FTIR光谱结合数据挖掘方法构建死亡时间推断数学模型  

Establishment of the Mathematical Model for PMI Estimation Using FTIR Spectroscopy and Data Mining Method

作　　者：王磊 秦新潮 林汉成 邓恺飞 罗仪文 孙其然 杜秋香[1] 王振原 托娅 孙俊红[1] WANG Lei;QIN Xin-chao;LIN Han-cheng;DENG Kai-fei;LUO Yi-wen;SUN Qi-ran;DU Qiu-xiang;WANG Zhen-yuan;TUO Ya;SUN Jun-hong(School of Forensic Medicine, Shanxi Medical University, Taiyuan 030001, China;Shanghai Key Laboratory of Forensic Medicine, Shanghai Forensic Service Platform, Academy of Forensic Science, Shanghai 200063, China;School of Basic Medical Science, Shanghai University of Medicine and Health Sciences, Shanghai 201318, China;Linwei Branch of Weinan Public Security Bureau, Weinan 714000, China;Department of Forensic Science, Health Science Center, Xi’an Jiaotong University, Xi’an 710061, China)

机构地区：[1]山西医科大学法医学院,山西太原030001 [2]司法鉴定科学研究院,上海市法医学重点实验室,上海市司法鉴定专业技术服务平台,上海200063 [3]上海健康医学院基础医学院,上海201318 [4]渭南市公安局临渭分局,陕西渭南714000 [5]西安交通大学医学部法医学院,陕西西安710061

出　　处：《法医学杂志》2018年第1期1-6,12,共7页Journal of Forensic Medicine

基　　金：“十三五”国家重点研发计划资助项目(2016YFC0800702);国家自然科学基金资助项目(81571852,81601645);上海市法医学重点实验室资助项目(17DZ2273200);上海市司法鉴定专业技术服务平台资助项目(16DZ2290900)

摘　　要：目的利用傅里叶变换红外(Fourier transform infrared,FTIR)光谱技术结合数据挖掘方法分析死后大鼠脾组织FTIR光谱与死亡时间的关系,推断大鼠死亡时间。方法大鼠脱臼处死,尸体置于20℃环境中,于不同时间点取大鼠脾组织,采集FTIR检测数据,数据预处理后应用数据挖掘方法进行分析。结果大鼠脾组织光谱吸收峰强随死亡时间延长发生变化,峰位没有改变;主成分分析结果示前三个主成分累积贡献率为96%,各时间点光谱样本具有明显聚类趋势;偏最小二乘判别分析和支持向量机分类方法可将不同死亡时间光谱样本进行有效四分类(0~24h、48~72h、96~120h和144~168h);偏最小二乘回归分析构建的死亡时间推断模型决定系数(R^2)为0.96,校正均方根误差和交叉验证均方根误差分别为9.90h和11.39h,预测集R^2达到0.97,预测均方根误差为10.49h。结论 FTIR光谱技术结合数据挖掘方法可对大鼠脾组织进行有效定性和定量分析,可建立分类判别和偏最小二乘回归模型,对死亡时间进行准确推断。Objective To analyse the relationship between Fourier transform infrared(FTIR)spectrum of rat’s spleen tissue and postmortem interval(PMI)for PMI estimation using FTIR spectroscopy combined with data mining method.Methods Rats were sacrificed by cervical dislocation,and the cadavers were placed at20℃.The FTIR spectrum data of rats’spleen tissues were taken and measured at different time points.After pretreatment,the data was analysed by data mining method.Results The absorption peak intensity of rat’s spleen tissue spectrum changed with the PMI,while the absorption peak position was unchanged.The results of principal component analysis(PCA)showed that the cumulative contribution rate of the first three principal components was96%.There was an obvious clustering tendency for the spectrum sample at each time point.The methods of partial least squares discriminant analysis(PLS-DA)and support vector machine classification(SVMC)effectively divided the spectrum samples with different PMI into four categories(0-24h,48-72h,96-120h and144-168h).The determination coefficient(R2)of the PMI estimation model established by PLS regression analysis was0.96,and the root mean square error of calibration(RMSEC)and root mean square error of cross validation(RMSECV)were9.90h and11.39h respectively.In prediction set,the R2was0.97,and the root mean square error of prediction(RMSEP)was10.49h.Conclusion The FTIR spectrum of the rat’s spleen tissue can be effectively analyzed qualitatively and quantitatively by the combination of FTIR spectroscopy and data mining method,and the classification and PLS regression models can be established for PMI estimation.

关 键 词：法医病理学 谱学 傅里叶变换红外 数据挖掘 主成分分析 支持向量机 死亡时间 大鼠 

分 类 号：DF795.1[医药卫生—法医学]