Raman spectroscopy and non-linear Raman spectroscopic techniques are increasingly employed across various disciplines, including chemical analytics, life sciences, and medicine. The applications in these fields rely on artificial intelligence (AI)-based methods to translate measured data into high-level information and knowledge within the application domain. The high-level information depends on the specific task and sample characteristics, such as disease types, tissue types, and other properties like constituent concentrations.

To achieve this translation, specialized data pipelines must be constructed for each measurement modality, comprising experimental design, sample size planning, data pre-treatment, data pre-processing, chemometric and machine learning-based data modeling, model transfer methods, and transfer learning. Almost every step in the data pipeline can be optimized using AI-based methods, including machine learning and deep learning.

This talk will highlight common pitfalls encountered when generating data pipelines for linear and non-linear Raman spectroscopic measurement techniques and discuss strategies to avoid them.