Synthetic Tumor Manipulation: With Radiomics Features
This work addresses the need for generating unlimited realistic synthetic tumors to advance medical imaging research and potential clinical applications, representing an incremental improvement by combining existing methods like GANs with radiomics conditioning.
The authors tackled the problem of generating realistic synthetic tumors with detailed control over subregions by introducing RadiomicsFill, a generator conditioned on radiomics features, and demonstrated its capability to produce diverse, realistic tumors and fine-tune specific features like 'Pixel Surface' and 'Shape Sphericity' in glioma patient experiments.
We introduce RadiomicsFill, a synthetic tumor generator conditioned on radiomics features, enabling detailed control and individual manipulation of tumor subregions. This conditioning leverages conventional high-dimensional features of the tumor (i.e., radiomics features) and thus is biologically well-grounded. Our model combines generative adversarial networks, radiomics-feature conditioning, and multi-task learning. Through experiments with glioma patients, RadiomicsFill demonstrated its capability to generate diverse, realistic tumors and its fine-tuning ability for specific radiomics features like 'Pixel Surface' and 'Shape Sphericity'. The ability of RadiomicsFill to generate an unlimited number of realistic synthetic tumors offers notable prospects for both advancing medical imaging research and potential clinical applications.