Investigating the Adoption of Raman Spectroscopy in Routine Oncology Workflows: A SciX Interview with Nick Stone, Part 2

Transferability between spectrometers is a critical challenge for the implementation of advanced diagnostic technologies like Raman spectroscopy in clinical analysis. There have been several studies recently that have explored this topic at length. In one study, a research team from Norway studied a method using partial least squares (PLS) models to assist in transferring calibrations between different Raman and near-infrared (NIR) spectrometers (1).

In clinical analysis, vibrational spectroscopy for analytical measurements has become more popular. The reason for this trend is that spectroscopy not only allows for rapid analysis, but it is also inexpensive (1). These two reasons benefit the patient and the physician, because it helps expedite patient treatment and therefore improve patient outcomes.

Spectrometers, even those of the same make and model, can produce variations in spectral data because of differences in calibration, environmental conditions, and operator expertise (1). Calibration techniques attempt to solve this issue between instruments when data transferability is required. Historically, transferability has involved extensive computational corrections and data pre-processing to align spectra from different instruments, adding layers of complexity and creating barriers to widespread clinical adoption (1,2).

Efforts to improve transferability have focused on developing standardized protocols and calibration techniques to reduce variability between spectrometers (2). Studies like the one led by Dr. Nick Stone have demonstrated that, by adhering to common data acquisition protocols, it is possible to maintain high diagnostic accuracy when transferring Raman spectra across multiple clinical centers (3,4). For instance, Stone’s work in esophageal cancer detection showed no significant drop in model performance or log-loss when applying data from one center to another, despite using spectrometers at different locations (4). Notably, this was achieved without the need for complex computational corrections.

In part two of our conversation with Stone, he discusses these issues and answers the following questions:

• How significant is the issue of transferability between spectrometers in clinical settings, and what are the potential implications for patient outcomes if this issue is not resolved?
• How does machine learning play a role helping to support multi-center clinical diagnostics? What are some potential benefits and drawbacks of using ML in this case?

To view the rest of our coverage of the upcoming SciX 2024 Conference, click here: https://www.spectroscopyonline.com/conferences/scix

References

1. Tengstrand, E.; Solberg, L. E.; Dankel, K.; et al. Calibration Transfer Between Different Spectrometers by Wavelength Correspondence. Vib. Spectrosc. 2024, 132, 103667. DOI: 10.1016/j.vibspec.2024.103667
2. Workman, Jr., J. Calibration Transfer, Part II: The Instrumentation Aspects. Spectroscopy 2013, 28 (5). https://www.spectroscopyonline.com/view/calibration-transfer-part-ii-instrumentation-aspects (accessed 2024-09-30).
3. Wetzel, W. Investigating the Adoption of Raman Spectroscopy in Routine Oncology Workflows: A SciX Interview with Nick Stone, Part 1. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/investigating-the-adoption-of-raman-spectroscopy-in-routine-oncology-workflows-a-scix-interview-with-nick-stone-part-1 (accessed 2024-09-30).
4. Chasse, J. Raman-Based Oesophageal Tissue Classification and Modern Machine Learning in the Support of Multi-Center Clinical Diagnostics: An Interview with 2024 Charles Mann Award Winner Nick Stone. Spectroscopy. Available at: https://www.spectroscopyonline.com/view/raman-based-oesophageal-tissue-classification-and-modern-machine-learning-in-the-support-of-multi-center-clinical-diagnostics-an-interview-with-2024-charles-mann-award-winner-nick-stone (accessed 2024-09-30).