A Multi-Modal Risk Score to Predict Progression of Oral Premalignant Lesions to Oral Squamous Cell Carcinoma

Oral cavity squamous cell carcinoma (OSCC) incidence is increasing worldwide. A significant number of patients have oral premalignant lesions (OPL) prior to their cancer diagnosis, and patients with OPL have a 45x increased risk of developing OSCC. However, there are significant challenges in risk stratifying these lesions. Currently, risk stratification is based on dysplasia grade determined by traditional histologic review of hematoxylin and eosin-stained slides, but this has poor prognostic value as interrater agreement among pathologists is low. The purpose of this study is to validate a multi-stage, deep learning histologic and epigenetic biomarker to identify OPLs with high risk of progression to invasive OSCC.