Hyperspectral Imaging Pushes PAT into a New Era

A New Vision for PAT

In their recent publication, “Hyperspectral image and chemometrics. A step beyond classical spectroscopic PAT tools” (1), authors Anna de Juan and Rodrigo Rocha de Oliveira from the Chemometrics Group, Department of Chemical Engineering and Analytical Chemistry, Universitat de Barcelona, argue that hyperspectral imaging (HSI) represents a leap forward in industrial process monitoring. Their article, published in Analytical and Bioanalytical Chemistry, emphasizes that HSI merges the strengths of spectroscopy and vision systems—delivering spatially resolved chemical information unattainable with traditional PAT tools (1).

As the authors note in the original text, the dual nature of HIS, “rich spatial and chemical information,” positions it as a transformative technology for industrial quality control, automation, and mechanistic understanding (1). PAT has been applied using spectroscopic optical technologies across biopharmaceutical and other applications for many years (1–3).

From Single-Point Sensors to Imaging Spectroscopy

Classical spectroscopic PAT relies on sensors that gather information from limited sampling spots. While powerful, these point-based measurements “lack applicability” without chemometrics to interpret them, as the authors explain in their introduction. HSI solves this limitation by associating a full spectrum with every pixel across an image. The result is a data-rich “cube” containing two spatial dimensions and one spectral dimension (1).

This cube can be interpreted through a Beer–Lambert-like bilinear model, in which spectra derive from additive contributions of pure components. According to de Juan and de Oliveira, this inherent linearity explains why established chemometric methods, such as principal component analysis (PCA) and partial least squares (PLS), translate naturally into the HSI domain (1).

Furthermore, the authors highlight that pushbroom (line-scanning) systems now acquire hyperspectral data at speeds compatible with real-time PAT monitoring.

Quantitative Mapping: Beyond Bulk Composition

One of the advancements emphasized in the original work is pixel-level quantitative prediction. Instead of using a single averaged spectrum for bulk composition, HSI enables spatial concentration maps via multivariate calibration.

This capability has proven critical in industries where distribution matters as much as quantity—especially food. The authors cite moisture-mapping in meat and salmon fillets, as well as spatial monitoring of cheese ripening. Such examples illustrate why the “distribution of constituents” is now considered an essential factor in modern quality metrics (1).

Qualitative Decisions through Imaging Classification

The article also explores qualitative PAT applications—such as sorting, defect detection, and foreign object identification—where class membership must be assigned to specific locations. De Juan and de Oliveira describe both clustering and supervised classification strategies, noting that hyperspectral information becomes indispensable when chemical differences between classes are subtle (1).

They reference a compelling example from Amigo and colleagues involving microplastic sorting. Pixel-based classification sufficed for broad categories, but object-level averaging achieved superior accuracy in distinguishing chemically similar subclasses—demonstrating the importance of multiscale approaches.

Unmixing for Process Understanding

Among the most powerful aspects of HSI described in the article is “image unmixing,” which identifies pure spectral signatures and their spatial distributions directly from raw data. The authors detail how this approach has enabled real-time monitoring of thermal-induced polymorphic transformations in pharmaceuticals. By unmixing image sequences across temperatures, researchers could visualize polymorph evolution spatially and track concentration changes quantitatively (1).

Such examples highlight HSI’s role in mechanistic process understanding—providing insights far beyond what classical sensors can deliver.

Future Directions: Compression and Multimodal Fusion

Despite rapid progress, the authors note that challenges remain. Image data are large, and industrial adoption will require efficient on-the-fly compression and smart data selection. They also predict major potential in multimodal imaging—combining, for example, NIR, Raman, and fluorescence—but acknowledge that current multimodal systems remain costly and computationally complex.

References

(1) de Juan, A.; de Oliveira, R. R. Hyperspectral Image and Chemometrics. A Step Beyond Classical Spectroscopic PAT Tools. Anal. Bioanal. Chem. 2025. DOI: 10.1007/s00216-025-06154-x.

(2) Claßen, J.; Aupert, F.; Reardon, K. F.; Solle, D.; Scheper, T. Spectroscopic Sensors for In-Line Bioprocess Monitoring in Research and Pharmaceutical Industrial Application. Anal. Bioanal. Chem. 2017, 409, 651–666. DOI: 10.1007/s00216-016-0068-x.

(3) Rolinger, L.; Rüdt, M.; Hubbuch, J. A Critical Review of Recent Trends, and a Future Perspective of Optical Spectroscopy as PAT in Biopharmaceutical Downstream Processing. Anal. Bioanal. Chem. 2020, 412, 2047–2064. DOI: 10.1007/s00216-020-02407-z.