摘要
应用傅里叶变换近红外光谱透射技术结合偏最小二乘法(PLS)建立了人血清中7种生化成分的定标模型,利用内部交叉验证和自动优化功能对定标模型进行了优化,确定了最优建模参数。模型对人血清中总胆固醇、甘油三酯、总蛋白、白蛋白、载脂蛋白B、低密度脂蛋白胆固醇、葡萄糖定标样品集的预测值与化学值的相关系数r分别为0.9011、0.9593、0.9249、0.761、0.8831、0.5191、0 9148,预测校正标准误差RMSECV分别为15mg/dL,21.6mg/dL,2 66g/L,3 96g/L,0.091g/L,16.2mg/dL,0.49mmol/L。
<Abstrcat>Seven biochemical compounds in human serum were analyzed rapidly using Fourier transform near infrared transmittance spectroscopy(NIRS) combined with partial least square regrassion(PLS). The optimum parameters of mathematics models for them were studied using leave-one-out cross validation method. The correlation coefficients(r) between the true values and the NIRS prediction values for total cholesterol, glycerinate, total protein, albumin, apolipoprotein B, low-density lipoprotein(LDL) cholesterol and glucose of the calibration sample sets were 0.9011, 0.9593, 0.9249, 0.761, 0.8831, 0.5191, 0.9148 and the root mean square errors of cross validation (RMSECV) for them were 15.0 mg/dL, 21.6 mg/dL, 2.66 g/L, 3.96 g/L, 0.091 g/L, 16.2 mg/dL,0.49 mM/L respectively.
出处
《分析试验室》
CAS
CSCD
北大核心
2005年第7期17-20,共4页
Chinese Journal of Analysis Laboratory
基金
中国科学院长春光学精密机械与物理研究所应用光学国家重点实验室开放基金
吉林省科技发展计划项目(20040324 2)
关键词
近红外光谱
检测
人血清
生化成分
Near-infrared spectroscopy(NIRS)
Determination
Human serum
Biochemical components
