Open Collections

Dubljevic, Natalia

Radiomics is a term that refers to the extraction of features that provide quantitative information from medical images. These radiomics features capture information about tissues and lesions such as shape and heterogeneity. They act as imaging biomarkers, have been shown to have the ability to predict the biological characteristics of lesions, and can assess patient prognosis or response to treatment. However, determining features with high predictive power and robustness is a difficult task. It is hypothesized that utilizing features calculated under a range of perturbations may provide improved predictive power. This allows the predictive model access to a large range of features of which the most beneficial may be leveraged. Furthermore, new information may be introduced by the change of a feature under a given parameter. This approach is referred to as ‘tensor radiomics’. Three implementations given the task of predicting head and neck cancer patient outcome were tested using discretization bin sizes as the varying parameter. Initial results are inconclusive, and further experimentation is required to give a comprehensive assessment of tensor radiomics.

Text

University of British Columbia. PHYS 449

Vancouver : University of British Columbia Library

10.14288/1.0413121

Science, Faculty of; Physics and Astronomy, Department of

Undergraduate

DSpace