Prognosis Prediction in Head and Neck Squamous Cell Carcinoma by Radiomics and Clinical Information

Area of Science:

• Oncology
• Radiology
• Machine Learning

Background:

• Head and neck squamous cell carcinoma (HNSCC) is a complex cancer with variable patient outcomes.
• Prognosis in HNSCC is influenced by tumor heterogeneity.
• Radiotherapy (RT) is a common treatment modality for HNSCC.

Purpose of the Study:

• To predict 5-year overall survival in HNSCC patients receiving radiotherapy.
• To integrate radiomic and clinical data using machine learning for improved prognostic accuracy.
• To compare the performance of combined models against radiomic-only and clinical-only models.

Main Methods:

• Utilized HNSCC radiotherapy planning CT images and clinical data from The Cancer Imaging Archive.
• Extracted radiomic features and analyzed clinical data independently using five machine learning algorithms.
• Developed a probability-weighted enhanced model (PWEM) by ensembling radiomic and clinical models.

Main Results:

• The study analyzed 299 HNSCC cases.
• The PWEM achieved an Area Under the Curve (AUC) of 0.86, outperforming individual radiomic and clinical models.
• T stage, age, and disease site were identified as key clinical predictors of prognosis.

Conclusions:

• A combined radiomic-clinical machine learning model offers superior performance for predicting HNSCC survival.
• This integrated approach surpasses the predictive power of models based on radiomic or clinical factors alone.
• Further validation with larger prospective studies is recommended for clinical implementation.