Module 15: Connectome Proofreading and Quality Control

🔍 Common Error Types

Segmentation errors (splits, merges) and synapse mislabels affect downstream analysis. Learn to spot patterns in raw EM imagery and segment overlays.

• Split vs. merge errors
• Ghost synapses and missing links
• Boundary ambiguity and stitching artifacts

📊 Visualization Tools

Interactive viewers like Neuroglancer enable efficient quality control. Understand how to use layers and cross-sections for visual checks.

• Configuring layers in Neuroglancer
• Using 3D mesh and skeleton modes
• Spotting errors across slices

📈 Metrics and Fixes

Evaluate accuracy using F1 score, precision, recall, and consistency with ground truth or heuristics. Apply edits or flag errors for correction.

• Segment overlap metrics
• Topology-aware metrics
• Manual editing vs. AI-assisted correction

🌟 COMPASS Integration

• Knowledge: Common connectomics error modes
• Skills: Visual identification, annotation tools
• Character: Persistence, accountability
• Meta-Learning: Recognizing error patterns across datasets

📚 References & Resources

• Funke et al., 2018. Large Scale Image Segmentation with Structured Loss Based Deep Learning for Connectomics. ECCV.
• Motta et al., 2019. Dense connectomic reconstruction in layer 4 of the somatosensory cortex. Science.
• Colab: "Segmentation Proofreading with Neuroglancer"

✅ Assessment

• Locate and document at least 3 segmentation errors in a provided volume
• Use Neuroglancer to propose a correction
• Reflect on how quality control affects downstream analysis