NeuroRefiner: Morphology-Aware Multi-Agent Refinement for 3D Fluorescence Microscopy Neuron Segmentation

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Haiyang Yan, Jinyue Guo, Yanchao Zhang, Bingqing Wang, Zhenchen Li, Jing Liu, Jiazheng Liu, Linlin Li, Hua Han
• Venue: arXiv: cs.CV
• Published: 2026-08-10
• Collected: 2026-08-11
• URL: http://arxiv.org/abs/2608.09636v1
• PDF: https://arxiv.org/pdf/2608.09636v1
• DOI: UNKNOWN
• Registry group: academic_apis_and_preprints
• Evidence role: canonical_academic_index
• Language: en
• Country / region: UNKNOWN

Plain-Language Brief

Researchers propose NeuroRefiner, a multi-agent system for improving the accuracy of 3D neuron segmentation in fluorescence microscopy. The system addresses challenges posed by the sparse and elongated morphology of neurons, which can cause existing methods to produce fragmented results. NeuroRefiner consists of three agents that collaborate to diagnose topological errors, generate correction instructions, and validate refinement quality. The researchers also introduce TopoRefineNet, a tool for instruction-guided segmentation refinement, and claim that NeuroRefiner outperforms state-of-the-art methods, particularly on the challenging ZBFWB dataset.

What the Document Says

• NeuroRefiner is proposed as a multi-agent system for 3D neuron segmentation in fluorescence microscopy
• The sparse and elongated morphology of neurons poses challenges to existing segmentation methods
• NeuroRefiner addresses these challenges by formalizing a human expert workflow involving iterative global observation and local editing
• NeuroRefiner comprises three collaborative agents: one for diagnosing topological errors, another for generating correction instructions, and a third for validating refinement quality
• TopoRefineNet is proposed as a 3D U-Net-based tool for instruction-guided segmentation refinement
• Cross-modality feature fusion is used in TopoRefineNet to generate refined masks
• NeuroRefiner uses multi-round agent reasoning and voxel-level editing to produce topologically more accurate segmentations with enhanced interpretability
• Experiments on BigNeuron, CWMBS, and ZBFWB datasets demonstrate NeuroRefiner's performance improvement over state-of-the-art methods, particularly achieving a 3.02% improvement in F1 score on the ZBFWB dataset

Key Concepts Mentioned

• NeuroRefiner
• multi-agent system
• neuron segmentation
• fluorescence microscopy
• topological errors
• global observation
• local editing
• TopoRefineNet
• 3D U-Net
• cross-modality feature fusion
• F1 score

Claims Made by the Document

• NeuroRefiner addresses challenges posed by the sparse and elongated morphology of neurons
• NeuroRefiner outperforms state-of-the-art methods
• NeuroRefiner achieves a 3.02% improvement in F1 score on the ZBFWB dataset

Evidence Used

• Experiments on BigNeuron, CWMBS, and ZBFWB datasets

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Accurate 3D neuron segmentation in fluorescence microscopy is critical for neuroscience. However, the sparse and elongated morphology of neurons poses significant challenges to existing segmentation methods. These methods struggle to preserve both local details and global topology, leading to fragmented results. To address this, we propose NeuroRefiner, a multi-agent system that formalizes the human expert workflow involving iterative global observation and local editing. Specifically, NeuroRefiner comprises three collaborative agents dedicated to diagnosing topological errors, generating correction instructions, and validating refinement quality. To facilitate agent instruction-guided segmentation refinement, we propose TopoRefineNet, a dedicated 3D U-Net-based tool that leverages cross-modality feature fusion to generate refined masks. Through multi-round agent reasoning and voxel-level editing, NeuroRefiner produces topologically more accurate segmentations with enhanced interpretability. Experiments on the BigNeuron, CWMBS, and ZBFWB datasets demonstrate that NeuroRefiner outperforms state-of-the-art methods, notably achieving a 3.02% improvement in F1 score on the challenging ZBFWB dataset.