Evaluating Normalization Methods for Robust Spectral Performance Assessments of Hyperspectral Imaging Cameras
Biosensors · 04.01.2025 · 15 (1) · 20
DOI: 10.3390/bios15010020 ↗Kurz gefasst
Fachartikel in der Zeitschrift Biosensors, online erschienen am 04.01.2025. Verfasst von Siavash Mazdeyasna und 5 weiteren Autor:innen. Beteiligte Einrichtung: Center for Devices and Radiological Health, U.S. Food and Drug Administration, Silver Spring, MD 20993, USA. Hauptthemen laut MEDLINE: Hyperspectral Imaging. Die Arbeit wurde bisher 28-mal zitiert (Crossref) und nennt 37 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: U.S. Food and Drug Administration, U.S. Food and Drug Administration’s Critical Path funding.
Abstract (Originalsprache)
Hyperspectral imaging (HSI) technology, which offers both spatial and spectral information, holds significant potential for enhancing diagnostic performance during endoscopy and other medical procedures. However, quantitative evaluation of HSI cameras is challenging due to various influencing factors (e.g., light sources, working distance, and illumination angle) that can alter the reflectance spectra of the same target as these factors vary. Towards robust, universal test methods, we evaluated several data normalization methods aimed at minimizing the impact of these factors. Using a high-resolution HSI camera, we measured the reflectance spectra of diffuse reflectance targets illuminated by two different light sources. These spectra, along with the reference spectra from the target manufacturer, were normalized with nine different methods (e.g., area under the curve, standard normal variate, and centering power methods), followed by a uniform scaling step. We then compared the measured spectra to the reference to evaluate the capability of each normalization method in ensuring a consistent, standardized performance evaluation. Our results demonstrate that normalization can mitigate the impact of some factors during HSI camera evaluation, with performance varying across methods. Generally, noisy spectra pose challenges for normalization methods that rely on limited reflectance values, while methods based on reflectance values across the entire spectrum (such as standard normal variate) perform better. The findings also suggest that absolute reflectance spectral measurements may be less effective for clinical diagnostics, whereas normalized spectral measurements are likely more appropriate. These findings provide a foundation for standardized performance testing of HSI-based medical devices, promoting the adoption of high-quality HSI technology for critical applications such as early cancer detection.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Center for Devices and Radiological Health, U.S. Food and Drug Administration, Silver Spring, MD 20993, USA.
- Chemical and Biomolecular Engineering, University of South Alabama, Mobile, AL 36688, USA.
ORCID-Profile: Siavash Mazdeyasna, Mohammed Shahriar Arefin, Andrew Fales, Silas J Leavesley, Quanzeng Wang
Förderung
- U.S. Food and Drug Administration
- U.S. Food and Drug Administration’s Critical Path funding
Zitieren
APA
Mazdeyasna, S., Arefin, M. S., Fales, A., Leavesley, S. J., Pfefer, T. J., & Wang, Q. (2025). Evaluating Normalization Methods for Robust Spectral Performance Assessments of Hyperspectral Imaging Cameras. Biosensors, 15(1), 20. https://doi.org/10.3390/bios15010020
Vancouver
Mazdeyasna S, Arefin MS, Fales A, Leavesley SJ, Pfefer TJ, Wang Q. Evaluating Normalization Methods for Robust Spectral Performance Assessments of Hyperspectral Imaging Cameras. Biosensors. 2025;15(1):20. doi:10.3390/bios15010020
BibTeX
@article{Mazdeyasna2025,
author = {Mazdeyasna, Siavash and Arefin, Mohammed Shahriar and Fales, Andrew and Leavesley, Silas J and Pfefer, T Joshua and Wang, Quanzeng},
title = {Evaluating Normalization Methods for Robust Spectral Performance Assessments of Hyperspectral Imaging Cameras},
journal = {Biosensors},
year = {2025},
volume = {15},
number = {1},
pages = {20},
doi = {10.3390/bios15010020}
}
EUVIMED