From voxels to knowledge: a practical guide to the segmentation of complex electron microscopy 3D-data.

Area of Science:

• Cellular and Molecular Biology
• Microscopy and Imaging Techniques
• Structural Biology

Background:

• Modern 3D electron microscopy provides detailed cellular ultrastructure.
• Analyzing complex cellular volumes requires effective feature extraction and segmentation.
• Data characteristics significantly influence the choice of segmentation approach.

Purpose of the Study:

• To evaluate different 3D segmentation approaches for electron microscopy data.
• To develop a decision-making framework for selecting optimal segmentation strategies.
• To enable better visualization, quantification, and comprehension of cellular 3D organization.

Main Methods:

• Utilized six 3D ultrastructural datasets from electron tomography, FIB-SEM, and SBF-SEM.
• Applied four segmentation methods: manual model building, manual tracing, semi-automated, and automated algorithms.
• Compared segmentation outcomes based on dataset characteristics like signal-to-noise ratio and feature density.

Main Results:

• The optimal segmentation approach varied depending on specific data set characteristics.
• No single method universally outperformed others across all datasets.
• A combination of objective data properties and user criteria determines the most successful strategy.

Conclusions:

• A proposed triage scheme aids in selecting appropriate 3D segmentation methods for electron microscopy.
• This framework facilitates informed decisions for analyzing complex cellular ultrastructure.
• Effective segmentation is crucial for accurate visualization and quantitative analysis in 3D cell biology.