A computational pathology-based AI framework for predicting PD-L1 expression and prognosis in HNSCC

Abstract

Aim: To investigate interobserver variability in programmed cell death ligand 1 (PD-L1) combined positive score (CPS) assessment in head and neck squamous cell carcinoma (HNSCC) and to develop an artificial intelligence (AI)-based model for predicting PD-L1 expression and patient prognosis from hematoxylin and eosin (H&E)-stained slides.

Methods: Fifty HNSCC specimens were independently evaluated for PD-L1 by pathologists with different experience levels. Agreement was assessed using Fleiss’ and Cohen’s κ. Whole-slide images were processed into tiles for deep learning using DenseNet121. Tile-level features were integrated via two machine learning pipelines to construct whole-slide prediction models. Multiple algorithms were tested, with performance evaluated in validation and testing cohorts. Prognostic value was analyzed using AI-derived risk stratification.

Results: Interobserver agreement was low (Fleiss’ κ = 0.34), indicating substantial variability in CPS assessment. DenseNet121 achieved moderate predictive performance (AUC 0.641 in validation, 0.616 in testing). AI models significantly improved prediction accuracy, with logistic regression demonstrating the best performance (AUC 0.900 in validation, 0.851 in testing). AI-derived prediction scores effectively stratified overall survival, with multiple models showing significant prognostic discrimination (P < 0.05).

Conclusions: AI models integrating deep learning and machine learning can accurately predict PD-L1 expression and stratify prognosis in HNSCC, outperforming tile-level deep learning alone.

Keywords

head and neck squamous cell carcinoma / programmed cell death ligand 1 / deep learning / computational pathology / artificial intelligence

References

[1] Dunn LA, Ho AL, Pfister DG. Head and Neck Cancer: A Review. Jama. 2026; 335: 531-41.

[2] Mody MD, Rocco JW, Yom SS, Haddad RI, Saba NF. Head and neck cancer. Lancet. 2021; 398: 2289-99.

[3] Rumgay H, Colombet M, Ramos da Cunha A, Filho AM, Warnakulasuriya S, Conway DI, et al. Global incidence of lip, oral cavity, and pharyngeal cancers by subsite in 2022. CA: Cancer J Clin. 2025; 76: e76.

[4] Cai X, Zhang J, Jing F, Zhou X, Zhang H, Li T. Clinical and prognostic features of multiple primary cancers with oral squamous cell carcinoma. Arch Oral Biol. 2023; 149: 105661.

[5] Soerjomataram I, Cabasag C, Bardot A, Fidler-Benaoudia MM, Miranda-Filho A, Ferlay J, et al. Cancer survival in Africa, central and south America, and Asia (SURVCAN-3): a population-based benchmarking study in 32 countries. Lancet Oncol. 2023; 24: 22-32.

[6] Cai X, Huang J. Distant metastases in newly diagnosed tongue squamous cell carcinoma. Oral Dis. 2019; 25: 1822-8.

[7] Alsahafi E, Begg K, Amelio I, Raulf N, Lucarelli P, Sauter T, et al. Clinical update on head and neck cancer: molecular biology and ongoing challenges. Cell Death Dis. 2019; 10: 540.

[8] Chang JYF, Tseng CH, Lu PH, Wang YP. Contemporary Molecular Analyses of Malignant Tumors for Precision Treatment and the Implication in Oral Squamous Cell Carcinoma. J Pers Med. 2021; 12: 12.

[9] Van den Bossche V, Zaryouh H, Vara-Messler M, Vignau J, Machiels JP, Wouters A, et al. Microenvironment-driven intratumoral heterogeneity in head and neck cancers: clinical challenges and opportunities for precision medicine. Drug Resist Updates. 2022; 60: 100806.

[10] Cai XJ, Zhang HY, Zhang JY, Li TJ. Bibliometric analysis of immunotherapy for head and neck squamous cell carcinoma. J Dent Sci. 2023; 18: 872-82.

[11] Burtness B, Harrington KJ, Greil R, Soulières D, Tahara M, de Castro G Jr, et al.; KEYNOTE-048 Investigators. Pembrolizumab alone or with chemotherapy versus cetuximab with chemotherapy for recurrent or metastatic squamous cell carcinoma of the head and neck (KEYNOTE-048): a randomised, open-label, phase 3 study. Lancet. 2019; 394: 1915-28.

