Poster #099
Developing a Neural Network to Identify Nerves and Neuromuscular Junctions
Mentor: Will Wang, PhD
Understanding the detailed structure of motor nerves within muscle tissue has long been a challenge. However, 3D imaging methods using tissue clearing and light sheet microscopy have allowed better visualization of innervation by immunofluorescence. Although new 3D images can be captured, they are data-intensive and require computational approaches for analysis. In order to accurately identify nerves and neuromuscular junctions, we need to train algorithms to classify images based on morphological features that we recognize with our eyes. With that said, we are using convolutional neural networks that mimic our vision, with strong applications of computer vision, to detect nerves and neuromuscular junctions. When trained over a set of 3D images of the mouse EDL muscle, a neural network can accurately distinguish the nerves and neuromuscular junctions better than traditional image analysis and machine learning approaches.