Automated PD-L1 Scoring Using Artificial Intelligence in Head and Neck Squamous Cell Carcinoma

Simple Summary Immunotherapy forms an emerging and successful field in cancer therapy using checkpoint inhibitors (e.g., anti PD-L1, anti PD-1), preventing immune escape of the tumor. However, these drugs are often only effective in a subpopulation of patients. To identify such patients, various so-called PD-L1 scores based on PD-L1 expression by immunohistochemistry in tumor tissue had been established. However, these scores may vary between different human investigators, which could negatively influence treatment decisions. The aim of our work was to obtain reproducible and reliable PD-L1 scores using artificial intelligence. Our results show comparable performance between human-human and human-machine interactions and could provide a deeper insight into the function and limitations of automated scoring by artificial intelligence. This could serve as a basis to improve patient selection for checkpoint inhibitors in the future. Immune checkpoint inhibitors (ICI) represent a new therapeutic approach in recurrent and metastatic head and neck squamous cell carcinoma (HNSCC). The patient selection for the PD-1/PD-L1 inhibitor therapy is based on the degree of PD-L1 expression in immunohistochemistry reflected by manually determined PD-L1 scores. However, manual scoring shows variability between different investigators and is influenced by cognitive and visual traps and could therefore negatively influence treatment decisions. Automated PD-L1 scoring could facilitate reliable and reproducible results. Our novel approach uses three neural networks sequentially applied for fully automated PD-L1 scoring of all three established PD-L1 scores: tumor proportion score (TPS), combined positive score (CPS) and tumor-infiltrating immune cell score (ICS). Our approach was validated using WSIs of HNSCC cases and compared with manual PD-L1 scoring by human investigators. The inter-rater correlation (ICC) between human and machine was very similar to the human-human correlation. The ICC was slightly higher between human-machine compared to human-human for the CPS and ICS, but a slightly lower for the TPS. Our study provides deeper insights into automated PD-L1 scoring by neural networks and its limitations. This may serve as a basis to improve ICI patient selection in the future.

Automated summary

This study aimed to achieve reproducible and reliable PD-L1 scoring using artificial intelligence to enhance patient selection for checkpoint inhibitors. Results showed comparable performance between human-human and human-machine interactions, providing deeper insights into the function and limitations of automated scoring by artificial intelligence.