Deep Learning for Analysis of Collagen Fiber Organization in Scar Tissue

The evaluation of the morphology and organization collagen fibers is critical in understanding wound healing tissue remodeling after a thermal injury skin. However, histological analysis conducted by pathologists often labor-intensive limited to qualitative evaluations scoring within narrow field view. In this study, we propose convolutional neural network (CNN) model classify Masson's trichrome (MT)-stained histology images burn-induced scar characterize microstructures normal quantitative manner. created on vivo rodent models prepared for MT-stained slides healing. A CNN developed, trained, tested with various sizes classification characterization. proposed classifies both (i.e., without burn, as control) at scales over 97% accuracy. features acquired from visually characterizes density directional variance distributed dermal layers whole images. deep learning technique can provide an objective reliable method rapidly assess quantify repair remodeling.