New imaging method measures proteins and lipids in kidney tissue

A pipeline using normalized stimulated Raman imaging and convolutional neural networks provides quantitative measurements of mouse kidney tissue.

Researchers developed a pipeline for automated histopathology in the mouse kidney. This method uses normalized stimulated Raman imaging (NoRI) to make quantitative measurements of proteins and lipids.

Conventional histology relies on fixation, embedding, sectioning, and staining. These methods can distort cellular architecture, extract lipids, and introduce variability.

NoRI computationally corrects protein, lipid, and water Raman signals. This allows for quantitative biomass measurements while preserving tissue architecture.

Convolutional neural networks (CNN) perform morphological feature extraction from the images. These networks accurately classified tubule types with an F1=0.93F1=0.93 score.

The system also classified anatomical regions with an F1=0.91F1=0.91 and biological sex with an F1=0.97F1=0.97. Female tubules showed greater cytoplasmic lipid of $+6.5$ to 7.37.3 mg/mL and nuclear protein of $+26.3$ mg/mL.

In an acute kidney injury (AKI) model, the pipeline captured lipid and protein organization over 25 d postinjury. NoRI images outperformed raw stimulated Raman scattering (SRS) images in distinguishing early injury time points.

Key facts

  • NoRI enables quantitative, label-free protein and lipid measurements at high spatial resolution.
  • CNN analysis of NoRI images classified biological sex with an F1=0.97F1=0.97.
  • NoRI outperformed raw SRS images in distinguishing ischemia-reperfusion injury between 24 to 48 h.

All sections

Edition