Department of Inorganic and Analytical Chemistry, University of Szeged, Dóm square 7, Szeged, 6720, Hungary.
Institute of Chemical Technologies and Analytics, TU Wien, Getreidemarkt 9/164, 1060, Vienna, Austria.
Sci Rep. 2023 Jun 21;13(1):10089. doi: 10.1038/s41598-023-37258-y.
The present study demonstrates the importance of converting signal intensity maps of organic tissues collected by laser-induced breakdown spectroscopy (LIBS) to elemental concentration maps and also proposes a methodology based on machine learning for its execution. The proposed methodology employs matrix-matched external calibration supported by a pixel-by-pixel automatic matrix (tissue type) recognition performed by linear discriminant analysis of the spatially resolved LIBS hyperspectral data set. On a swine (porcine) brain sample, we successfully performed this matrix recognition with an accuracy of 98% for the grey and white matter and we converted a LIBS intensity map of a tissue sample to a correct concentration map for the elements Na, K and Mg. Found concentrations in the grey and white matter agreed the element concentrations published in the literature and our reference measurements. Our results revealed that the actual concentration distribution in tissues can be quite different from what is suggested by the LIBS signal intensity map, therefore this conversion is always suggested to be performed if an accurate concentration distribution is to be assessed.
本研究证明了将激光诱导击穿光谱（LIBS）采集的有机组织的信号强度图转换为元素浓度图的重要性，并提出了一种基于机器学习的执行方法。所提出的方法采用基于外部校准的矩阵匹配，由通过对空间分辨 LIBS 高光谱数据集的线性判别分析进行的逐像素自动矩阵（组织类型）识别来支持。在猪（猪）脑样本上，我们成功地以 98%的准确度对灰质和白质执行了这种矩阵识别，并且我们将组织样本的 LIBS 强度图转换为 Na、K 和 Mg 元素的正确浓度图。在灰质和白质中发现的浓度与文献中公布的元素浓度和我们的参考测量值一致。我们的结果表明，组织中的实际浓度分布可能与 LIBS 信号强度图所暗示的浓度分布有很大不同，因此如果要评估准确的浓度分布，总是建议进行这种转换。
