Prediction of human papillomavirus status patients with oropharyngeal squamous cell carcinoma using machine learning methods

In head and neck squamous cell carcinomas (HNSCC), accurate determination of HPV status plays a critical role in both treatment planning and clinical decision-making processes. Particularly in oropharyngeal cancers, HPV status can significantly influence the staging of the disease. Although several laboratory tests are available for determining HPV status, they are often costly and time-consuming. Detection of E6/E7 mRNA from biopsy samples is considered the gold standard; however, this method is not feasible in clinical practice due to procedural complexity and resource requirements. Furthermore, while polymerase chain reaction (PCR)-based techniques are frequently employed, their sensitivity is high, yet specificity remains relatively low. HPV-positive HNSCC cases, which are associated with better prognosis and therapeutic response, necessitate the implementation of personalized treatment strategies. Clinicians face a challenging decision-making process in balancing the oncological benefits of aggressive surgical interventions with their potential morbidity risks. In this thesis, an artificial intelligence-based approach is proposed to predict HPV status. The study evaluates modern machine learning models—namely CatBoost, LightGBM, and XGBoost—which are known for their tolerance to missing data, using a limited dataset of patient records containing irregularities and incomplete values. Additionally, genetic algorithm-based hyperparameter optimization was employed to enhance model performance. The findings indicate that, despite structural limitations within the dataset, the proposed AI models possess substantial potential for accurately predicting HPV status. This research also highlights the feasibility of integrating artificial intelligence into oncology-focused decision support systems, suggesting that more advanced implementations may offer significant contributions to clinical care pathways in the future.systems in the field of oncology and that advanced applications can contribute to patient care processes.