数字图像比色法对胆红素检测的多种模型研究  

Analysis on multiple models for bilirubin detection by digital image colorimetry

作　　者：李平 吕圆 杨赛男[1] 罗鑫 王诗琪 戴斌[1] LI Ping;LV Yuan;YANG Sai-nan;LUO Xin;WANG Shi-qi;DAI Bin(College of Medicine&Technology,Hunan Polytechnic of Environment and Biology,Hengyang 421005,China)

机构地区：[1]湖南环境生物职业技术学院医药技术学院,湖南衡阳421005

出　　处：《化学研究与应用》2024年第10期23-32,共10页Chemical Research and Application

基　　金：湖南省自然科学基金项目(2023JJ60207)资助;湖南省教育厅科学研究项目(21B0874)资助;衡阳市科技创新计划项目(202121034379)资助;衡阳市科技计划项目(202010031493)资助;衡阳市科技指导性项目(S2018F9031018292)资助;衡阳市指导性计划项目(衡科发[2021]51号-37资助;湖南环境生物职业技术学院培优计划科研专项项目(PY2021-09)资助;湖南环境生物职业技术学院青年基金项目(ZK2021-03)资助。

摘　　要：数字图像比色法的关键在于建立适合的定量参数。已有多种颜色空间和数学模型用于定量参数的建立。本文以胆红素检测为例,利用智能手机采集图像,对HSV、L^(*)a^(*)b^(*)、CMYK以及RGB颜色空间的28种数学模型进行比较和分析,并考察了各模型在灵敏度、线性相关系数、抗干扰能力、重复性方面的表现。除IB外,其他27种数学模型的决定系数R^(2)均高于0.9。不同数学模型在灵敏度方面存在差异。RGB颜色空间中,选择与反应体系颜色呈互补色的颜色通道以及该通道衍生的数学模型定量,可以获得较高的灵敏度。L^(*)a^(*)b^(*)和CMYK颜色空间中,选择与反应体系颜色接近的颜色通道定量,能获得较高的灵敏度。抗干扰研究中,引入空白样本的数学模型能够实现更好的抗光强干扰效果。其中I_(R0)-I_(R)、I_(G0)-I_(G)、(I_(0)-I)_(total)、[(I-I_(blank))_(total)^(2)]^(0.5)、-lg(Gray/Gray_(up))模型的测试结果在不同光强下的绝对偏差小于4.59%。IG及其衍生数学模型的抗色温能力较强,其中I_(G)、255-I_(G)、I_(G)/(I_(R)+I_(G)+I_(B))模型在不同色温下的测试结果的绝对偏差小于5.84%。欧氏距离模型(I_(R)^(2)+I_(G)^(2)+I_(B)^(2))0.5对光强和色温均表现出较强的抗干扰能力。对28种模型进行检出限测试,吸光度模型、L^(*)、a^(*)模型的检出限均小于1.5μM,I_(G)及其色彩通道建立的数学模型所得检出限相对其他通道低。研究结果表明,图像比色法数学模型的选择,应根据定义颜色空间的特点选择最佳的颜色通道,并优化数学模型,可以提高检测性能。为提高模型的抗干扰能力,可采用欧式距离模型,或引入空白样本信号值。The key to digital image colorimetry by smartphone is to establish appropriate quantitative parameters.Multiple color space and mathematical models have been employed for this purpose.In this study,bilirubin detection was employed as an example,and images were captured using a smartphone.A comparison and comprehensive analysis of 28 mathematical models based on the HSV,L^(*)a^(*)b^(*),CMYK,and RGB color spaces were conducted,along with an examination of the performance of these models in terms of sensitivity,linearity,interference resistance,and repeatability.Except for I_(B),the coefficient of determination(R^(2))of the oth-er 27 mathematical models were all above 0.9.Sensitivity varied among different mathematical models.In the RGB color space,se-lecting a color channel that is complementary to the reaction system color and quantifying it using its derived models can yield high-er sensitivity.In the L^(*)a^(*)b^(*) and CMYK color space,selecting a color channel that closely matches the reaction system color also results in higher sensitivity.In the anti-interference study,the introduction of a data model for blank samples demonstrated superior resiliene againt ight intesity interferene ncluding I_(RO)-I_(R)、I_(GO)-I_(G)、(I_(O)-I)_(total)、[(I-I_(blank)^(2)_(total))]^(05)、-1g(Cray/Gray_(up)),modela and the absolute deviation of test results at different light intensities was less than 4.59%.Ic and its derived mathematical models exhibited robust resistance to color temperature,including I_(G),255-I_(G),I_(G)/(I_(R)+I_(G)+I_(B))models,and the absolute deviation of test results under different color temperatures was less than 5.84%.The Euclidean distance model(I_(R)^(2)+I_(G)^(2)+I_(B)^(2))^(0.5) displayed an excellent anti-interfer-ence ability against both light intensity and color temperature.The detection limits of the 28 models were conducted,and the absor-bance model,L^(*)and a^(*)models had detection limits below 1.5μM.The Ic channel and its associated mathematical models yielded lower detection lim

关 键 词：数字图像比色法 数学模型 智能手机 胆红素检测 

分 类 号：O657.3[理学—分析化学]