other targeting agents, have been increasing (1, 2). The greater the number of ICBbased clinical trials, the more difficult visual inspections by trained pathologists have become for the evaluation of immune effects by ICB, such as tumor-infiltrating lymphocyte (TIL) measurement and programmed death-ligand 1 (PD-L1) staining.In fact, many pathologists have intensively studied methods to enumerate TIL numbers accurately and efficiently and demonstrated that a well-functioning immune-scoring system could contribute to determine the TIL status of tumors (3-6). However, manual visual inspections are time-consuming and require manual labor; therefore, an efficient and user-friendly automatic immunohistochemistry (IHC) imaging analysis system is urgently needed. Specifically, since the development of whole slide scanners that can generate ultralarge 2D images in the field of digital pathology, open source bioimage analysis software has been developed, such as ImageJ, Fiji, Icy, and CellProfiler (7-10). However, in terms of high variability and limited reproducibility, it has been difficult for computational researchers to develop state-of-the-art imaging software 419 This article is freely accessible online.