MorphoBranch: A Fine-Structure-Preserving Workbench for Morphometric Analysis of Branched Cellular Structures

Song Zhiying; Ling Hanyi; Wu Junyi; Jiang Yangbo

eess.IVOct 1, 2026

MorphoBranch: A Fine-Structure-Preserving Workbench for Morphometric Analysis of Branched Cellular Structures

Authors: Song Zhiying, Ling Hanyi, Wu Junyi, Jiang Yangbo

Organizations: School of Data Science, Zhejiang University of Finance and Economics, Hangzhou, 310018, China · College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, 310027, China · School of Information Technology and Artificial Intelligence, Zhejiang University of Finance and Economics, Hangzhou, 310018, China

Abstract

Background and Objectives: Fluorescence-labeled cellular arbors provide readouts of neuronal and microglial morphology, but fine and weakly labeled processes are prone to fragmentation and false connections that bias skeleton-based measurements. We present MorphoBranch, a fine-structure-preserving, human-reviewable workbench for morphometry of branched cellular structures. Methods: MorphoBranch combines a deterministic Morphometry Engine with an LLM-assisted Refinement Engine. The Mor- phometry Engine implements an image-to-graph workflow integrating multiscale structural evidence extraction, hysteresis segmen- tation, evidence-constrained skeleton refinement, and graph-based morphometry. The Refinement Engine maps natural-language requests to registered actions for parameter adjustment, preview execution, metric reporting, and unsupported-request handling, while image processing and quantitative computation remain deterministic and reviewable. Results: MorphoBranch was evaluated on two public neuronal axon datasets, AxonMIP and AxonStack, and the in-house Cell- Morph dataset of microglial fluorescence images. It achieved the highest Skeleton F1 and clDice and the lowest length-estimation error among the evaluated methods on all three datasets, while also achieving the highest Dice and IoU on AxonMIP and Axon- Stack. Across 150 natural-language tasks, the Refinement Engine achieved a 94.0% end-to-end success rate. Conclusions: These results demonstrate that MorphoBranch provides a reproducible, human-reviewable workflow for mor- phometric analysis of branched cellular structures. It supports fine-structure-preserving quantification across neuronal axon and microglial fluorescence images while maintaining inspectable and reproducible analysis workflows.

Figures & tables

Figure 1: Overall architecture of MorphoBranch. The workbench integrates a deterministic Morphometry Engine with an optional LLM-assisted Refinement Engine. The Morphometry Engine performs data preparation, structure extraction, topology refinement, and morphometric analysis, whereas the Refinement Engine translates natural-language requests into validated workflow adjustments through registered actions. Quantitative results, quality-control visualizations, refinement feedback, and reproducibility records are exported for review and downstream analysis.

Figure 2: Structure extraction workflow. The prepared analysis image $I(x,y)$ is denoised, corrected for slowly varying background, and transformed into a multiscale structural evidence map $R$ . Hysteresis thresholding identifies high-confidence seed regions and connected low-threshold structures, producing a binary structure mask while preserving weak, continuous processes.

Figure 3: Evidence-constrained skeleton refinement. Candidate bridges between skeleton endpoints are screened using endpoint proximity, directional consistency, and continuous image evidence. A candidate repair is accepted only when all criteria are satisfied, allowing image-supported interruptions to be repaired while rejecting unsupported shortcuts.

Category	Descriptor	Definition	Interpretation
Geometry	Foreground area	$A=\|M\|s_{x}s_{y}$ , where $\|M\|$ is the number of foreground pixels	Area occupied by the extracted structure
	Total skeleton length	$L_{\mathrm{total}}=\sum_{e\in E}\ell(e)$ , where $\ell(e)$ is the length of edge $e$	Overall extent of the skeletonized processes
	Skeleton density	$\rho=L_{\mathrm{total}}/A$	Skeleton extent normalized by foreground area
Topology	Connected components	Number of connected subgraphs in the refined graph $G^{*}$	Structural continuity and potential fragmentation
	Endpoints	$N_{\mathrm{end}}=\|\{v\in V:\deg(v)=1\}\|$	Peripheral terminations and potential skeleton breaks
	Junctions	$N_{\mathrm{junc}}=\|\{v\in V:\deg(v)\geq 3\}\|$ after junction consolidation	Branching complexity and potential false connections

Table 1: Core morphometric descriptors reported by MorphoBranch, organized into four categories characterizing geometry, topology, branch hierarchy, and spatial complexity of the extracted structures.

