Broadband coherent anti-Stokes Raman scattering (BCARS) micro-/spectroscopy is a powerful method of chemical imaging that is self-referenced to a co-generated background signal; thus, enabling direct, unnormalized comparison between microscopy platforms and samples. The workflow required to extract the self-referenced Raman spectra, however, has typically required milliseconds per spectrum; thus, hindering real-time analysis, visualization, and user interactivity. In this work, we demonstrate a new workflow using linear machine learning methods that is intrinsically interpretable, informed by the physics of the problem, and yet enables real-time processing with improved quantitation. Additionally, we will demonstrate how this can translate into automated region-of-interest selection.
|