结直肠癌肿瘤标志物多分割点设置算法  

Multi-cutoff level setting algorithm for various tumors markers in colorectal cancer

作　　者：施京华[1] 苏强[1] 顾平[2] 黄钢[2] 朱岩[3] 

机构地区：[1]上海交通大学工业工程与物流工程系,上海200240 [2]上海交通大学医学院附属仁济医院,核医学科,上海200001 [3]清华大学经济管理学院,北京100084

出　　处：《清华大学学报（自然科学版）》2009年第5期777-780,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(70672077);上海交通大学医工(理)交叉研究基金重点项目(YG2007ZD10)

摘　　要：为了提高诊断正确率,大量医学研究集中于探讨肿瘤标志物正常值与非正常值的合适划分,即确定分割点。为此,该文提出了一种肿瘤标志物多分割点设置算法。通过正交实验和缩小搜索空间的方法,进一步提高了诊断正确率和优化收敛速度。在结直肠癌实际诊断数据中的应用效果表明,该算法能有效地提高诊断正确率,其诊断结果优于医学上的单分割点的诊断结果。There have been many studies to determine the best cutoff levels for tumor markers to differentiate between normal and abnormal values. This paper presents a multi-cutoff level setting algorithm for various tumors markers to improve diagnostic accuracy. The accuracy and convergence of the algorithm are further improved using an orthogonal experiment and limiting the search space. Experiments demonstrate that this algorithm effectively improves the diagnostic accuracy for eolorectal cancer by various tumor markers and is better than the single cutoff level normally used in medicine.

关 键 词：粗糙集 遗传算法 结直肠癌 肿瘤标志物 诊断正确率 正交实验 

分 类 号：R735.3[医药卫生—肿瘤] TP18[医药卫生—临床医学]