Pith. sign in

REVIEW 4 major objections 4 minor 46 references

Preserving instance continuity and length in segmentation through connectivity-aware loss computation

T0 review · 4 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that two connectivity-aware loss functions reduce segmentation discontinuities in elongated biological structures and thereby improve downstream length estimation.

desk verdict Clearly specified loss variants and honest baselines, but the paper's own overlapping-instances metric contradicts its headline claim in the main experiment; needs error bars and a revised claim before the evidence is convincing. read the letter →

arxiv 2509.03154 v1 pith:YIWNVPYG submitted 2025-09-03 cs.CV

Karol Szustakowski , Luk Frank , Julia Esser , Jan Gründemann , Marie Piraud This is my paper · ORCID
classification cs.CV
keywords axoninitialsegmentconnectivity-awarelossNegativeCenterlineSimplifiedTopologytubularstructuresegmentationinstancecontinuitylight-sheetfluorescencemicroscopylengthdistribution
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Segmentation models optimized for voxel-wise accuracy tend to split elongated biological structures wherever the input signal fades, and every split corrupts downstream measurements such as axon length. This paper proposes two loss functions that make continuity itself the optimization target: Negative Centerline Loss penalizes the fraction of the label's centerline that the prediction fails to cover, and Simplified Topology Loss re-applies binary cross-entropy inside the specific regions that would connect or disconnect instances. On 3D light-sheet fluorescence microscopy of axon initial segments, both losses reduce the number of prediction fragments per labeled instance and improve the Wasserstein distance between predicted and labeled length distributions, with the centerline loss best in the full-data setting and the topology loss best when training data are scarce. The paper additionally reports that downscaling the in-plane resolution by three improves connectivity, a result that runs against the usual assumption that higher resolution is better. The central claim is that structural priors embedded in the loss, rather than more parameters or more data, are the effective lever for continuity in elongated-structure segmentation.

What carries the argument

Negative Centerline Loss is a differentiable loss whose support is the label's soft skeleton; it computes |(1−P)∘L_CL|/|L_CL|, so the gradient is nonzero only on the centerline and largest precisely at gaps in the prediction. Simplified Topology Loss is a non-differentiable region-selection procedure—threshold, dilation, connected-component labeling—that finds label regions bordering at least two distinct prediction regions, as well as prediction regions overlapping no label, and then applies binary cross-entropy only inside those regions. The skeletonization is the soft-skeleton pooling routine from clDice run to convergence, with O(nd) worst-case time; the region finder runs in amortized n

What would settle it

Segment synthetic tubular objects with known true lengths and controlled signal dropout using the proposed losses, then compare the graph-diameter length estimates against the known lengths; if the Wasserstein-distance improvement disappears or reverses when true length is known, the downstream-length claim is an artifact of the proxy. As a second check, re-measure real AIS length by manually tracing from soma to tip on the same instances and see whether the ranking of losses survives.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that two simple losses can make a U-Net segment elongated structures as connected instances rather than as fragments, and that this translates into better length statistics downstream. Negative Centerline Loss computes the soft skeleton of the label and measures the fraction of that skeleton not intersected by the prediction; its gradient is therefore concentrated exactly where a gap breaks the instance. Simplified Topology Loss binarizes the prediction, dilates it, labels connected components, and identifies label regions that touch at least two distinct prediction regions—the bridges that, if added, would join fragments—then retrains the n

Load-bearing premise

The length-based evaluation assumes that the longest path through a skeletonization of the segmentation equals the object's true AIS length, and that the manual labels are complete enough to define the target centerline; if either assumption fails, the reported length improvements may not reflect true biological length.