[12] Doroshow DB, Bhalla S, Beasley MB, Sholl LM, Kerr KM, Gnjatic S, et al. PD-L1 as a biomarker of response to immune-checkpoint inhibitors. Nat Rev Clin Oncol. 2021; 18: 345-62.

[13] Cohen EEW, Bell RB, Bifulco CB, Burtness B, Gillison ML, Harrington KJ, et al. The Society for Immunotherapy of Cancer consensus statement on immunotherapy for the treatment of squamous cell carcinoma of the head and neck (HNSCC). J ImmunoTher Cancer. 2019; 7: 184.

[14] Machiels JP, Tao Y, Licitra L, Burtness B, Tahara M, Rischin D, et al.; KEYNOTE-412 Investigators. Pembrolizumab plus concurrent chemoradiotherapy versus placebo plus concurrent chemoradiotherapy in patients with locally advanced squamous cell carcinoma of the head and neck (KEYNOTE-412): a randomised, double-blind, phase 3 trial. Lancet Oncol. 2024; 25: 572-87.

[15] Harrington KJ, Burtness B, Greil R, Soulières D, Tahara M, de Castro G Jr, et al. Pembrolizumab With or Without Chemotherapy in Recurrent or Metastatic Head and Neck Squamous Cell Carcinoma: Updated Results of the Phase III KEYNOTE-048 Study. J Clin Oncol. 2023; 41: 790-802.

[16] De Keukeleire SJ, Vermassen T, Deron P, Huvenne W, Duprez F, Creytens D, et al. Concordance, Correlation, and Clinical Impact of Standardized PD-L1 and TIL Scoring in SCCHN. Cancers. 2022; 14: 2431.

[17] Kondo Y, Suzuki S, Ono S, Goto M, Miyabe S, Ogawa T, et al. In Situ PD-L1 Expression in Oral Squamous Cell Carcinoma Is Induced by Heterogeneous Mechanisms among Patients. Int J Mol Sci. 2022; 23: 4077.

[18] Akhtar M, Rashid S, Al-Bozom IA. PD−L1 immunostaining: what pathologists need to know. Diagn Pathol. 2021; 16: 94.

[19] Cai X, Zhang H, Wang Y, Zhang J, Li T. Digital pathology-based artificial intelligence models for differential diagnosis and prognosis of sporadic odontogenic keratocysts. Int J Oral Sci. 2024; 16: 16.

[20] Cai X, Li L, Yu F, Guo R, Zhou X, Zhang F, et al. Development of a Pathomics-Based Model for the Prediction of Malignant Transformation in Oral Leukoplakia. Lab Investig. 2023; 103: 100173.

[21] Valanarasu JMJ, Xu H, Usuyama N, Kim C, Wong C, Argaw P, et al. Multimodal AI generates virtual population for tumor microenvironment modeling. Cell. 2026; 189: 386-400.e19.

[22] Song AH, Williams M, Williamson DFK, Chow SSL, Jaume G, Gao G, et al. Analysis of 3D pathology samples using weakly supervised AI. Cell. 2024; 187: 2502-20.e17.

[23] Cai XJ, Peng CR, Ding CY, Cui YY, Gao L, Xu ZX, et al. Tumor cell- and infiltrating immune cell-based supervised learning artificial intelligence multimodal platform for tumor prognosis. npj Precis Oncol. 2025; 9: 348.

[24] Wagner SJ, Reisenbüchler D, West NP, Niehues JM, Zhu J, Foersch S, et al. Transformer-based biomarker prediction from colorectal cancer histology: A large-scale multicentric study. Cancer Cell. 2023; 41: 1650-61.e4.

[25] Wang S, Rong R, Zhou Q, Yang DM, Zhang X, Zhan X, et al. Deep learning of cell spatial organizations identifies clinically relevant insights in tissue images. Nat Commun. 2023; 14: 7872.

[26] Jiang S, Li X, Huang L, Xu Z, Lin J. Prognostic value of PD-1, PD-L1 and PD-L2 deserves attention in head and neck cancer. Front Immunol. 2022; 13: 988416.

[27] Strati A, Koutsodontis G, Papaxoinis G, Angelidis I, Zavridou M, Economopoulou P, et al. Prognostic significance of PD-L1 expression on circulating tumor cells in patients with head and neck squamous cell carcinoma. Ann Oncol. 2017; 28: 1923-33.

