论文部分内容阅读
多光谱活体荧光成像技术正逐渐成为生物医学研究的关键技术,但是荧光物质光谱之间的串扰和自发荧光现象严重影响了荧光影像的解译。混合光谱分解对于去除活体多光谱荧光影像自发荧光效应和进行多种荧光信号分离是非常有效的技术,但是光谱分解的前提是获得了各种荧光物质的光谱。基于多元曲线解析交替最小二乘法(MCR-ALS)计算框架,提出包括非负、等式、闭合性、单峰、波段范围及归一化的多约束条件的荧光纯光谱估计方法,利用估计的纯光谱和线性混合光谱模型得到不同荧光信号的分离,去除自发荧光背景对起标记目的的荧光物质信号的干扰。Dirichle分布随机混合构造的不同信噪比和纯净水平的荧光蛋白混合光谱数据分析结果反映出在混合问题严重、有噪声影响的情况下,该算法要比传统端元光谱分析方法的精度高10倍以上。活体鼠多光谱量子荧光影像的实验也证明了该算法在信号分离上的有效性。
Multispectral living-body fluorescence imaging technology is becoming the key technology in biomedical research. However, the crosstalk and autofluorescence between the fluorescence spectra seriously affect the interpretation of fluorescence images. Mixed spectral decomposition is a very effective technique for removing autofluorescence and separating multiple fluorescent signals in living multispectral fluorescence images. However, the premise of spectral decomposition is to obtain the spectra of various fluorescent substances. Based on the framework of Multivariate Curve Analysis and Alternating Least Squares (MCR-ALS) calculation, a new method of estimating fluorescence pure spectrum based on the multi-constraint conditions of nonnegative, equality, closedness, unimodal, band range and normalization is proposed. The separation of different fluorescence signals is achieved by the pure spectral and linear mixed spectral models, removing the interference of the autofluorescent background to the fluorescent substance signal for the purpose of labeling. Dirichle distribution Randomly mixed structures of different signal-to-noise ratios and pure levels of fluorescence protein mixed spectral data analysis results show that the algorithm is more than 10 times more accurate than traditional end-point spectral analysis methods under the conditions of severe mixing and noise effects the above. The experiment of multi-spectral quantum fluorescence images of living mice also proved the effectiveness of the algorithm in signal separation.