Deep learning-based decoding of axonal ultrastructure in gene-edited mice using electron microscopy imaging

TL;DR

A deep learning framework for automated analysis of axonal and myelin structures in electron microscopy images to accelerate neurological disease research.

Problem

Myelin abnormalities or loss can lead to severe neurological impairments, but large-scale quantitative analysis of axonal components resolved by electron microscopy is difficult and time-consuming.

Approach

The authors developed a machine learning framework that automatically recognizes and quantifies multiple features of axons and myelin, including mitochondrial density and periaxonal area. They applied this framework to spinal cord fibers from variably hypomyelinated mice for validation.

Results & Contribution

The study showed that reductions in myelin sheath thickness and length correlate with changes in mitochondrial density and periaxonal area. This framework is expected to contribute to future insights into the changing relationships between axons, myelin, and mitochondria during neurological plasticity and myelin disease progression.