bioRxivNeurosciencePreprint
Automated Reconstruction of a Serial-Section EM Drosophila Brain with Flood-Filling Networks and Local Realignment
Peter H. LiiD, Larry F. Lindsey, Michał JanuszewskiiD, Zhihao ZhengiD, Alexander Shakeel BatesiD, István TaisziD, Mike Tyka, Matthew Nichols, Feng LiiD, Eric PerlmaniD, Jeremy Maitin-ShepardiD, Tim M. BlakelyiD, Laramie LeavittiD, Gregory S.X.E. JefferisiD, Davi D. BockiD, Viren JainiD
Abstract
Reconstruction of neural circuitry at single-synapse resolution is a key target for improving understanding of the nervous system in health and disease. Serial section transmission electron microscopy (ssTEM) is among the most prolific imaging methods employed in pursuit of such reconstructions. We demonstrate how Flood-Filling Networks (FFNs) can be used to computationally segment a forty-teravoxel whole-brain Drosophila ssTEM volume. To compensate for data irregularities and imperfect global alignment, FFNs were combined with procedures that locally re-align serial sections as well as dynamically adjust and synthesize image content. The proposed approach produced a largely merger-free segmentation of the entire ssTEM Drosophila brain, which we make freely available. As compared to manual tracing using an efficient skeletonization strategy, the segmentation enabled circuit reconstruction and analysis workflows that were an order of magnitude faster.
- organism: Drosophila
- assay: imaging
- research focus: cell type annotation
- research focus: clustering
- research focus: trajectory inference
- research focus: differential expression
- approach: deep learning
- approach: statistical model
Abstract summary
Claude Opus 5.5 · xhigh effort · 18 Oct 2026 · from the abstract · CC BY-NC-ND 4.0
Flood-filling networks with local section realignment segment a forty-teravoxel whole-brain Drosophila electron microscopy volume, making circuit reconstruction an order of magnitude faster than manual tracing.
Mapping circuits at single-synapse resolution means segmenting serial-section electron microscopy volumes of enormous size, and data irregularities and imperfect global alignment get in the way of automated segmentation. Pairing flood-filling networks with local realignment and on-the-fly image repair produced a segmentation of an entire fly brain that the authors make freely available.
Key results
ClaimedA whole Drosophila brain imaged by serial-section transmission electron microscopy was segmented automatically, and the segmentation is reported to be largely free of merge errors.
Applied to a forty-teravoxel whole-brain volume and described as largely merger-free; no split or merge rates, skeleton-based accuracy metrics or proofreading effort are reported in the abstract. · Abstract
ClaimedRobustness came from pairing the segmentation network with procedures that realign neighbouring sections locally and adjust or synthesise image content where the data are irregular, rather than relying on a single global alignment.
Stated as the approach taken to handle data irregularities and imperfect global alignment; no ablation separating the contribution of realignment from that of the network is given. · Abstract
SuggestedDownstream circuit reconstruction and analysis on top of the segmentation ran much faster than tracing the same circuits by hand.
Reported as an order of magnitude faster than manual tracing with an efficient skeletonization strategy; no timings, neuron counts or annotator numbers back the comparison in the abstract. · Abstract
Methods, data and reproducibility
Flood-Filling Networks are run over a forty-teravoxel whole-brain Drosophila serial-section transmission electron microscopy volume, combined with local re-alignment of adjacent sections and dynamic adjustment or synthesis of image content where the data are defective; the resulting segmentation is compared against manual skeleton tracing for reconstruction speed.
- Whole-brain Drosophila serial-section transmission electron microscopy volume · Drosophila · n = forty-teravoxel volume
- Methods & tools
- Flood-Filling Networks
- Reproducibility gaps
- The segmentation is said to be freely available, but no URL, repository or accession is printed, and no code, model weights or compute requirements are described.
Limitations (2)
- Segmentation quality is described qualitatively as largely merger-free, with no error rates or metrics, so how much proofreading a user would still face is unknown. · noted in this summary
- The speed comparison rests on one manual tracing strategy and is quoted only as an order of magnitude, without task definitions or measured times. · noted in this summary
Read this if you reconstruct circuits from serial-section electron microscopy and need automated segmentation that survives imperfect alignment at whole-brain scale.
- Citations
- 60
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- High-Precision Automated Reconstruction of Neurons with Flood-filling Networks
Michał Januszewski, Joergen M. R. Kornfeld, … Viren Jain·Nature MethodsPeer-reviewed·Jul 11, 2018
- En bloc preparation of Drosophila brains enables high-throughput FIB-SEM connectomics
Zhiyuan Lu, C. Shan Xu, … Ian A. Meinertzhagen·Frontiers in Neural CircuitsPeer-reviewed·Dec 16, 2022
- Petascale neural circuit reconstruction: automated methods
Thomas Macrina, Kisuk Lee, … H. Sebastian Seung·bioRxiv·Aug 4, 2021
- Whole-Brain Serial-Section Electron Microscopy In Larval Zebrafish
David G. C. Hildebrand, Marcelo Cicconet, … Florian Engert·NaturePeer-reviewed·May 1, 2017
- Accelerated EM Connectome Reconstruction using 3D Visualization and Segmentation Graphs
Philip M. Hubbard, Stuart Berg, … Stephen M. Plaza·bioRxiv·Jan 17, 2020
- Synapse Detection Efficiency in EM Drosophila Connectomics
Louis K. Scheffer·bioRxiv·Oct 16, 2025