Institute of Pathology, Medical Faculty, University Hospital Cologne, Cologne, Germany.
Department of Translational Genomics, Center of Integrated Oncology Cologne-Bonn, Medical Faculty, University of Cologne, Cologne, Germany.
Clin Cancer Res. 2021 Feb 15;27(4):1131-1138. doi: 10.1158/1078-0432.CCR-20-3596. Epub 2020 Dec 1.
Human papillomavirus (HPV) in oropharyngeal squamous cell carcinoma (OPSCC) is tumorigenic and has been associated with a favorable prognosis compared with OPSCC caused by tobacco, alcohol, and other carcinogens. Meanwhile, machine learning has evolved as a powerful tool to predict molecular and cellular alterations of medical images of various sources.
We generated a deep learning-based HPV prediction score (HPV-ps) on regular hematoxylin and eosin (H&E) stains and assessed its performance to predict HPV association using 273 patients from two different sites (OPSCC; Giessen, = 163; Cologne, = 110). Then, the prognostic relevance in a total of 594 patients (Giessen, Cologne, HNSCC TCGA) was evaluated. In addition, we investigated whether four board-certified pathologists could identify HPV association ( = 152) and compared the results to the classifier.
Although pathologists were able to diagnose HPV association from H&E-stained slides (AUC = 0.74, median of four observers), the interrater reliability was minimal (Light Kappa = 0.37; = 0.129), as compared with AUC = 0.8 using the HPV-ps within two independent cohorts ( = 273). The HPV-ps identified individuals with a favorable prognosis in a total of 594 patients from three cohorts (Giessen, OPSCC, HR = 0.55, < 0.0001; Cologne, OPSCC, HR = 0.44, = 0.0027; TCGA, non-OPSCC head and neck, HR = 0.69, = 0.0073). Interestingly, the HPV-ps further stratified patients when combined with p16 status (Giessen, HR = 0.06, < 0.0001; Cologne, HR = 0.3, = 0.046).
Detection of HPV association in OPSCC using deep learning with help of regular H&E stains may either be used as a single biomarker, or in combination with p16 status, to identify patients with OPSCC with a favorable prognosis, potentially outperforming combined HPV-DNA/p16 status as a biomarker for patient stratification.
人乳头瘤病毒（HPV）在口咽鳞状细胞癌（OPSCC）中具有致瘤性，与由烟草、酒精和其他致癌物引起的 OPSCC 相比，HPV 与预后较好相关。同时，机器学习已发展成为一种强大的工具，可以预测各种来源的医学图像的分子和细胞变化。
我们在常规苏木精和伊红（H&E）染色上生成了一种基于深度学习的 HPV 预测评分（HPV-ps），并使用来自两个不同地点的 273 名患者（OPSCC；吉森， = 163；科隆， = 110）评估了其预测 HPV 相关性的性能。然后，在总共 594 名患者（吉森、科隆、HNSCC TCGA）中评估了其预后相关性。此外，我们还研究了四位认证病理学家是否能够识别 HPV 相关性（ = 152），并将结果与分类器进行了比较。
尽管病理学家能够从 H&E 染色载玻片上诊断 HPV 相关性（AUC = 0.74，四位观察者的中位数），但与 HPV-ps 在两个独立队列中（ = 273）的 AUC = 0.8 相比，观察者间的可靠性非常低（Light Kappa = 0.37； = 0.129）。HPV-ps 在来自三个队列（吉森、OPSCC、HR = 0.55， < 0.0001；科隆、OPSCC、HR = 0.44， = 0.0027；TCGA、非 OPSCC 头颈部、HR = 0.69， = 0.0073）的 594 名患者中均能识别出预后较好的个体。有趣的是，当与 p16 状态结合时，HPV-ps 进一步对患者进行分层（吉森，HR = 0.06， < 0.0001；科隆，HR = 0.3， = 0.046）。
使用常规 H&E 染色的深度学习检测 OPSCC 中的 HPV 相关性，可以作为单一生物标志物使用，也可以与 p16 状态结合使用，以识别预后较好的 OPSCC 患者，其作为患者分层的生物标志物可能优于 HPV-DNA/p16 状态的组合。
