Analysis of variance in spectroscopic imaging data from human tissues

Area of Science:

• Biomedical Engineering
• Spectroscopy
• Data Science

Background:

• Fourier transform infrared (FT-IR) spectroscopic imaging offers potential for analyzing cell types and diseases.
• Current limitations in understanding performance variability hinder technological advancement and clinical adoption.
• Data variance from biological diversity and measurement noise impacts FT-IR analysis reliability.

Purpose of the Study:

• To identify and quantify sources of data variation in FT-IR spectroscopic imaging.
• To establish a framework for improving FT-IR technology performance.
• To guide the development of statistically valid studies for clinical translation.

Main Methods:

• Utilized a high-throughput tissue microarray (TMA) platform for diverse data collection.
• Applied comprehensive analysis of variance (ANOVA) models to quantify variation sources.
• Estimated explained variation portions to identify key discriminating spectral metrics.

Main Results:

• Quantified primary sources contributing to data variation in FT-IR spectra.
• Identified the most effective spectral metrics for distinguishing between samples.
• Highlighted specific areas for technological improvement in FT-IR imaging.

Conclusions:

• Provides a quantitative framework for understanding and mitigating data variation in FT-IR spectroscopy.
• Offers guidelines for designing robust studies and advancing clinical translation of FT-IR imaging.
• Enhances the reliability and applicability of FT-IR spectroscopic imaging in biomedical research.