Intra-class variability in diffuse reflectance spectroscopy: application to porcine adipose tissue

Area of Science:

• Biomedical Optics
• Spectroscopy
• Tissue Characterization

Background:

• Optical diffuse reflectance spectroscopy (DRS) is valuable for biological tissue study and diagnosis.
• Accurate tissue classification relies on robust training datasets for DRS applications.
• Measurement variability within and between samples is often overlooked in DRS analyses.

Purpose of the Study:

• To evaluate spectral variations within and between ex-vivo porcine adipose tissue samples from different animals.
• To assess the impact of this variability on tissue classification accuracy.
• To determine if spectral variability is random or influenced by biological factors.

Main Methods:

• Applied normalization, principal component analysis (PCA) for dimensionality reduction, and variability control algorithms.
• Analyzed spectral variations within and between samples using PCA.
• Calculated Mahalanobis distance and performed ANOVA tests to quantify variability.

Main Results:

• PCA revealed significant dataset variability, even after applying variability removal algorithms.
• Projected data clustered by animal in the principal component space.
• Variability was confirmed to be non-random, dependent on anatomical location and individual animal.

Conclusions:

• Spectral variability in DRS measurements of adipose tissue is significant and not random.
• This variability is influenced by the specific animal and anatomical origin.
• Accounting for non-random variability is crucial for improving the accuracy of DRS-based tissue classification.