Radiomics is the use of tumor texture, as seen in pre-treatment computed tomography, positron emission tomography or other images, to understand important information about individual tumors, such as response to treatment. We have demonstrated that it is possible to use pre-treatment images to categorize patients as low-risk (good survival) and high-risk (poor survival). Additionally, we have some very exciting data that shows that radiomics approaches can predict whether increasing the radiotherapy dose will improve or reduce the individual patient's overall survival. However, before these radiomics model scan be implemented clinically, validation in independent patient datasets is essential. One big hurdle to this is the fact that patients are not all imaged on a single CT scanner (or PET scanner, etc.), but on a wide range of different scanners (different manufacturers, models, etc.), and the calculated value of tumor texture can be affected by which scanner is used to image patient. This means that before we can properly validate radiomics models, and apply them to real-world situations (meaning outside of the well-controlled, single-institution trial setting), it is important to understand the magnitude of these variabilities, and the impact they have on radiomics models. This knowledge will help direct future research to minimize their impact on the creation, independent validation, and future use of radiomics models. This work is relevant to many different treatment types, including chemo- radiotherapy, immunotherapy, etc.
The proposed project is relevant to public health because it provides data on uncertainties in quantitative image features extracted from CT images taken on multiple CT scanners, and assesses the impact of these on radiomics models for patient risk stratification. This understanding is vital as we try to optimize radiomics models using large image sets and then validate using independent datasets. The results are translatable to different treatment approaches, including radiotherapy, chemotherapy, surgery and immunotherapy.
| Ger, Rachel B; Craft, Daniel F; Mackin, Dennis S et al. (2018) Practical guidelines for handling head and neck computed tomography artifacts for quantitative image analysis. Comput Med Imaging Graph 69:134-139 |
| Mackin, Dennis; Ger, Rachel; Dodge, Cristina et al. (2018) Effect of tube current on computed tomography radiomic features. Sci Rep 8:2354 |
| Ger, Rachel B; Zhou, Shouhao; Chi, Pai-Chun Melinda et al. (2018) Comprehensive Investigation on Controlling for CT Imaging Variabilities in Radiomics Studies. Sci Rep 8:13047 |
