Synthetic CT generation based on CBCT using respath‐cycleGAN

• DOI: 10.1002/mp.15684
• Publication Date: 2022-04-30T03:49:57Z
• Authors: Liwei Deng; Jie Hu; Jing Wang; Sijuan Huang; Xin Yang
• Keywords: 03 medical and health sciences; 0302 clinical medicine; Radiotherapy Planning, Computer-Assisted; Image Processing, Computer-Assisted; Humans; Radiotherapy Dosage; Spiral Cone-Beam Computed Tomography; Cone-Beam Computed Tomography; Signal-To-Noise Ratio; Artifacts

Cone-beam computed tomography (CBCT) plays an important role in radiotherapy, but the presence of a large number artifacts limits its application. The purpose this study was to use respath-cycleGAN synthesize CT (sCT) similar planning (pCT) from CBCT for future clinical practice.The method integrates respath concept into original cycleGAN, called respath-cycleGAN, map pCT. Thirty patients were used training and 15 testing.The mean absolute error (MAE), root square (RMSE), peak signal noise ratio (PSNR), structural similarity index (SSIM), spatial nonuniformity (SNU) calculated assess quality sCT generated CBCT. Compared with images, MAE improved 197.72 140.7, RMSE 339.17 266.51, PSNR 22.07 24.44, while SSIM increased 0.948 0.964. Both visually quantitatively, is superior without respath. We also performed generalization test head-and-neck (H&N) model on pelvic data set. results again showed that our superior.We developed good In practice, may be develop radiotherapy plans.