Normalized Raman Imaging for Studies of Tissue Physiology of the Kidney.

Histology is the cornerstone of clinical pathology and an essential tool for many areas of medicine. Nevertheless, conventional histological methods, which rely on fixation, embedding, secitoning, and staining, distort cellular architecture, extract key molecules such as lipids, and introduce variability that severely limits reproducibility. Here, we present Normalized Raman Imaging (NoRI), a form of stimulated Raman scattering applied to organ pathology. NoRI enables quantitative, label-free measurements of proteins and lipids at high spatial resolution. NoRI overcomes heterogeneous light scattering from homogeneous tissues by computationally correcting each signal by the combined Raman signals of protein, lipid, and water. This enables quantitative biomass measurements while preserving tissue architecture, thereby facilitating advanced analysis by convolutional neural networks and feature discovery. Here we apply NoRI to the mouse kidney, showing that such imaging can be used to accurately classify tubule types (median F1 [harmonic mean of precision and recall against manual annotation]=0.93), anatomical regions (F1=0.91), and biological sex (F1=0.97) from regions as small as 132.5μm