对近年来采用多种光谱学方法对癌诊断研究工作进行了简要综述。 利用电感耦合等离子体原子发射光谱分析(ICP-AES)法对人血清中微量元素含量进行了测定, 将双向联想记忆神经网络(BAM)用于建立微量元素与肺癌、 肝癌和胃癌之间的关系, 建立了分类鉴定模型; 利用近红外光谱技术非破坏无损检测的特性, 采用化学计量学方法对子宫内膜癌组织的近红外光谱(NIRS)进行特征变量提取, 用模糊规则专家系统建立诊断模型, 取得了满意的结果; 建立了一种基于粒子群优化(PSO)方法对近红外光谱变量进行选择, 然后采用支持向量机(SVM)建立诊断模型; 太赫兹时域光谱(THz-TDS)技术是近年来受到重视的一种非破坏无损检测方法, 由于其辐射能量低, 因此对于生物样品检测具有很好的发展前景。 采用THz-TDS技术对子宫颈癌组织进行测试, 对这些癌组织的THz光谱采用导数光谱、 正交信号校正(OSC)等光谱预处理方法减少干扰成分和变量选择后, 分别采用模糊规则专家系统(FuRES)、 模糊优化联想记忆网络(FOAM)、 支持向量机(SVM)和偏最小二乘判别分析(PLS-DA)等方法建立分类模型, 均取得了较满意的结果。 这些研究结果为研究癌的发生和发展提供了有用的信息, 为癌的早期诊断提供了新的方法。

Studies on cancer diagnosis using various spectroscopic methods combined with chemometrics are briefly reviewed. Elemental contents in serum samples were determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES), bidirectional associative memory (BAM) networks were used to establish diagnosis models for the relationships between elemental contents and lung cancer, liver cancer, and stomach cancer, respectively. Near infrared spectroscopy (NIRS) is a non-destructive detection technology. Near infrared spectra of endometrial carcinoma samples were determined and spectral features were extracted by chemoometric methods, a fuzzy rule-based expert system (FuRES) was used for establishing diagnosis model, satisfactory results were obtained. We also proposed a novel variable selection method based on particle swarm optimization (PSO) for near infrared spectra of endometrial carcinoma samples. Spectra with optimized variable were then modeled by support victor machine (SVM). Terahertz technology is an emerging technology for non-destructive detection, which has some unique characteristics. Terahertz time domain spectroscopy (THz-TDS) was used for cervical carcinoma measurement. Absorption coefficients were calculated from the measured time domain spectra and then processed with derivative, orthogonal signal correction (PC-OSC) to reduce interference components, and then fuzzy rule-based expert system (FuRES), fuzzy optimal associative memory (FOAM), support victor machine (SVM), and partial least squares discriminant analysis (PLS-DA) were used for diagnosis model establishment. The above results provide useful information for cancer occurring and development, and provide novel approaches for early stage diagnosis of various cancers.