Spectral unmixing of Raman microscopic images of single human cells using Independent Component Analysis

TL;DR

This paper demonstrates that Independent Component Analysis (ICA) effectively unmixes Raman hyperspectral images of human cells, enabling detailed, label-free visualization of cellular components with minimal preprocessing.

Contribution

The study introduces ICA as a novel, effective, and minimally preprocessed method for spectral unmixing and image clustering in Raman microscopy of single human cells.

Findings

ICA reconstructs false color maps highlighting cellular structures

ICA requires minimal preprocessing and is label-free

ICA outperforms PCA in unmixing Raman spectral data

Abstract

Application of independent component analysis (ICA) as an unmixing and image clustering technique for high spatial resolution Raman maps is reported. A hyperspectral map of a fixed human cell was collected by a Raman micro spectrometer in a raster pattern on a 0.5um grid. Unlike previously used unsupervised machine learning techniques such as principal component analysis, ICA is based on non-Gaussianity and statistical independence of data which is the case for mixture Raman spectra. Hence, ICA is a great candidate for assembling pseudo-colour maps from the spectral hypercube of Raman spectra. Our experimental results revealed that ICA is capable of reconstructing false colour maps of Raman hyperspectral data of human cells, showing the nuclear region constituents as well as subcellular organelle in the cytoplasm and distribution of mitochondria in the perinuclear region. Minimum preprocessing requirements and label-free nature of the ICA method make it a great unmixed method for extraction of endmembers in Raman hyperspectral maps of living cells.