Superiority of Radiomics Information Compared to Clinical Factors in Characterization of Human Papilloma Virus (HPV) Status in Patients With Oropharyngeal Squamous Cell Carcinomas
Bagher-Ebadian H, Siddiqui F, Ghanem AI, Zhu S, Lu M, Movsas B, and Chetty IJ. Superiority of Radiomics Information Compared to Clinical Factors in Characterization of Human Papilloma Virus (HPV) Status in Patients With Oropharyngeal Squamous Cell Carcinomas. Int J Radiat Oncol Biol Phys 2021; 111(3):e405-e406.
Int J Radiat Oncol Biol Phys
Purpose/Objective(s): To investigate the potential predictive value of patient clinical factors versus radiomic features extracted from CT images for characterization of Human Papilloma Virus (HPV) status for patients with oropharyngeal squamous cell carcinoma (OPSCC).
Materials/Methods: One hundred twenty-eight OPSCC patients with known HPV-status (60-HPV+ and 68-HPV-, confirmed by immunohistochemistry-P16-protein testing) were retrospectively studied. Radiomic features (11 feature-categories) were extracted in 3D from contrast-enhanced (CE)-CT images of gross-tumor-volumes using ‘in-house’ software (ROdiomX) developed and validated following the image-biomarker-standardization-initiative (IBSI) guidelines. Six categories of clinical factors were investigated: Age-at-Diagnosis, Gender, Total-Charlson comorbidity score, Alcohol-Use, Smoking-History, and T-Stage, according to AJCC 7th-edition. An Elastic Net technique combined with a Generalized-Linear-Model (Lasso-GLM) were applied to perform L1 and L2 regularizations in the radiomic and clinical feature spaces to identify and rank the optimal feature subsets with most representative information for prediction of HPV. Elastic-Net GLM classifiers based on clinical factors only, radiomics only, and combined clinical and radiomics (ensemble/integrated) were constructed using random-permutation-sampling. Tests of significance (One-way ANOVA), average Area-Under-Receiver-Operating-Characteristic (AUC), and Positive and Negative Predictive values (PPV and NPV) were computed to estimate the generalization-error and prediction performance of the classifiers.
Results: Five clinical factors, including T-stage, smoking status, and age; and 21 radiomic features, including tumor morphology, textural information, and intensity contrast were found to be statistically significant discriminators between HPV positive and negative cohorts. Performances for prediction of HPV status for the 3 classifiers were: Radiomics Elastic-Net-GLM: AUC/PPV/NPV = 0.799/0.775/0.802; Clinical Elastic-Net-GLM: 0.673/0.738/0.679, and Integrated/Ensemble Elastic-Net-GLM: 0.912/0.884/0.860. Results imply that the radiomics-based classifier significantly outperforms that using clinical factors only, and that the combination of both radiomics and clinical factors yields even higher predictive performance.
Conclusion: Albeit subject to confirmation in a larger cohort, this pilot study presents encouraging results in support of the role of radiomic features towards characterization of HPV in patients with OPSCC.