NeuronProofreader is a Python library for automated proofreading of neuron reconstructions. It corrects both split (over-fragmented neurons) and merge (erroneously joined neurons) errors in large-scale connectomics datasets. The library processes neuron fragments in SWC format using a multimodal pipeline that identifies and corrects topological errors in whole-brain reconstructions.
Figure: NeuronProofreader reconnects fragmented neuron segments into coherent traces.
The split correction proofreading pipeline consists of three main steps:
a. Graph Construction: Reads neuron fragments stored as SWC files and loads them into a Networkx graph.
b. Proposal Generation: Generates potential connections between nearby fragments.
c. GNN-Based Inference: Predicts whether to accept or reject proposals based on the geometric and image-based features.
Figure: Visualization of split correction pipeline.
In progress...
To use the software, in the root directory, run
pip install -e .To do...
For any inquiries, feedback, or contributions, please do not hesitate to contact us. You can reach us via email at [email protected] or connect on LinkedIn.
NeuronProofer is licensed under the MIT License.