论文部分内容阅读
应用激光镊子拉曼光潜(LTRS)技术,俘获单个红细胞并收集其拉曼光谱。将主成分分析(PCA)算法和反向传播BP网络预测模型相结合,进行地中海贫血(简称地贫)红细胞类型的判别。PCA结果显示,正常对照与中间型α-地贫(HbH-CS)基本可区分,但正常对照与重型β-地贫,HbH-CS与重型β-地贫间差异不明显。将归一化处理的前5个主成分进行BP网络训练及预测,结果发现,正常对照与HbH-CS间预测正确率高达97.90%,正常对照与重型β-地贫,HbH-CS与重型β-地贫间预测正确率分别为90.72%和86.28%。该结果与平均拉曼光谱及主成分分析结果基本吻合。取不同的实验条件下收集的光谱进行同样的分析,3种组合的预测正确率略有不同,分别为95.28%,92.08%,91.85%,但呈现基本相同的规律。
Laser Raman spectroscopy (LTRS) was used to capture single RBCs and collect their Raman spectra. The principal component analysis (PCA) algorithm and backpropagation BP network prediction model combined to determine the type of thalassemia (thalassemia) red blood cells. PCA results showed that the normal control and intermediate alpha-thalassemia (HbH-CS) basically distinguishable, but no significant difference between the normal control and severe β-thalassemia, HbH-CS and severe β-thalassemia. The first five principal components of the normalized BP neural network training and prediction, and found that the normal control and HbH-CS prediction accuracy as high as 97.90%, normal control and severe β-thalassemia, HbH-CS and heavy beta The correct prediction rates of thalassemia intervertebrates are 90.72% and 86.28% respectively. The results are in good agreement with the average Raman spectra and principal component analysis. Taking the spectra collected under different experimental conditions for the same analysis, the prediction accuracy rates of the three combinations were slightly different, which were 95.28%, 92.08% and 91.85%, respectively, but showed basically the same laws.