Best of the Week: Interview with Juergen Popp, Microplastic Detection, Machine Learning Models

This week, Spectroscopy published various articles touch upon several important application areas such as environmental analysis and biophotonics. These pieces also feature discussions of integrating artificial intelligence (AI) into the analysis. Several key techniques are highlighted, including near-infrared spectroscopy (NIR), Raman spectroscopy, and Fourier transform infrared (FT-IR) spectroscopy. Happy reading!

Recapping Photonics West: Integrating AI and Raman to Improve Tumor Classification

At Photonics West 2024, Juergen Popp, who is the Scientific Director at the Leibniz Institute for Photonics Technology, discussed how AI-enhanced biophotonics is transforming cancer and infection diagnostics. His talk covered label-free spectroscopic techniques for tumor margin control, tumor typing, and personalized treatment planning (1). In this interview, Popp discusses the role of Raman spectroscopy in infection treatment, including rapid pathogen identification, resistance profiling, immune response analysis, and treatment evaluation, highlighting its potential for improving precision medicine and patient outcomes (1).

Detection of Colorless Microplastics in the Environment Using NIR Spectroscopy and Machine Learning

Researchers from Tongji University and the Shanghai Institute of Pollution Control developed a novel method using NIR hyperspectral imaging (NIR-HSI) and machine learning (ML) to detect colorless microplastics, which are often overlooked in environmental surveys. The researchers tested four ML models, finding that a two-stage classification approach improved accuracy to 99% (2). The method, which requires no labor-intensive sample preparation, enables large-scale plastic pollution monitoring and industrial applications like waste sorting (2). This breakthrough enhances environmental assessments, ensuring colorless microplastics are accurately accounted for, and supports more effective plastic waste management and sustainability initiatives.

Spectroscopy and GPC to Evaluate Dissolved Organic Matter

Researchers from Beijing University of Civil Engineering and Architecture and China Construction Fifth Engineering Division evaluated the effectiveness of sludge-filled filters in removing dissolved organic matter (DOM) from road runoff using gel permeation chromatography (GPC), UV–vis spectroscopy, and excitation-emission matrix (EEM) fluorescence spectroscopy (3). The filters achieved a 70–80% DOM removal rate, efficiently targeting macromolecular and hydrophobic compounds. The sludge promoted microbial activity, enhancing degradation (3). Although promising, further research is needed to address potential clogging and long-term performance. Integrating sludge-based filters with existing urban water management systems could improve runoff treatment and protect aquatic environments (3).

Distinguishing Horsetails Using NIR and Predictive Modeling

Horsetails (genus Equisetum) are ancient plants dating back 325 million years to the Carboniferous period, with 15 species found worldwide. They are linked to coal formation, but species identification is challenging due to morphological similarities. Knut Baumann of the University of Technology Braunschweig used NIR spectroscopy to differentiate species. In this interview, Baumann talks about his research and its implications for species classification and medicinal applications (4).

Blood-Glucose Testing: AI and FT-IR Claim Improved Accuracy to 98.8%

A recent study showed how FT-IR spectroscopy can significantly improve non-invasive blood-glucose testing. In the study, the team replaced traditional single-pass attenuated total reflection (ATR) with a multiple-reflection ATR (MATR) setup, increasing sensitivity (5). They also integrated a quantum cascade laser (QCL) for precise glucose detection and applied two-dimensional correlation spectroscopy (2D-COS) to minimize interference (5). Machine learning (ML) algorithms further improved classification accuracy to 98.8% (5). Validated across 7,200 test spectra, this approach offers a promising, highly accurate alternative to invasive glucose monitoring, advancing diabetes management and early detection with non-invasive technology.

References

1. Wetzel, W. Recapping Photonics West: Integrating AI and Raman to Improve Tumor Classification. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/enhancing-tumor-classification-with-ai-and-raman-a-conversation-with-juergen-popp (accessed 2025-02-05).
2. Wetzel, W. Detection of Colorless Microplastics in the Environment Using NIR Spectroscopy and Machine Learning. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/detection-of-colorless-microplastics-in-the-environment-using-nir-spectroscopy-and-machine-learning (accessed 2025-02-05).
3. Chasse, J. Spectroscopy and GPC to Evaluate Dissolved Organic Matter. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/spectroscopy-and-gpc-to-evaluate-dissolved-organic-matter (accessed 2025-02-05).
4. Wetzel, W. Distinguishing Horsetails Using NIR and Predictive Modeling. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/distinguishing-horsetails-using-nir-and-predictive-modeling (accessed 2025-02-05).
5. Workman, Jr., J. Blood-Glucose Testing: AI and FT-IR Claim Improved Accuracy to 98.8%. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/blood-glucose-testing-ai-and-ft-ir-claim-improved-accuracy-to-98-8- (accessed 2025-02-05).