中文摘要：

目的：针对腹部CT序列图像因邻近器官对比度低以及肝脏形状不一致等造成的肝脏分割困难问题，提出一种基于超限学习机的腹部CT序列图像肝脏自动分割方法。方法：首先，在预处理阶段利用阈值法和形态学操作去除肌肉、脂肪、肋骨和脊椎；然后，对预处理结果图像求取均值、标准差和距离变换，提取有效的训练特征；最后，将3个特征归一化处理后作为超限学习机的输入层，并进行学习最终得出分割结果。结果：通过对9个腹部CT序列图像进行肝脏分割实验，并与其他3种方法进行比较，本文算法具有明显优势。结论：本文算法能对腹部CT序列图像中的肝脏进行准确有效地分割。

英文摘要：

Objective To solve the difficulty of liver segmentation for abdominal computed tomography （CT） sequence images caused by low contrast of adjacent organs and different liver shapes by proposing an automatic liver segmentation for abdominal CT sequence images based on extreme learning machine （ELM）. Methods The threshold value method and morphological operation were firstly applied to remove muscle, fat, ribs and spine in the preprocessing stage. And then, the mean value, standard deviation and distance transform of the preprocessed image were calculated to extract the effective training features. Finally, these three features processed by normalization were applied as the input layer of ELM, and the segmentation results were obtained by learning. Results Liver segmentation was carded on nine abdominal CT sequence images. And compared with the other three methods, the proposed method had obvious advantages. Conclusion The proposed method can segment livers in abdominal CT sequence images accurately and effectively.