BackgroundTraditional histopathologic evaluation of peripheral nerve using brightfield microscopy is resource-intensive, necessitating complex sample preparation. Label-free imaging techniques paired with artificial intelligence-based image reconstruction and segmentation may facilitate peripheral nerve histomorphometry. New methodHerein, the utility of label-free phase contrast techniques paired with artificial intelligence-based image processing for imaging of mammalian peripheral nerve is demonstrated. ResultsFresh frozen murine sciatic nerve sections were imaged in transmission modalities using differential interference and phase contrast microscopy and in epifluorescent modality following staining with myelin-specific dye. Deep learning was employed to predict epifluorescent images from transmitted phase contrast images, and machine learning employed for automated segmentation of myelinated axons for reporting of axons counts and g-ratios. Comparison with existing methodsConventional peripheral nerve histomorphometry employs resource intensive resin embedding, ultra-microtome sectioning, and staining steps. Herein we demonstrate feasibility of high-throughput nerve histomorphometry via label-free phase contrast imaging of frozen sections. ConclusionsClinical applications of label-free phase contrast microscopy paired with deep learning algorithms are discussed.
Read full abstract