Editorial extensions

If this is right

  • If the central claim holds, segmentation models trained with these losses produce fewer split instances per labeled AIS, so automated length distributions match manual annotations more closely than baseline training does.
  • In the full-data 3x-downscaled setting, Negative Centerline Loss lowers the Wasserstein distance of length distributions from 55.5 to 42.9 and raises precision from 0.822 to 0.862 without meaningfully changing Dice.
  • Simplified Topology Loss retains most of the continuity benefit and gives the best results in the halved-dataset run (Wasserstein 167.4 vs 244.3 for the baseline), making it the recommended choice when labels are scarce.
  • Both losses add roughly 13% to per-epoch training time relative to the baseline, so the continuity gain comes at modest compute cost.
  • Because the losses are agnostic to imaging modality and structure shape, the authors expect them to transfer to other elongated-structure tasks, such as vascular or road-network segmentation, where continuity is important.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural extension the authors leave implicit: Simplified Topology Loss's region finder already identifies exactly where the model is fragmented, so it could double as an active-learning acquisition function that tells annotators which locations to correct.
  • The downscaling result hints that continuity errors in 3D patch-based training are driven by limited context or receptive field rather than by resolution; an explicit test would be to train at full resolution with larger patches or dilated convolutions and see whether the gap to downscaled runs closes.
  • The same 'connectivity as the objective' logic could apply across time rather than space—e.g., video instance segmentation under occlusion—where the analogue of signal dropout is a missing frame and the analogue of length is track duration.
  • A direct validation of the length proxy—comparing graph diameter against manually traced soma-to-tip paths on the same instances—would show whether the reported length improvements reflect true biological length or an artifact of the measurement.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper addresses a practical problem in biomedical segmentation of elongated structures: preserving instance continuity and length when signal dropout causes disconnected predictions. It proposes two new loss functions, Negative Centerline Loss (penalizing uncovered label centerline) and Simplified Topology Loss (reapplying BCE in regions causing discontinuities), and evaluates them on a 3D light-sheet fluorescence expansion microscopy dataset of axon initial segments using an nnU-Net architecture. Comparisons are made against a baseline nnU-Net, clDice, and a persistence-diagram topology loss, under both original and 3x-downscaled conditions, plus a small-dataset setting. The paper reports improved Wasserstein distance of length distributions and precision, and claims the losses reduce segmentation discontinuities per instance. The source code is released as an nnU-Net fork.

Significance. If the central claim were established, the paper would make a useful, computationally modest contribution: two simple losses with clear algorithmic descriptions, an open implementation, and a real biological dataset where continuity matters. The paper also raises an interesting subsidiary point about downscaling improving connectivity. However, the evidence as presented is not yet convincing for the headline claim. The direct metric for per-instance discontinuity does not improve for the main method, no uncertainty quantification is provided, and important hyperparameters and experimental choices are tuned on the validation signal. The length-proxy and label-quality issues further complicate the downstream interpretation. The work is potentially valuable, but it needs a substantial revision to align claims with evidence.

major comments (4)
  1. [Abstract and Table 2, Section 2.4] The abstract claims the losses 'reduce the number of segmentation discontinuities per instance.' The metric that most directly measures this is 'overlapping instances' (average number of prediction instances per label instance, ideally 1). In the main 3x-downscaled run, Negative Centerline Loss gives 1.042, which is worse than the baseline's 1.038; Simplified Topology Loss is 1.054. The reported precision and Wasserstein improvements are indirect: precision can improve by eliminating false-positive instances rather than fewer splits, and Wasserstein distance concerns the aggregate length distribution, which can shift for reasons other than within-instance continuity. A direct per-instance split count, or a re-framing of the claim away from 'fewer discontinuities,' is needed.
  2. [Section 2.3 and Table 1] The auxiliary loss weights w_eval are selected 'empirically to make w_eval as large as possible while still maintaining good segmentation quality improvement as measured by the Dice validation metric during the first epochs.' This means each method is compared after tuning its weighting on the validation signal. No sensitivity analysis or fixed-weight protocol is reported, so part of the observed Wasserstein/precision gain could reflect this tuning rather than the loss design itself. The authors should report results for a range of weights or justify that the chosen weights are not the source of the differences.
  3. [Section 2.5 and Figure 3] The length-based claims rest on a proxy: skeleton graph diameter of each instance after discarding border-intersecting instances. The paper acknowledges there is no standardized AIS length measurement and does not validate this proxy against manual tracing or an independent length estimate. In addition, Figure 3's caption states that in some regions models 'outperform the ground truth label' because annotators missed dim AIS. Since precision, recall, and length metrics all use these labels as ground truth, and since Negative Centerline Loss is defined against the label centerline, incomplete labels can penalize correct bridging behavior. The authors should quantify label uncertainty or demonstrate robustness to it.
  4. [General experimental reporting (all tables)] All performance metrics are reported as 5-fold means without error bars, confidence intervals, or significance tests. Several comparisons, e.g., overlapping instances 1.038 vs. 1.042, are small relative to likely fold-to-fold variance. The claim that one loss is preferable to another requires at least per-fold results or a paired statistical test. This is especially important because the experimental setup fixes many choices (downscaling, spacing correction, loss weights) after exploratory analysis on the same data.
minor comments (4)
  1. [Section 2.4] The 'Strictly standardized mean difference' (SSMD) metric is never defined or referenced. Please provide its formula and explain how it is computed from the length distributions.
  2. [Section 2.5] The length computation uses voxel counts with z-stack spacing considered, but it is not stated whether the skeletonization is performed on downsampled-then-upsampled masks or at original resolution. Clarify the exact pipeline used for the downscaled experiments.
  3. [Section 2.1 / Algorithm 2] The soft-skeleton routine is a key component, but the pooling window sizes and the stopping criterion are not fully specified. The text says 'until convergence,' while the pseudocode loops while |I| > 0; please clarify how this behaves for 3D inputs and how it differs from clDice's fixed-iteration variant.
  4. [Introduction, Section 1.1] The AISuite URL is given in the text but not in the references. Also, the reference formatting for [3] is incomplete ('Neurophotonics, 6:1').