Figure 4: LLM-assisted refinement workflow. Natural-language requests are interpreted using the current project state and translated into structured action plans. Proposed actions are validated against the registered capabilities and executed in an isolated preview workflow. Users may inspect, revise, restore, or confirm the preview, and only confirmed adjustments are committed to deterministic execution with provenance recording.

Figure 5: MorphoBranch workbench interface. The local web interface integrates project and parameter management, intermediate-output inspection, morphometric analysis, and LLM-assisted refinement within a unified user-facing environment.

Category	Evaluation focus	Representative examples
Metric explanation and query	Interpretation and retrieval of available morphometric measurements without unsupported biological inference.	“What does total skeleton length mean?” “Can you diagnose whether this cell is healthy?”
Parameter refinement	Mapping user requests to registered parameters, bounded editing, preview execution, and handling of invalid changes.	“Show the top five paths and run a preview.” “Set the maximum gap to an invalid value.”
Analysis execution	Invocation of registered analysis actions and appropriate handling of missing prerequisites or unavailable results.	“Run a preview and report the main metrics.” “Report the metrics before any preview has been run.”
Export and reproducibility	Retrieval of generated artifacts, analysis configurations, and reproducibility information without fabricating unavailable outputs.	“Report the parameters used for this result.” “Where are the exported measurement files stored?”
Safety handling	Recognition and handling of requests outside registered capabilities, including unsupported reconstruction, unsafe file operations, and unregistered tools.	“Reconstruct the full 3D arbor from this 2D image.” “Run an unregistered external tool.”

Table 2: Task categories used for LLM-assisted refinement evaluation. Each category contained 30 natural-language tasks.

Dataset	Method	Dice $\uparrow$	IoU $\uparrow$	Skeleton F 1 $\uparrow$	clDice $\uparrow$	Length error (%) $\downarrow$
AxonMIP [ 3 ]	Otsu threshold [ 16 ]	$0.455\pm 0.147$	$0.305\pm 0.112$	$0.571\pm 0.199$	$0.549\pm 0.201$	$109.92\pm 165.65$
	Sauvola threshold [ 18 ]	$0.116\pm 0.053$	$0.062\pm 0.030$	$0.144\pm 0.089$	$0.120\pm 0.061$	$1860.95\pm 1184.43$
	Frangi ridge [ 8 ]	$0.478\pm 0.168$	$0.328\pm 0.135$	$0.622\pm 0.192$	$0.534\pm 0.190$	$38.05\pm 24.43$
	Meijering ridge [ 13 ]	$0.521\pm 0.091$	$0.357\pm 0.081$	$0.704\pm 0.080$	$0.666\pm 0.106$	$46.06\pm 40.36$
	MorphoBranch	$\mathbf{0.565\pm 0.086}$	$\mathbf{0.398\pm 0.079}$	$\mathbf{0.720\pm 0.083}$	$\mathbf{0.683\pm 0.110}$	$\mathbf{20.01\pm 11.86}$
AxonStack [ 2 ]	Otsu threshold [ 16 ]	$0.282\pm 0.139$	$0.171\pm 0.097$	$0.377\pm 0.214$	$0.372\pm 0.215$	$497.51\pm 506.13$

Table 3: Segmentation and skeleton-preservation performance of the Morphometry Engine on neuronal axon and microglial morphology datasets. Values are reported as mean

\pm

standard deviation.

Dataset	Input size (pixels)	Runtime / image (s)
AxonMIP	$512\times 512$	$0.189\pm 0.007$
AxonStack	$512\times 512$	$0.193\pm 0.016$
CellMorph	Variable	$0.138\pm 0.105$

Table 4: Runtime of the MorphoBranch pipeline. CellMorph contains images with variable dimensions (width: 250–664 pixels; height: 167–620 pixels).

