骨关节炎软骨的近红外光谱学研究及分期诊断

NIR Spectroscopic Study and Staging Diagnosis of Osteoarthritic Articular Cartilage

骨关节炎是一种威胁中老年人群公共健康和生活质量的重大医学疾病。骨关节炎的早期病变主要表现在细胞外基质成分含量的变化,患者自身很难发现,现有的临床方法和实验方法也不能较准确地识别骨关节炎的早期病变。近年来,傅里叶变换近红外(FTNIR)光谱技术因为其分析速度快、成本低、易于穿透组织获得样本的光谱信息等特点已被用于手术导航、无损检测和疾病诊断等各个领域。基于以上优势,采用FTNIR技术对不同深度分区(表层区、过渡区、深层区)的健康和骨关节炎的关节软骨进行NIR光谱采集和预处理,结合主成分分析(PCA)和Fisher判别(FDA)分别研究不同的预处理方式对判别结果的影响、不同深度下基质成分含量的变化以及骨关节炎分期识别。比较其他2种(基线校正、二阶导数3次多项式25点Savitzky-Golay平滑)预处理方式,同分区中一阶导数2次多项式21点Savitzky-Golay平滑预处理的判别结果最优,其中表层区的识别率高达95%(初始案例)和90%(交互验证案例);表层区的判别结果优于过渡区,更优于深层区,恰可证明骨关节炎的早期病变主要发生在表层区。在骨关节炎分期识别中,经数据优化后模型的初始案例识别率为100%,交互验证识别率为93.3%,预测集的识别率为87.5%。结果表明:NIR光谱的一阶导数预处理结合PCA-FDA方法能有效地鉴别关节软骨病变与否并进行骨关节炎的分期诊断,对骨关节炎监测和早期诊断研究具有重要意义,并有潜力应用于骨关节炎的原位分期和早期临床诊断。

Osteoarthritis(OA)is a major medical disease that threatens the middle-aged and aged people’s public health and quality of life.The early lesions of OA are mainly shown in the content changes of extracellular matrix components,which are difficult to detect by patients themselves,even the existing clinical and experimental methods.In recent years,Fourier transform near-infrared(FTNIR)spectroscopy has been used for the field of surgical navigation,non-destructive testing and various disease diagnoses due to its fast speed,low cost,ease of penetrating tissue to obtain spectral information,etc.Based on the above advantages,FTNIR technology was used to collect and analyze NIR spectra of healthy and multi-period OA articular cartilage at different depth zones(superficial zone,transitional zone and deep zone)in this paper.Then,principal component analysis(PCA)and Fisher discrimination algorithm(FDA)were combined for studying the influence of different preprocessing methods on discrimination results,the changes of matrix composition at different zones,and the diagnosis of multi-period OA.Compared to other two preprocessing methods(baseline correction,second-derivative cubic polynomials 25-point Savitzky-Golay smoothing)at same zone,the preprocessing of first-derivative quadratic polynomials 21-point Savitzky-Golay smoothing shows the best discrimination results,and the recognition rates at the superficial zone are as high as 95%and 90%for initial case and cross validation case,respectively.The discrimination results at the superficial zone are better than those at the transitional zone,and far better than those at the deep zone,which proves that the early lesions in OA mainly occurs at superficial zone.In the multi-period OA recognition,the recognition rates of initial case,cross-validation and prediction sets through optimized data by the FDA model are 100%,93.3%,and 87.5%,respectively.The results indicate that the first derivative pretreatment of the NIR spectra combined with the PCA-FDA method can effectively identify whether the articular cartilage is diseased and which period it is,which is of great significance in study of monitoring OA and early diagnosis.NIR technology with the appropriate spectral analysis method can be applied to the in situ staging and early clinical diagnosis of OA.