SNEMI3D: 3D Neurite Segmentation in Electron Microscopy Images

SNEMI3D is a challenge dataset for training machine-learning algorithms to automatically segment neurites in 3D from electron microscopy (EM) image stacks. The large training dataset consists of manually delineated neurites from mouse cortex, produced by the Lichtman Lab at Harvard University and annotated by Daniel R. Berger. The challenge was organized in conjunction with the ISBI 2013 conference to gauge the state-of-the-art in automated neurite segmentation.

Use Cases

• Training 3D semantic segmentation models based on manually delineated neurites in EM images.
• Benchmarking automated neurite segmentation methods based on object classification accuracy in three dimensions.
• Developing algorithms to handle anisotropic resolution challenges inherent in serial-section EM imaging.
• Studying neural connectivity in mouse cortex based on the manually annotated image volumes.

Strengths

• Provides a large training dataset with manual expert annotations for neurites, enabling supervised learning.
• Includes a separate test dataset for objective method comparison and ranking.
• Data originates from a reputable source, the Lichtman Lab at Harvard University, with corresponding findings published in Cell (2015).

Limitations

• Row count, file formats, and column-level documentation are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.
• The EM imaging technique produces anisotropic volumes where the z-direction has low resolution, which may present a challenge for some 3D analysis methods.

Provenance

Source

Lichtman Lab at Harvard University (Daniel R. Berger, Richard Schalek, Narayanan 'Bobby' Kasthuri, Juan-Carlos Tapia, Kenneth Hayworth, Jeff W. Lichtman).

Collection Method

Electron microscopy imaging of mouse cortex, with manual annotation of neurites by Daniel R. Berger.