Novel data learning framework for integrating radiology and pathology data for discovering prognostic biomarkers and predicting outcomes in head and neck cancer.
Head and neck cancer (HNC) is the seventh most common cancer in the world, with 1.1 million new diagnoses reported annually. In the US, the incidence is over 54,000 cases per year, resulting in over 11,000 annual deaths. HNC often spreads to lymph nodes in the neck, which may result in multiple regions of importance for radiology and pathology imaging. However, most data learning models for cancer diagnosis process medical images independently and are targeted at one specific region. Accurate grade classification and biomarker identification of HNC may inform clinicians as to the risk level of patients and inform treatment routes.
Researchers at Emory developed Swin Transformer-based MultiModal and Multi-Region Data Fusion Framework (SMuRF) to help predict outcomes in head and neck cancer. This software can simultaneously process images from multiple anatomical sites, such as the primary tumor and associated lymph nodes, at the micro- and macro- scales. SMuRF integrates these multi-modal and multi-scale images for risk stratification and to identify specific regions of high prognostic value.