Fig. 12

Illustration from top to bottom represents original CT images, ground truth, segmentation results by the proposed segmentation framework without \({\mathcal {L}}_b\) and with \({\mathcal {L}}_b\), respectively. Images a–e refer to five samples, respectively