Pipeline Overview

Sample preparation: tissue block and ultrathin sections

Stage 1 Sample Preparation

Biological tissue is prepared for electron microscopy imaging through fixation, staining, and embedding.

Inputs

  • Fresh tissue sample
  • Fixation protocols

Outputs

  • Resin-embedded tissue block
  • Serial ultrathin sections

Methods

Chemical fixation Heavy metal staining Resin embedding Ultramicrotomy
Electron microscopy imaging

Stage 2 EM Imaging

Serial section electron microscopy captures high-resolution images of the tissue volume.

Inputs

  • Serial tissue sections
  • Imaging parameters

Outputs

  • Raw EM image tiles
  • Imaging metadata

Methods

Serial section TEM SEM (ATUM) FIB-SEM
Image processing: stitching and alignment

Stage 3 Image Processing

Raw images are stitched, aligned, and corrected to create a contiguous 3D volume.

Inputs

  • Raw EM image tiles
  • Section ordering

Outputs

  • Aligned image volume
  • Multi-resolution pyramid

Methods

Tile stitching Section alignment Artifact correction Intensity normalization
Segmentation: distinct objects colored separately

Stage 4 Segmentation

Machine learning models identify and label individual neurons and cellular structures in the volume.

Inputs

  • Aligned EM volume
  • Training data

Outputs

  • Supervoxel segmentation
  • Initial neuron IDs
  • Merge/split predictions

Methods

Flood-filling networks Affinity prediction Agglomeration
Proofreading: inspect and correct errors

Stage 5 Proofreading

Human annotators correct segmentation errors—splits and mergers—to produce accurate neuron reconstructions. Primary tracer team activity.

Inputs

  • AI segmentation
  • EM imagery
  • Error predictions

Outputs

  • Corrected neuron segments
  • Validated root IDs
  • Proofread status flags

Methods

WebKnossos Neuroglancer SAM2 (AI assist)
Annotation: labeling features along a neuron

Stage 6 Annotation

Synapses, cell types, and other biological features are labeled on the proofread segmentation. Primary tracer team activity.

Inputs

  • Proofread neurons
  • EM imagery
  • Protocol guidelines

Outputs

  • Synapse locations
  • Pre/post assignments
  • Cell type labels
  • Neurotransmitter predictions

Methods

Neuroglancer CAVE Tampermonkey
Analysis: connectivity graph

Stage 7 Analysis

Connectivity data is queried, analyzed, and visualized to understand neural circuits.

Inputs

  • Synapse tables
  • Neuron IDs
  • Cell type annotations

Outputs

  • Connectivity matrices
  • Circuit diagrams
  • Statistical summaries

Methods

tracer_tools CAVE API Python / pandas Jupyter
Publication: paper with figures

Stage 8 Publication

Findings are documented and shared through papers, preprints, and public data releases.

Inputs

  • Analysis results
  • Figures and visualizations
  • Methods documentation

Outputs

  • Peer-reviewed papers
  • Preprints
  • Public datasets

Pipeline Tools

Tool Stage Purpose
Ultramicrotome + diamond knife 1 Cutting the resin-embedded block into serial ultrathin (~40 nm) sections
TEMCA (TEM Camera Array) 2 High-throughput serial-section transmission EM imaging (the modality behind FAFB and BANC)
SEAMLeSS / AlignTK 3 Stitching and elastic alignment of the section images into a contiguous 3D volume
CloudVolume 3, 4 Chunked volume storage and multi-resolution pyramid generation (precomputed format)
Flood-filling networks (FFN) 4 Automated machine-learning segmentation of neurons into supervoxels
PyChunkedGraph 4, 5 Editable segmentation graph that makes proofreading splits and merges possible
WebKnossos 5, 6 Volume annotation, proofreading, segmentation correction
Neuroglancer 5, 6 3D visualization, synapse annotation, data viewing
CAVE 6, 7 Annotation database, versioned data queries
tracer_tools 7 Python API for CAVE queries, batch processing, ID management
Tampermonkey 5, 6 Browser userscripts for WebKnossos, Neuroglancer, and Spelunker workflow enhancement

Supported Datasets

Dataset Organism Datastack Status
FlyWire FAFB Drosophila (whole brain) flywire_fafb_production Active
BANC Drosophila (brain + nerve cord) brain_and_nerve_cord Active