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the proposed losses and reported metrics are connected empirically, not by definition, and no load-bearing self-citation chain or fitted-input-as-prediction step is present.

full rationale

This paper's central claims are empirical, not derived by definition. Negative Centerline Loss is an explicit algorithmic objective: it computes the fraction of the label's soft-skeleton not intersecting the prediction (Algorithm 1). Simplified Topology Loss re-applies binary cross-entropy on regions where label-prediction differences border at least two prediction components (Algorithm 3). Neither loss is defined in terms of the reported evaluation metrics (Wasserstein distance of skeleton-diameter lengths, SSMD, precision, recall, or overlapping-instance count), so the favorable Table 2 results cannot reduce by construction to the training objective. The evaluation is performed on held-out folds and on upsampled masks, providing independent empirical evidence. There is also no load-bearing self-citation chain: the baselines (clDice, topology loss, Topograph) and the soft-skeleton subroutine are external prior works, cited normally, and none of them is by the current authors. The disclosed empirical choices—loss weights selected during early validation and 3x downscaling chosen after exploratory analysis—are tuned inputs rather than outputs disguised as predictions; they are limitations for generalizability, not circularity. The skeptic's observation that Negative Centerline worsens the direct overlapping-instances metric (1.042 vs 1.038) in the flagship downscaled run is a substantive contradiction of the abstract's discontinuity claim, but that is an empirical validity problem, not a definitional or self-citation circularity. Likewise, the graph-diameter length proxy and the possibility of incomplete napari labels are assumptions affecting biological interpretation, not circular definitions of the loss or of the metric. In short, the derivation chain from loss definitions to reported metrics is self-contained; no circular step is exhibited.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The paper adds no invented physical entities. Its claims rest on a handful of domain assumptions: the skeleton-graph diameter is treated as AIS length, manual labels are trusted in signal-dropout regions, and the one-to-one instance matching defines the reported metrics. The free parameters are the per-loss evaluation weights (Table 1) and the 3x downscaling choice.

free parameters (2)
  • weval (weight of the auxiliary loss) = clDice 3, Topology 1, Negative Centerline 3, Simplified Topology 4
    Section 2.3 and Table 1: weights selected empirically to maximize the auxiliary loss while keeping Dice acceptable in early epochs.
  • XY downscaling factor = 3
    Section 3.1: chosen from exploratory analysis; it improved continuity, while z-downscaling and default spacing correction were not used.
assumptions (4)
  • domain assumption Graph diameter of the skeleton of a segmentation mask is a valid proxy for AIS length
    Section 2.5: length is computed by skeletonization, graph construction, and diameter; no validation against manual tracing or biological ground truth.
  • domain assumption Manual napari annotations are accurate enough in signal-dropout regions to train a continuity-preserving loss
    Section 2.5 and Figure 3: annotators missed dim AIS; the paper notes models sometimes 'outperform the ground truth label'.
  • domain assumption One-to-one instance matching is a faithful basis for precision, recall, and overlap metrics
    Section 2.4: instances can be matched once; splits lower precision but do not lower recall if the label is matched with at least one part.
  • ad hoc to paper Penalizing uncovered label centerline regions is sufficient to teach models to bridge missing-signal gaps
    Core design premise of Negative Centerline Loss (Algorithm 1); plausible but not independently validated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Preserving instance continuity and length in segmentation through connectivity-aware loss computation." pith.science (2026). https://pith.science/paper/YIWNVPYG

@misc{pith2026250903154,
  author       = {Pith},
  title        = {Pith review of: Preserving instance continuity and length in segmentation through connectivity-aware loss computation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YIWNVPYG}},
  note         = {Machine review of arXiv:2509.03154}
}
read the original abstract

In many biomedical segmentation tasks, the preservation of elongated structure continuity and length is more important than voxel-wise accuracy. We propose two novel loss functions, Negative Centerline Loss and Simplified Topology Loss, that, applied to Convolutional Neural Networks (CNNs), help preserve connectivity of output instances. Moreover, we discuss characteristics of experiment design, such as downscaling and spacing correction, that help obtain continuous segmentation masks. We evaluate our approach on a 3D light-sheet fluorescence microscopy dataset of axon initial segments (AIS), a task prone to discontinuity due to signal dropout. Compared to standard CNNs and existing topology-aware losses, our methods reduce the number of segmentation discontinuities per instance, particularly in regions with missing input signal, resulting in improved instance length calculation in downstream applications. Our findings demonstrate that structural priors embedded in the loss design can significantly enhance the reliability of segmentation for biological applications.

