Free-form tumor synthesis in computed tomography images via richer generative adversarial network

作者：

Highlights：

• The first work to support user-customized tumor synthesis using inpainting model.

• A novel richer-convolution enhanced dilated–gated GAN to achieve the task.

• A new hybrid loss function for authentic tumor synthesis.

• Comprehensive evaluation on the wide range of public 3D tumor/lesion datasets.

摘要

•The first work to support user-customized tumor synthesis using inpainting model.•A novel richer-convolution enhanced dilated–gated GAN to achieve the task.•A new hybrid loss function for authentic tumor synthesis.•Comprehensive evaluation on the wide range of public 3D tumor/lesion datasets.

论文关键词：Medical image synthesis,dilated–gated convolution,Generative adversarial network,Richer convolutional feature,3D free-form synthesis

论文评审过程：Received 17 September 2020, Revised 28 December 2020, Accepted 5 January 2021, Available online 9 February 2021, Version of Record 19 February 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.106753