中文摘要：

针对手动分割颈部淋巴结的局限性,提出一种在较少人为干预下,准确批量分割颈部淋巴结的算法。采用遍历阈值算法提取疑似淋巴结种子点,加入基于统计学的强制停止条件,对种子点进行区域生长,依据淋巴结和周围软组织生理特征及其图像特性,提出颈部淋巴结判决算法。实验结果表明,颈部淋巴结体积的正确分割比TPVP为94.00%,阳性预测值PV为97.11%,无漏判,精确度达86.33%,该算法能实现颈部淋巴结精确批量分割,为临床诊断、治疗和预后提供有效辅助依据。

英文摘要：

To solve the limitation of manual segmentation,the algorithm segmenting the neck lymph nodes in batches accurately with less human intervention was proposed.Candidate lymph node seeds were detected using traversal thresholding algorithm.Region growing method was used for detected seeds with a hard statistic-based constraint as stop criteria.Considering the physiological characteristics and image properties of lymph node and soft tissue,an algorithm was proposed to accurately identify the lymph nodes.Experimental results verify that the correct segmentation ratio of lymph nodes volume（TPVP）is 94.00%,the positive predictive value（PV）is 97.11%.All lymph nodes are detected,and the detection accuracy is 86.33%.The proposed semi-automatic segmentation system can provide effective diagnosis,treatment and prognosis aid to clinicians.