Figures

Figures reproduced from arXiv: 2509.03154 by the authors.

Figure 1
Figure 1. An example of model outputting disconnected labels due [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Comparison of the supports of selected loss functions. The striped pattern of [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparison of the segmentations resulting from the proposed loss functions, displayed as a maximum projection from multiple [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 45 canonical work pages

  1. [1]

    Optimizing the Dice Score and Jaccard Index for Medical Image Segmentation: Theory & Practice

    Jeroen Bertels, Tom Eelbode, Maxim Berman, Dirk Van- dermeulen, Frederik Maes, Raf Bisschops, and Matthew B. Blaschko. Optimizing the dice score and jaccard index for medical image segmentation: Theory & practice. CoRR, abs/1911.01685, 2019. 1

    work page Pith review arXiv 1911

  2. [2]

    Clough, Israel Valverde, Giovanni Montana, and Andrew P

    Nick Byrne, James R. Clough, Israel Valverde, Giovanni Montana, and Andrew P. King. A persistent homology-based topological loss for cnn-based multiclass segmentation of cmr. IEEE Transactions on Medical Imaging , 42(1):3–14,

    work page

  3. [3]

    Light-sheet fluo- rescence expansion microscopy: Fast mapping of neural cir- cuits at super resolution

    Jana B ¨urgers and Juan Rodriguez-Gatica. Light-sheet fluo- rescence expansion microscopy: Fast mapping of neural cir- cuits at super resolution. Neurophotonics, 6:1, 2019. 1, 2

    work page 2019

  4. [4]

    Tillberg, and Edward S

    Fei Chen, Paul W. Tillberg, and Edward S. Boyden. Expan- sion microscopy. Science, 347(6221):543–548, 2015. 1

    work page 2015

  5. [5]

    Dumitrescu, Mark D

    Adna S. Dumitrescu, Mark D. Evans, and Matthew S. Grubb. Evaluating tools for live imaging of structural plasticity at the axon initial segment. Frontiers in Cellular Neuroscience, V olume 10 - 2016, 2016. 5

    work page 2016

  6. [6]

    Topological persistence and simplification

    Herbert Edelsbrunner, David Letscher, and Afra Zomoro- dian. Topological persistence and simplification. 2003. 2

    work page 2003

  7. [7]

    Hoogenraad

    Am ´elie Fr´eal and Casper C. Hoogenraad. The dynamic axon initial segment: From neuronal polarity to network home- ostasis. Neuron, 113(5):649–669, 2025. 1

    work page 2025

  8. [8]

    Contribution of axon initial segment structure and channels to brain pathology

    Juan Garrido. Contribution of axon initial segment structure and channels to brain pathology. Cells, 12:1210, 2023. 1

    work page 2023

Show all 46 references
  1. [9]

    Activity-dependent re- location of the axon initial segment fine-tunes neuronal ex- citability

    Matthew Grubb and Juan Burrone. Activity-dependent re- location of the axon initial segment fine-tunes neuronal ex- citability. Nature, 465:1070–4, 2010. 1

    2010

  2. [10]

    Neuroscience: A plastic axonal hotspot

    Jan Gr ¨undemann and Michael Hausser. Neuroscience: A plastic axonal hotspot. Nature, 465:1022–3, 2010. 1

    2010

  3. [11]

    Impact of loss function in deep learning methods for accurate retinal vessel segmenta- tion, 2022

    Daniela Herrera, Gilberto Ochoa-Ruiz, Miguel Gonzalez- Mendoza, and Christian Mata. Impact of loss function in deep learning methods for accurate retinal vessel segmenta- tion, 2022. 1

    2022

  4. [12]

    Topology-preserving deep image segmentation, 2019

    Xiaoling Hu, Li Fuxin, Dimitris Samaras, and Chao Chen. Topology-preserving deep image segmentation, 2019. 2

    2019

  5. [13]

    Heterogeneity of the axon initial segment in interneurons and pyramidal cells of rodent visual cortex

    Felix H ¨offlin, Alexander Jack, Christian Riedel, Julia Mack- Bucher, Johannes Roos, Corinna Corcelli, Christian Schultz, Petra Wahle, and Maren Engelhardt. Heterogeneity of the axon initial segment in interneurons and pyramidal cells of rodent visual cortex. Frontiers in Cel...

