主成分分析结合马氏距离法剔除深静脉血栓红外光谱异常样品

Rejection of Abnormal Deep Venous Thrombosis Infrared SpectraSamples by Principal Component Analysis Combined with Martingale Distance Method

推断深静脉血栓(deep venous thrombosis,DVT)形成时间一直是法医学中的热点和难点问题。傅里叶变换红外光谱(Fourier transform infrared,FTIR)具有快速、高敏、无损等特点,为DVT形成时间推断提供了新的思路。采用FTIR测定大鼠DVT红外光谱数据,立足于血栓光谱数据建模分析的角度,采用主成分分析(Principal Component Analysis,PCA)结合马氏距离法进行异常样品的筛选和剔除,探究该方法对血栓红外光谱建模能力的提升效能。结果表明利用PCA结合马氏距离法能够有效地筛选剔除血栓组织红外光谱样品集中存在的异常样品,提高血栓红外光谱建模的预测能力。当筛选权重为1时,筛选剔除异常样品后的模型预测精度最优。为DVT的红外光谱分析提供了方法参考,为利用FTIR技术推断大鼠DVT形成时间奠定了理论基础。

Inferring the formation time of deep venous thrombosis(DVT)has been a hot and difficult problem in forensic medicine.Fourier transform infrared(FTIR)spectroscopy,which is fast,highly sensitive and non-destructive,provides a new idea for inferring the time of DVT formation.FTIR was used to determine the infrared spectral data of rat DVT,and based on the perspective of modeling and analysis of thrombus spectral data,Principal Component Analysis(PCA)combined with the martingale distance method was used for screening and rejection of abnormal samples to investigate the effectiveness of this method in improving the model ing ability of thrombus infrared spectra.The results showed that PCA combined with the distance method could effectively screen out the abnormal samples in the thrombus infrared spectroscopy sample set and improve the prediction ability of thrombus infrared spectroscopy modeling.When the screening weight was 1,the model prediction accuracy after screening out the abnormal samples was optimal.It provides a methodological reference for the infrared spectroscopic analysis of DVT and lays a theoretical foundation for inferring the formation time of DVT in rats using FTIR technique.