Quantifying axonal features of human superficial white matter from three-dimensional multibeam serial electron...

Area of Science:

• Neuroscience
• Microscopy
• Computational Biology

Background:

• Short-range association fibers in superficial white matter are crucial for human cognition.
• Microscopic characterization of these fibers is vital for understanding cortico-cortical connectivity.
• Previous imaging limitations hindered detailed analysis of axonal features in large volumes.

Purpose of the Study:

• To perform detailed morphological characterization of short-range association fibers in the human superficial white matter.
• To overcome imaging challenges by using advanced electron microscopy and automated analysis.
• To provide a comprehensive dataset for future neuroscience research.

Main Methods:

• Acquired multi-beam scanning electron microscopy data at 4 × 4 × 33 nm³ resolution.
• Segmented myelin and myelinated axons using deep convolutional neural networks (CNNs) with transfer learning.
• Analyzed a 200 × 200 × 112 μm³ volume of human superficial white matter.

Main Results:

• Successfully segmented 128,285 myelinated axons, including 70,321 >10 μm and 2102 >100 μm.
• Observed significant local variations in axon diameter (beading) and direction (undulation).
• Determined average morphometric parameters for axons >10 μm: inner diameter ~0.5 µm, outer diameter ~1 µm, g-ratio ~0.5.

Conclusions:

• This study provides novel insights into axonal morphometry in the human superficial white matter.
• The generated large-scale 3D electron microscopy dataset and segmentation results are valuable resources.
• The findings contribute to a better understanding of cortico-cortical connectivity and cognitive function.