KNN方法在癌症中红外光谱检测中的应用  

Application of KNN Method to Cancer Diagnosis Using Fourier-Transform Infrared Spectroscopy

作　　者：李响[1] 李庆波[1] 徐怡庄[2] 张广军[1] 吴瑾光[2] 杨丽敏[2] 凌晓锋[3] 周孝思[3] 王健生[4] 

机构地区：[1]北京航空航天大学仪器科学与光电工程学院,北京100083 [2]北京大学化学与分子工程学院稀土材料化学及应用国家重点实验室,北京100871 [3]北京大学第三医院普外科,北京100083 [4]西安交通大学第一医院外科,陕西西安710061

出　　处：《光谱学与光谱分析》2007年第3期439-443,共5页Spectroscopy and Spectral Analysis

基　　金：国家"863-703"专题资助;航天科技创新基金;教育部博士点基金资助

摘　　要：红外光谱主要是研究分子中以化学键联结的原子之间的振动光谱,它能够在分子水平上揭示正常组织和癌组织之间存在的差异。文章利用化学计量学中的有关知识通过计算机自动对未知样本光谱进行判别分析。首先应用平滑处理,基线校正(SNV方法),奇异值剔除(RHM方法)等算法对光谱数据进行预处理,然后采用K-最近邻法(简称KNN法)实现未知样本的自动判别,提高了癌症红外光谱检测的准确度。文章对63例胃组织样品进行了傅里叶变换红外光谱判别分析,与病理检验结果比较,准确度达到91.7%。Early diagnosis and early medical treatments are the keys to save the patients＇ lives and improve their living quality. Fourier transform infrared （FTIR） spectroscopy can be used to distinguish malignant from normal tissues at the molecular level. In the present paper, programs were made with chemometrics method of pattern recognition to classify unknown tissue samples. Spectral data were pretreated by using smoothing, SNV and RHM method. Cross validation was used to test the discrimination effect of KNN method. A total of 63 gastric tissue samples were employed in this study, including 26 cases of normal tissue samples and 37 cases of cancerous tissue samples. The recognition results of the KNN method showed that the correctness of classification achieved 91.7%.

关 键 词：KNN方法 癌症 FTIR 模式识别 

分 类 号：R730.4[医药卫生—肿瘤]