Radiomic analysis has exponentially increased the amount of quantitative data extractable from a single medical image. However, the effect of various image acquisition conditions on the reproducibility or robustness of these features is understudied. Specifically, when generating a predictive model to be used in a multi-institutional setting, it must be robust to voxel size changes. This study aims to develop a task-specific robustness preselection step for incorporation into radiomics pipeline to improve the generalizability of a model applied to a testing set of dissimilar resolution.
This abstract and the presentation materials are available to members only; a login is required.