Elettra-Sincrotrone Trieste, Strada Statale 14 - km 163.5, 34149, Basovizza, Trieste, Italy; AGH University of Science and Technology, Faculty of Physics and Applied Computer Science, al. Mickiewicza 30, 30-059, Kraków, Poland.
Elettra-Sincrotrone Trieste, Strada Statale 14 - km 163.5, 34149, Basovizza, Trieste, Italy.
Anal Chim Acta. 2020 Mar 22;1103:143-155. doi: 10.1016/j.aca.2019.12.070. Epub 2020 Jan 3.
Model-based algorithms have recently attracted much attention for data pre-processing in tissue mapping and imaging by Fourier transform infrared micro-spectroscopy (FTIR). Their versatility, robustness and computational performance enabled the improvement of spectral quality by mitigating the impact of scattering and fringing in FTIR spectra of chemically homogeneous biological systems. However, to date, no comprehensive algorithm has been optimized and automated for large-area FTIR imaging of histologically complex tissue samples. Herein, for the first time, we propose a unique, integrated and fully-automated Multiple Linear Regression Multi-Reference (MLR-MR) method for correcting linear baseline effects due to diffuse scattering, for compensating substrate thickness inhomogeneity and accounting for sample chemical heterogeneity in FTIR images. In particular, the algorithm uses multiple-reference spectra for histologically heterogeneous biological samples. The performance of the procedure was demonstrated for FTIR imaging of chemically complex rat brain frontal cortex tissue samples, mounted onto Ultralene® films. The proposed MLR-MR correction algorithm allows the efficient retrieval of "pure" absorbance spectra and greatly improves the histological fidelity of FTIR imaging data, as compared with the one-reference approach. In addition, the MLR-MR algorithm here presented opens up the possibility for extracting information on substrate thickness variability, thus enabling the indirect evaluation of its topography. As a whole, the MLR-MR procedure can be easily extended to more complex systems for which Mie scattering effects must also be eliminated.
基于模型的算法最近在傅里叶变换红外微光谱（FTIR）的组织映射和成像的数据预处理中引起了广泛关注。它们的多功能性、鲁棒性和计算性能通过减轻化学均匀生物系统的 FTIR 光谱中散射和边缘的影响，提高了光谱质量。然而，迄今为止，还没有针对组织学复杂的组织样本的大面积 FTIR 成像进行优化和自动化的综合算法。在此，我们首次提出了一种独特的、集成的、全自动的多元线性回归多参考（MLR-MR）方法，用于校正由于漫散射引起的线性基线效应，补偿基底厚度不均匀性，并解释 FTIR 图像中的样品化学异质性。特别是，该算法使用多个参考光谱对组织学不均匀的生物样品进行分析。该程序的性能已通过对化学复杂的大鼠大脑额皮质组织样本进行 FTIR 成像来验证，这些样本安装在 Ultralene®薄膜上。与单参考方法相比，所提出的 MLR-MR 校正算法允许有效地提取“纯”吸收光谱，并大大提高 FTIR 成像数据的组织学保真度。此外，这里提出的 MLR-MR 算法为提取基底厚度变化信息提供了可能性，从而能够间接评估其形貌。总的来说，MLR-MR 程序可以很容易地扩展到需要消除米氏散射效应的更复杂的系统。
