中药指纹图谱分析及其FPGA实现  

FPGA based Fingerprint Analysis of Traditional Chinese Medicines

作　　者：沈晋慧[1] 曹辉[1] 马栋萍[1] 

机构地区：[1]北京联合大学生物化学工程学院,北京100023

出　　处：《中国医学物理学杂志》2009年第4期1327-1330,共4页Chinese Journal of Medical Physics

基　　金：北京市教委科研基金资助项目(No.KM200711417014)

摘　　要：目的:研究用于中药色谱指纹图谱相似性分析的算法及其在FPGA上实现硬化。方法:液相色谱指纹图谱的原始数据存在很多种漂移和数据干扰,在进行相似度分析之前必须去除这些干扰因素,我们使用带阻滤波消除原始信号中的噪声,使用遗传算法实现图谱间指纹峰的匹配,并采用相关系、夹角余玄和相关系数作为相似度判别的标准。上述算法均可移植到FPGA上,借助FPGA内部计算与操作全过程可完全并行化的能力,实现该算法的高速运算操作。结果:经相关领域专家检验,认为夹角余弦与相关系数的评价结果比较接近,欧氏距离测度则具有较好的综合评价能力。结论:用3种相似性测度计算不同批次丹参粉针注射液指纹图谱相似性的结果表明,夹角余弦测度更适于分析评价不同批次间药品质量的稳定性。整套算法能够无缝移植到FPGA硬件架构上,实现中药指纹图谱的现场分析,在中药炮制过程中实时分析炮制情况,指导中药炮制达到最优控制。Objective： An analysis algorithm of traditional Chinese medicines fingerprint similarity degree and it＇s FPGA circuit method was designed in the article. Methods： The raw data of high performance liquid chromatography fingerprints consists retention time drift, vertical drift and noise which will bring disturbance to similarity degree analysis, so we need to remove them. Using band-hinder filter to remove the noise and using genetic algorithm to correcting retention time drift and vertical drift so the fingers on two fingerprints are matching. The parameters we use are correlation coefficient, angle cosine and euclidean distance. All algorithms we discussed above can implemented successfully on FPGA which can use in field control with it＇s highly parallel computing technique to realize fast calculation. Results： The experts of related fields considered that the results of correlation coefficient and angle cosine are very similar, and euclidean distance has better comprehensive ability. Conclusions： Using three parameters to calculate the similarity of two salvia powder injection fingerprint The result shows that the angle cosine is more suitable for measuring the similarity of fingerprimts. All algorithms can put on the FPGA circuit for field analysis and real-time control of chinese herbal medicine processing.

关 键 词：指纹图谱分析 FPGA 遗传算法 

分 类 号：R318.6[医药卫生—生物医学工程]