Purpose

Although the characterization of the chemical and spatial distribution of compounds within a pharmaceutical tablet is still not a routine task, applying Raman spectroscopy with data analysis methods gives the possibility to obtain in-depth information on tablet quality rapidly. However, constraints such as analysis time, laser intensity, and spot size influence the quality of acquired spectra resulting in low signal-to-noise ratio spectra. Therefore, this study proposes a method to characterize a solid heterogeneous pharmaceutical product (e.g., a tablet) based on the product’s Raman chemical map.

Methods

In this work, surface Raman data were acquired using a simple and rapid method. An algorithm based on the hierarchical application of multivariate curve resolution with log-likelihood maximization combined with principal component analysis was used to blind identify the compounds and create a chemical map.

Results

Although the direct application of multivariate curve resolution algorithms did not allow a complete tablet characterization, the hierarchical application enabled individual compounds acquired from the mixed spectra to be identified and their chemical distribution in the tablet to be mapped without the use of external references. Results were successfully benchmarked against the EDXS analysis.

Conclusions

Innovations in multivariate methods could help overcome challenges and constraints in data acquisition. This method was, for example, found to be more robust against the presence of spectral outliers. It is promising for 3D analysis of real and complex samples.

Graphical abstract

中文翻译：

---

分层多元曲线分辨率与拉曼成像相结合，用于快速表征药片

目的

尽管表征药片中化合物的化学和空间分布仍不是一项常规任务，但将拉曼光谱与数据分析方法结合使用可以快速获得有关片剂质量的深入信息。然而，分析时间、激光强度和光斑大小等限制因素会影响所获得光谱的质量，从而导致光谱信噪比低。因此，本研究提出了一种基于产品的拉曼化学图来表征固体异质药物产品（例如，片剂）的方法。

方法

在这项工作中，使用简单快速的方法获取表面拉曼数据。一种基于多变量曲线分辨率的分层应用与对数似然最大化结合主成分分析的算法用于盲识别化合物并创建化学图。

结果

尽管直接应用多变量曲线分辨率算法无法对片剂进行完整表征，但分层应用程序可以识别从混合光谱中获取的单个化合物，并在不使用外部参考的情况下绘制它们在片剂中的化学分布。结果成功地与 EDXS 分析进行了基准测试。

结论

多元方法的创新可以帮助克服数据采集中的挑战和限制。例如，发现这种方法对光谱异常值的存在更加稳健。它有望用于真实和复杂样本的 3D 分析。

图形概要