First, we pretrained the encoder of a transformer-based network using a self-supervised approach on unlabeled abdominal computed tomography images. Subsequently, we fine-tuned the segmentation network ...