Automatic Detection of Subcellular-Scale Metabolic Dynamics via Spontaneous-Stimulated Raman Spatial Colocalization.

Cellular structural heterogeneity and low intrinsic contrast in label-free bright-field imaging hinder accurate localization of subcellular structures in nearly transparent specimens, thereby compromising both the precision and reproducibility of spontaneous Raman microscopy. Herein, we present a novel dual-modality, label-free three-dimensional (3D) automatic cell analysis platform that integrates spontaneous and stimulated Raman scattering (SRS) modes for morphological characterization and chemical mapping. The spatial distribution of specific molecules within cells was acquired via SRS 3D rapid imaging within 1 min. Subsequently, utilizing submicron positioning, spontaneous Raman spectroscopy was guided to automatically collect broadband, high spectral resolution Raman spectra spanning the fingerprint to high wavenumber regions, thereby obtaining the molecular component information on the specified intracellular organelles. We propose a novel method to measure the axial positioning consistency of spontaneous Raman detection. For the cell identification approach integrating Raman spectroscopy and machine learning, we demonstrate that enhanced axial positioning accuracy and sampling consistency of the system significantly improve the performance of the trained cell discrimination model relative to conventional Raman microscopy, achieving a cell identification accuracy over 98%. Using this platform, Raman spectroscopy based metabolic dynamics analysis of lipid droplet core components revealed that, under optimal culture conditions, lipid droplet cores in cell line samples were predominantly composed of the unsaturated oleate. In contrast, lipid droplet cores in clinical sample cells contained not only oleate but also a significant amount of the palmitate. Furthermore, the cholesterol ester content in lipid droplets within white blood cells of leukemia patients is significantly higher compared to that in normal samples. Deuterium labeling validated the de novo synthesis of oleate in lipid droplets of cell lines.