Multi-Dimensional Biomarkers using Integrated Molecular and Radiomic Data for Precision Head and Neck Oncology


  1. Perform molecular profiling on a large bank of head and neck cancer tissues, including tissue accumulated as part of multiple clinical trials.
  2. Integrate the molecular profile with image-extracted features or image-based “omics” commonly referred to as radiomics that will include both hand engineered and deep extracted features from a patient’s cross-sectional imaging scans.
  3. Use different machine learning approaches for predicting different clinical outcomes of interest, as well as deriving models for predicting certain molecular signatures and therapy response non-invasively using image-based features.



Head and neck cancers constitute a heterogeneous group of malignancies and are the sixth most common malignancy worldwide. In head and neck squamous cell carcinoma (HNSCC), other than incorporation of the molecular marker HPV, current staging and treatment planning is largely based on anatomic factors. The absence of more specific biomarkers poses a significant challenge for selection of optimal and timely therapy for our patients, including novel therapeutic options such as immunotherapy. Considering the diversity and variations in the genetic landscape, molecular pathogenesis, risk factors and treatment response that are encountered across different tumors and anatomic sites, there is a critical need for biomarkers that enable better tumor stratification and provide the basis for more precise and personalized therapies. This project will provide much needed additional insight in tumor pathogenesis that can be the basis for precision care, coupled with molecular and non-invasive image-based biomarkers for patient selection and treatment response prediction. Furthermore, the fundamental data analytic pipelines and approaches that will be developed will have broad utility to other cancer types and pathology. We believe that a project of this scale combining different types of omics, including integration of molecular and image-derived features overseen by domain experts, has transformative potential for precision personalized patient care. The combination of state-of-the-art molecular analysis with medical imaging, leveraging artificial intelligence or more specifically machine learning for developing classifiers, will be used to predict different outcomes of interest and develop tumor specific biomarkers.