汉译英：主要挑战包括面积变化以及CT图像中Loin区域和背景之间没有可见边界。我们研究中的内脏基准Ground Truth由一位专家使用Python语言开源软件库中名为LabelMe的半自动分割工具手动注释并根据其特征和语义信息对CT床和所有内脏组织进行二进制掩码注释。Ground Truth由背景的黑色像素和Loin的白色像素组成。CT图像样本及其相应注释如图4所示。

The main challenges include changes in area and lack of visible boundaries between the Loin region and the background in CT images. The ground truth in our study was manually annotated by an expert using a semi-automatic segmentation tool called LabelMe in Python open-source software library. The CT bed and all visceral tissues were annotated with binary masks based on their features and semantic information. The ground truth consisted of black pixels for the background and white pixels for the Loin. The CT image samples and their corresponding annotations are shown in Figure 4.