Efficient quantitative hyperspectral image unmixing method for large-scale Raman micro-spectroscopy data analysis

TL;DR

This paper introduces an efficient, multi-step computational method for large-scale hyperspectral image unmixing in Raman micro-spectroscopy, enabling rapid, accurate analysis of complex biological tissues without prior knowledge.

Contribution

The paper presents a novel, integrated unmixing approach combining noise filtering, background subtraction, and non-negative matrix factorization tailored for large-scale Raman hyperspectral data analysis.

Findings

Successfully applied to human atherosclerotic tissue images

Accurately retrieves biochemical composition without prior information

Operates with high speed suitable for large datasets

Abstract

Vibrational micro-spectroscopy is a powerful optical tool, providing a non-invasive label-free chemically specific imaging for many chemical and biomedical applications. However, hyperspectral image produced by Raman micro-spectroscopy typically consists of thousands discrete pixel points, each having individual Raman spectrum at thousand wavenumbers, and therefore requires appropriate image unmixing computational methods to retrieve non-negative spatial concentration and corresponding non-negative spectra of the image biochemical constituents. In this article, we present a new efficient Quantitative Hyperspectral Image Unmixing (Q-HIU) method for large-scale Raman micro-spectroscopy data analysis. This method enables to simultaneously analyse multi-set Raman hyperspectral images in three steps: (i) Singular Value Decomposition with innovative Automatic Divisive Correlation (SVD-ADC) which autonomously filters spatially and spectrally uncorrelated noise from data; (ii) a robust subtraction of fluorescent background from the data using a newly developed algorithm called Bottom Gaussian Fitting (BGF); (iii) an efficient Quantitative Unsupervised/Partially Supervised Non-negative Matrix Factorization method (Q-US/PS-NMF), which rigorously retrieves non-negative spatial concentration maps and spectral profiles of the samples' biochemical constituents with no a priori information and with great operation speed. Alternatively, the Q-US/PS-NMF is capable to work as a partially supervised method, when one or several samples' constituents are known, which significantly widens chemical and biomedical applications of the method. We apply the Q-HIU to the analysis of real large-scale Raman hyperspectral images of human atherosclerotic aortic tissues and our results show a proof-of-principle for the proposed method to retrieve the biochemical composition of the tissues.