Figure 6: Representative analysis of a CellMorph microglial image, showing intermediate processing outputs and branch-based morphometric visualizations.

Figure 7: A second representative analysis of a CellMorph microglial image with a distinct branching morphology.

Variant	Dice $\uparrow$	Skeleton F 1 $\uparrow$	clDice $\uparrow$	Length error (%) $\downarrow$
Full MorphoBranch	$\mathbf{0.503\pm 0.085}$	$\mathbf{0.612\pm 0.127}$	$\mathbf{0.614\pm 0.119}$	$48.50\pm 40.09$
w/o denoising	$0.502\pm 0.088$	$0.607\pm 0.135$	$0.607\pm 0.129$	$\mathbf{42.32\pm 38.31}$
w/o background correction	$0.402\pm 0.073$	$0.548\pm 0.117$	$0.554\pm 0.108$	$72.92\pm 49.65$
w/o multiscale ridge	$0.106\pm 0.194$	$0.131\pm 0.238$	$0.111\pm 0.206$	$87.02\pm 29.87$
w/o hysteresis	$0.442\pm 0.135$	$0.553\pm 0.189$	$0.519\pm 0.182$	$48.84\pm 23.48$

Table 5: Component ablation of the structure-extraction pipeline on AxonStack.

Figure 8: LLM-assisted refinement evaluation. (a) Task-level pass rates across five categories. (b) Capability-level scores for applicable tasks and end-to-end success.

Figure 9: Representative excerpts from LLM-assisted interaction cases. The examples illustrate metric explanation, parameter refinement, result interpretation, and safe handling of an unsupported 3D reconstruction request. Portions of the original responses are omitted for readability.

Capability	SNT [ 1 ]	Vaa3D [ 17 ]	NeuronMetrics [ 14 ]	NeurphologyJ [ 10 ]	MorphoBranch (ours)
Image-derived structure analysis	$\checkmark$	$\checkmark$	$\checkmark$	$\checkmark$	$\mathbf{\checkmark}$
Morphometric analysis	$\checkmark$	$\checkmark$	$\checkmark$	$\checkmark$	$\mathbf{\checkmark}$
Interactive visualization / inspection	$\checkmark$	$\checkmark$	$\checkmark$	$\checkmark$	$\mathbf{\checkmark}$
Native 3D tracing / reconstruction	$\mathbf{\checkmark}$	$\mathbf{\checkmark}$	–	–	–
Explicit workflow state and provenance	$\triangle$	–	–	–	$\mathbf{\checkmark}$
Natural-language workflow refinement	$\triangle$	–	–	–	$\mathbf{\checkmark}$

Table 6: Functional comparison of representative morphology-analysis tools. MorphoBranch emphasizes explicit workflow state and provenance together with natural-language workflow refinement.

\checkmark

: reported support;

\triangle

: later partial support; –: no identified support.

Explore similar work

Aug 31, 2026cs.CV

CORAL: A Benchmark for Structure-aware and Brain-wide Neuron Reconstruction in Light Microscopy

Automatic neuron reconstruction from light microscopy images is a central problem in computational neuroanatomy. While recent methods have achieved encouraging results on local image blocks, it remains unclear whether such progress translates to reconstruction that is both structurally accurate and scalable to the whole-brain scale. We present CORAL, the first benchmark for structure-aware evaluation of automatic neuron reconstruction from light microscopy images at both local and whole-brain scales. Built on a high-quality whole-brain fMOST dataset with carefully curated annotations, CORAL establishes two progressive tasks: block-level reconstruction, which evaluates reconstruction methods under limited spatial context, and brain-wide reconstruction, which assesses complete neuron reconstruction at the whole-brain scale. To account for topological correctness beyond geometric distance similarity, we introduce a structure-aware metric based on fiber prediction. To further achieve complete neuron reconstruction across the entire brain, we develop a brain-wide neuron tracing framework that extends arbitrary local reconstruction methods to the whole-brain scale through an iterative local-to-global process. Using this benchmark, we provide the first structure-aware comparison of mainstream methods for local neuron reconstruction and further evaluate their performance in brain-wide reconstruction. Our results underscore the importance of structure-aware evaluation and the need for more robust methods for complete neuron reconstruction.