[28] de Vicente JC, Rodríguez-Santamarta T, Rodrigo JP, Blanco-Lorenzo V, Allonca E, García-Pedrero JM. PD-L1 Expression in Tumor Cells Is an Independent Unfavorable Prognostic Factor in Oral Squamous Cell Carcinoma. Cancer Epidemiol Biomark Prev. 2019; 28: 546-54.

[29] Zeng Q, Klein C, Caruso S, Maille P, Laleh NG, Sommacale D, et al. Artificial intelligence predicts immune and inflammatory gene signatures directly from hepatocellular carcinoma histology. J Hepatol. 2022; 77: 116-27.

[30] Nero C, Boldrini L, Lenkowicz J, Giudice MT, Piermattei A, Inzani F, et al. Deep-Learning to Predict BRCA Mutation and Survival from Digital H&E Slides of Epithelial Ovarian Cancer. Int J Mol Sci. 2022; 23: 11326.

[31] Huang H, Zhou G, Liu X, Deng L, Wu C, Zhang D, et al. Contrastive learning-based computational histopathology predict differential expression of cancer driver genes. Brief Bioinform. 2022; 23: e23.

[32] Fujii S, Kotani D, Hattori M, Nishihara M, Shikanai T, Hashimoto J, et al. Rapid Screening Using Pathomorphologic Interpretation to Detect BRAF V600E Mutation and Microsatellite Instability in Colorectal Cancer. Clin Cancer Res. 2022; 28: 2623-32.

[33] Chen RJ, Lu MY, Williamson DFK, Chen TY, Lipkova J, Noor Z, et al. Pan-cancer integrative histology-genomic analysis via multimodal deep learning. Cancer Cell. 2022; 40: 865-78.e6.

[34] Qu H, Zhou M, Yan Z, Wang H, Rustgi VK, Zhang S, et al. Genetic mutation and biological pathway prediction based on whole slide images in breast carcinoma using deep learning. npj Precis Oncol. 2021; 5: 87.

[35] Bilal M, Raza SEA, Azam A, Graham S, Ilyas M, Cree IA, et al. Development and validation of a weakly supervised deep learning framework to predict the status of molecular pathways and key mutations in colorectal cancer from routine histology images: a retrospective study. Lancet Digit Health. 2021; 3: e763-72.

[36] Schmauch B, Romagnoni A, Pronier E, Saillard C, Maillé P, Calderaro J, et al. A deep learning model to predict RNA-Seq expression of tumours from whole slide images. Nat Commun. 2020; 11: 3877.

[37] Cai XJ, Peng CR, Cui YY, Li L, Huang MW, Zhang HY, et al. Identification of genomic alteration and prognosis using pathomics-based artificial intelligence in oral leukoplakia and head and neck squamous cell carcinoma: a multicenter experimental study. Int J Surg. 2025; 111: 426-38.

[38] Zaakouk M, Van Bockstal M, Galant C, Callagy G, Provenzano E, Hunt R, et al. Inter- and Intra-Observer Agreement of PD-L1 SP142 Scoring in Breast Carcinoma-A Large Multi-Institutional International Study. Cancers. 2023; 15: 1511.

[39] Fernandez AI, Robbins CJ, Gaule P, Agostini-Vulaj D, Anders RA, Bellizzi AM, et al. Multi-Institutional Study of Pathologist Reading of the Programmed Cell Death Ligand-1 Combined Positive Score Immunohistochemistry Assay for Gastric or Gastroesophageal Junction Cancer. Mod Pathol. 2023; 36: 100128.

[40] Park BJ, Mattox AK, Clayburgh D, Patel M, Bell RB, Yueh B, et al. Chemoradiation therapy alters the PD-L1 score in locoregional recurrent squamous cell carcinomas of the head and neck. Oral Oncol. 2022; 135: 106183.

[41] Braxton AM, Kiemen AL, Grahn MP, Forjaz A, Parksong J, Mahesh Babu J, et al. 3D genomic mapping reveals multifocality of human pancreatic precancers. Nature. 2024; 629: 679-87.

[42] Foersch S, Glasner C, Woerl AC, Eckstein M, Wagner DC, Schulz S, et al. Multistain deep learning for prediction of prognosis and therapy response in colorectal cancer. Nat Med. 2023; 29: 430-9.