Deep learning to decipher the progression and morphology of axonal degeneration

Background: Axonal degeneration (AxD) is a pathological hallmark of many neurodegenerative diseases. Deciphering the morphological patterns of AxD will help to understand the underlying mechanisms and to develop effective therapeutic interventions. Here, we evaluated the progression of AxD in cortical neurons using a novel microfluidic device in combination with a deep learning tool, the EntireAxon, that we developed for the enhanced-throughput analysis of AxD on microscopic images. Results: The EntireAxon convolutional neural network sensitively and specifically segmented the features of AxD, including axons, axonal swellings, and axonal fragments, and its performance exceeded that of human expert raters. In an in vitro model of AxD in hemorrhagic stroke induced by the hemolysis product hemin, we detected the concentration- and time-dependent degeneration of axons leading to a decrease in axon area, while the axonal swelling and axonal fragment area increased. Time course analysis revealed that axonal swellings preceded axon fragmentation, suggesting that swellings may be reliable predictors of AxD. Using a recurrent neural network, we further identified four morphological patterns of AxD (granular, retraction, swelling, and transport degeneration) in cortical axons subjected to hemin. Conclusions: These findings indicate a morphological heterogeneity of AxD under pathophysiologic conditions. The combination of the microfluidic device with the EntireAxon deep learning tool enable the systematic analysis of AxD but also unravel a so far unknown intricacy in which AxD can occur in a disease context.