Abstract

Functional tissue units (FTUs) composed of multiple cells like the glomerulus in the kidney nephron play important roles in health and disease. Histological staining is often used for annotation or segmentation of FTUs, but chemical stains can introduce artefacts through experimental factors that influence analysis. Secondly, many molecular -omics techniques are incompatible with common histological stains. To enable FTU segmentation and annotation in human kidney without the need for histological staining, we detail here the use of widefield autofluorescence (AF) microscopy as a simple, label-free modality that provides detailed renal morphology comparable to periodic acid-Schiff (PAS) stained tissue in both formalin-fixed paraffin-embedded (FFPE) and fresh frozen samples and with no tissue processing beyond sectioning. We demonstrate automated deep learning-based glomerular unit recognition and segmentation on PAS and AF images of the same tissue section from 9 fresh frozen samples and 9 FFPE samples. All training comparisons were carried out using registered AF microscopy and PAS stained whole slide images originating from the same section, and the recognition models were built with the exact same training and test examples. Measures of recognition performance, such as the Dice-Sorensen coefficient, the true positive rate, and the positive predictive value differed less than 2% between standard PAS and AF microscopy for both preservation methods. These results demonstrate that AF is a potentially powerful tool to study human kidney tissue, that it can serve as a label-free source for automated and manual annotation of tissue structures.