Clinical Spectroscopic Diagnostics

Hyperspectral unmixing is a well-established approach for analyzing spectroscopic images by recovering chemically interpretable component spectra and their spatial distributions. Despite its distinct analytical objective, unmixing is frequently discussed alongside or directly compared with more familiar multivariate techniques such as principal component analysis (PCA) and clustering, which address fundamentally different questions. Although these conceptual distinctions have long been recognized in chemometrics and hyperspectral imaging, they are often treated implicitly and are not always demonstrated side by side in a single, accessible framework, particularly for nonexpert users of spectroscopic imaging methods. In this work, we provide a concise qualitative illustration of these established principles using a Raman spectroscopic image of a single eukaryotic cell. Employing classical implementations of PCA, k-means and fuzzy c-means clustering, and hyperspectral unmixing based on the N-FINDR end-member extraction algorithm, we show how unmixing yields chemically interpretable pure component spectra and abundance maps, whereas PCA and clustering produce complementary but fundamentally different types of outputs. This contribution serves as a practical reference, clarifying the principles of unmixing and supporting its proper application and communication in spectroscopic studies.