Using Deep Learning to Predict Presence of Oropharyngeal Cancer with the RADCURE Dataset
Emory Adelman, Nehemiah Setiawan, Andrew Dahlstrom
Department of Biology
Faculty Supervisor: Sara El Alaoui
Using the DICOM dataset found in the cancer imaging archive RADCURE, we will examine medical image data to predict oropharyngeal cancer in subjects. This dataset consists of 3,346 head and neck cancer CT images collected from 2005-2017. Demographic, clinical, and treatment information is also in this dataset in a comma-separated value format. The dataset contains images with labels such as normal and non-normal tissue contours. Individual contour names were set, with "primary tumor volumes" and "19 organs-at-risk". Using the pre-labeled data, we will make a new deep-learning model. Our goal is to make a model proficient at prognostic diagnosis of cancers of the head and neck region to be used in a clinical setting to cross-check diagnosis.