Fig. 1.

Comparative analysis of planning CT (pCT), CBCT, synthetic CTs generated by UNet and ResUNet using non-adversarial and adversarial training, and the corresponding difference images relative to pCT. Figures a), b), and c) represent the best, median, and worst prediction results based on the image quality assessment. All images are displayed within the window of (-300, 500) HU. The yellow horizontal lines on the pCT indicate the locations from which the CT number line profiles were extracted (refer to supplementary figures SF6 and SF7 for details).