Zekang Yang, Jiamin Li, Zhenghua Li +3

Department of Computer Science and Technology, Institute for AI, BNRist, Tsinghua University, Beijing 100084, China · Tsinghua Laboratory of Brain and Intelligence (THBI), IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing 100084, China · School of Basic Medical Sciences, Tsinghua University, Beijing, 100084, China +2

Neural Reconstruction Microscopy Structure-Aware Coral

Ask PDF

Sep 9, 2026cs.CV

SkNeXt enables topology-guided neuronal reconstruction from petabyte-scale microscopy data

Recent advances in high-resolution fluorescence and electron microscopy have enabled nanoscale imaging across increasingly large brain volumes, but the resulting terabyte- to petabyte-scale datasets make complete neuronal reconstruction prohibitively expensive in computation, data movement, and manual proofreading. Here, we present SkNeXt, a topology-first framework for scalable neuronal reconstruction from large volumetric microscopy datasets. Instead of densely processing entire image volumes, SkNeXt first converts neuronal morphology into compact SWC skeletons that preserve long-range connectivity. Proofreading is therefore focused on sparse neuronal trees, allowing branch, continuity, and connectivity errors to be corrected before high-resolution reconstruction. The corrected skeletons then serve as persistent structural priors for recovering detailed morphology while preserving neuronal identity and topology. Crucially, SkNeXt also uses neuronal skeletons as spatial indices for selective data access, retrieving high-resolution image regions only along reconstructed trajectories and bypassing most background and signal-free volumes. This substantially reduces I/O and computational overhead, allowing reconstruction cost to scale with neuronal morphology rather than total dataset size. Using SkNeXt, we reconstructed neurons from a petabyte-scale super-resolution fluorescence dataset of the mouse brain on a single GPU within one week, without requiring exhaustive dense inference across the complete imaging volume.

Jiayi Ding, Hu Zhao

Chinese Institute for Brain Research, Beijing, China

Neural Reconstruction Microscopy Skeleton Topology Next

Ask PDF

Jul 6, 2026cs.CV

Probe-EM: Targeted Neuron Tracing via Training-Free Semantic Verification

Establishing large-scale, high-resolution neural connectivity maps is fundamental to elucidating the structural basis of brain function. However, when processing terabyte- or petabyte-scale electron microscopy data, over-segmentation inherent in automated reconstruction algorithms remains a critical bottleneck, requiring extensive manual proofreading spanning person-years. To alleviate the heavy reliance on annotated data and the limited flexibility of conventional tracing methods, we propose a training-free, targeted neuron tracing framework. Specifically, we introduce a skeleton-guided Heuristic Spatial Search paradigm that leverages geometric priors to iteratively reconstruct neuronal morphologies through a probing-verification cycle. To achieve robust zero-shot semantic verification, we further develop a Dimension-Aware Semantic Verification strategy built upon the foundation model NeuroSAM 2. This strategy resolves intra-slice splits via Planar Ensemble Consensus and inter-slice splits via Axial Spatio-Temporal Propagation. Notably, we integrate the proposed workflow into the Neuroglancer visualization platform, enabling an interactive human-in-the-loop proofreading system. Experimental results demonstrate that the proposed method outperforms supervised baselines and reduces manual proofreading time by 33.4%. The source code is publicly available at https://github.com/HeadLiuYun/Probe-EM.

Liuyun Jiang, Yanchao Zhang, Jinyue Guo +5

State Key Laboratory of Brain Cognition and Brain-inspired Intelligence Technology, Institute of Automation, Chinese Academy of Sciences, Beijing, China · School of Future Technology, University of Chinese Academy of Sciences · School of Artificial Intelligence, University of Chinese Academy of Sciences +1

Human Connectome Project Neural Reconstruction Neurons Electron Microscopy

Ask PDF