Keywords: Radiology, throat cancer, oncology, head-neck, prediction models, image analysis, treatment outcome

''Magnetic resonance based radiomics in oropharyngeal cancer''

Currently, patients with throat cancer are treated with radiotherapy combined with chemotherapy. This therapy is successful only in 55-75% of cases. The success of the treatment is partly explained by the size of the local tumor and the metastasis of the tumor to the lymph nodes and other parts of the body. However, these disease characteristics cannot fully explain treatment success, so other reliable "biomarkers" are needed to predict response to treatment.

This dissertation investigated whether quantitative features obtained from MRI imaging can be used as biomarkers to predict treatment outcome in throat cancer patients. Results show that it was possible to predict treatment outcome in advance and that the models are applicable across multiple (multicenter) patient datasets. However, extraction of these quantitative image features requires delineation of the tumor by an experienced radiologist, which is an intensive and time-consuming task. Therefore, this dissertation also explored the possibility of whether simpler methods to delineate the tumor region and methods that fully automatically recognize the tumor region are applicable to advance clinical implementation.

Click here for the full dissertation. 

Click here for the live stream.