Evaluating Normalization Methods for Robust Spectral Performance Assessments of Hyperspectral Imaging Cameras.

Area of Science:

• Medical Imaging
• Spectroscopy

Background:

• Hyperspectral imaging (HSI) provides spatial and spectral data, crucial for medical diagnostics.
• Quantitative evaluation of HSI cameras is difficult due to variability from light sources, distance, and angle.
• Standardized testing methods are needed for reliable HSI device performance assessment.

Purpose of the Study:

• To evaluate data normalization methods for minimizing variability in HSI camera performance testing.
• To assess the effectiveness of different normalization techniques in standardizing HSI camera evaluations.
• To provide a foundation for robust, universal test methods for medical HSI devices.

Main Methods:

• Measured reflectance spectra of diffuse targets using a high-resolution HSI camera and two light sources.
• Applied nine different data normalization methods (e.g., standard normal variate) followed by uniform scaling.
• Compared normalized spectra against manufacturer reference spectra to assess method performance.

Main Results:

• Normalization methods can mitigate the impact of certain factors on HSI camera evaluation.
• Performance varied across normalization methods; those using full spectral data (e.g., standard normal variate) performed better, especially with noisy spectra.
• Normalized spectral measurements appear more suitable for clinical diagnostics than absolute reflectance measurements.

Conclusions:

• Normalization is beneficial for standardizing HSI camera performance evaluation.
• Standard normal variate and similar methods are effective for robust HSI testing.
• Findings support the development of standardized HSI device testing for improved medical applications like early cancer detection.