Research map / Multi-agent systems and games / LLM agents
NeuroRefiner: Morphology-Aware Multi-Agent Refinement for 3D Fluorescence Microscopy Neuron Segmentation
HAIYANG YAN, Jinyue Guo, Yanchao Zhang, Bingqing Wang, Zhenchen Li, Jing Liu, Jiazheng Liu, linlin li, Hua Han
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Global Inspector alone (agents vs none) (ablation) | NeuroRefiner · ZBFWB | F1 (%): 76.39 -> 77.29 | No clear effect |
| add Refinement Advisor to Global Inspector (ablation) | NeuroRefiner · ZBFWB | F1 (%): 77.29 -> 83.47 | Improved |
| add Refinement Advisor to Global Inspector (ablation) | NeuroRefiner · CWMBS | F1 (%): 80.34 -> 87.14 | Improved |
| add Change Validator to Inspector plus Advisor (ablation) | NeuroRefiner · ZBFWB | F1 (%): 83.47 -> 86.13 | Improved |
| add Change Validator to Inspector plus Advisor (ablation) | NeuroRefiner · CWMBS | F1 (%): 87.14 -> 88.83 | Improved |
| add patch-level projections to global MIP input (ablation) | Global Inspector · ZBFWB | F1 (%): 77.36 -> 80.79 | Improved |
| add patch-level projections to global MIP input (ablation) | Global Inspector · CWMBS | F1 (%): 81.89 -> 85.09 | Improved |
| add 0-th Betti number input (ablation) | Global Inspector · ZBFWB | F1 (%): 80.79 -> 86.13 | Improved |
| use cross-attention in bottleneck instead of AdaLN (ablation) | TopoRefineNet · ZBFWB | F1 (%): 84.38 -> 86.13 | Improved |
| use cross-attention in bottleneck instead of channel-wise dot product (ablation) | TopoRefineNet · ZBFWB | F1 (%): 86.07 -> 86.13 | No clear effect |
| use larger Qwen3-4B text encoder instead of 0.6B (ablation) | TopoRefineNet · ZBFWB | F1 (%) | No clear effect |
| add two-stage curriculum training instead of single-stage on real defects (ablation) | TopoRefineNet · ZBFWB | F1 (%): 1.17% F1 degradation | Improved |
Closest papers by meaning
- Efficient Connectivity-Preserving Instance Segmentation with Supervoxel-Based Loss Function (AAAI 2025 · similarity 0.86)
- Learning Multimodal Volumetric Features for Large-Scale Neuron Tracing (AAAI 2024 · similarity 0.85)
- FISBe: A Real-World Benchmark Dataset for Instance Segmentation of Long-Range Thin Filamentous Structures (CVPR 2024 · similarity 0.85)
- Cross-Dimension Affinity Distillation for 3D EM Neuron Segmentation (CVPR 2024 · similarity 0.84)
- Pancakes: Consistent Multi-Protocol Image Segmentation Across Biomedical Domains (NeurIPS 2025 · similarity 0.83)
- MorphGrower: A Synchronized Layer-by-layer Growing Approach for Plausible Neuronal Morphology Generation (ICML 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