    2017

  6. [14]

    Jaeger, Simon Kohl, Jakob Wasserthal, Gregor Koehler, Tobias Norajitra, Sebastian Wirkert, and Klaus H

    Fabian Isensee, Jens Petersen, Andre Klein, David Zim- merer, Paul F. Jaeger, Simon Kohl, Jakob Wasserthal, Gregor Koehler, Tobias Norajitra, Sebastian Wirkert, and Klaus H. Maier-Hein. nnu-net: Self-adapting framework for u-net- based medical image segmentation, 2018. 4

    2018

  7. [15]

    Fabian Isensee, Tassilo Wald, Constantin Ulrich, Michael Baumgartner, Saikat Roy, Klaus Maier-Hein, and Paul F. Jaeger. nnu-net revisited: A call for rigorous validation in 3d medical image segmentation, 2024. 4

    2024

  8. [16]

    Sensory input drives rapid homeostatic scaling of the axon initial segment in mouse barrel cortex

    Nora Jamann, Dominik Dannehl, Nadja Lehmann, Robin Wagener, Corinna Thielemann, Christian Schultz, Jochen Staiger, Maarten Kole, and Maren Engelhardt. Sensory input drives rapid homeostatic scaling of the axon initial segment in mouse barrel cortex. Nature Communications, 12, 2021. 1

    2021

  9. [17]

    Manjunath

    Jiaxiang Jiang, Michael Goebel, Cezar Borba, William Smith, and B. Manjunath. A robust approach to 3d neu- ron shape representation for quantification and classification. BMC Bioinformatics, 24, 2023. 5

    2023

  10. [18]

    Iyer, Eric A

    Dakai Jin, Krishna S. Iyer, Eric A. Hoffman, and Punam K. Saha. A new approach of arc skeletonization for tree-like ob- jects using minimum cost path. In 2014 22nd International Conference on Pattern Recognition, pages 942–947, 2014. 5

    2014

  11. [19]

    Automated sorting of neu- ronal trees in fluorescent images of neuronal networks us- ing neurotreetracer

    Cihan Kayasandik, Pooran Negi, Fernanda Laezza, Manos Papadakis, and Demetrio Labate. Automated sorting of neu- ronal trees in fluorescent images of neuronal networks us- ing neurotreetracer. Scientific reports, 8(1), 2018. Publisher Copyright: © 2018 The Author(s). 2

    2018

  12. [20]

    Dynamic deep net- works for retinal vessel segmentation, 2019

    Aashis Khanal and Rolando Estrada. Dynamic deep net- works for retinal vessel segmentation, 2019. 1

    2019

  13. [21]

    Action potential generation requires a high sodium channel density in the axon initial segment

    Maarten Kole, Susanne Ilschner, Bjoern Kampa, Stephen Williams, Peter Ruben, and Greg Stuart. Action potential generation requires a high sodium channel density in the axon initial segment. Nature neuroscience, 11:178–86, 2008. 1

    2008

  14. [22]

    The electrical signifi- cance of axon location diversity

    Maarten HP Kole and Romain Brette. The electrical signifi- cance of axon location diversity. Current Opinion in Neuro- biology, 51:52–59, 2018. Cellular Neuroscience

    2018

  15. [23]

    Presynap- tic activity regulates na channel distribution at the axon ini- tial segment

    Hiroshi Kuba, Yuki Oichi, and Harunori Ohmori. Presynap- tic activity regulates na channel distribution at the axon ini- tial segment. Nature, 465:1075–8, 2010. 1

    2010

  16. [24]

    Measure anything: Real-time, multi- stage vision-based dimensional measurement using segment anything, 2024

    Yongkyu Lee, Shivam Kumar Panda, Wei Wang, and Mo- hammad Khalid Jawed. Measure anything: Real-time, multi- stage vision-based dimensional measurement using segment anything, 2024. 5

    2024

  17. [25]

    Alterations in the axon initial segment plasticity is involved in early pathogenesis in alzheimer’s disease

    Yu Li, Han Wang, Yiming Wang, Zhiya Chen, Yiqiong Liu, Wu Tian, Xinrui Kang, Abolghasem Pashang, Don Kulasiri, Xiaoli Yang, Hung Wing Li, and Yan Zhang. Alterations in the axon initial segment plasticity is involved in early pathogenesis in alzheimer’s disease. MedComm, 5(11):e768,

  18. [26]

    Berger, Alexander Weers, Nico Stucki, Daniel Rueckert, Ulrich Bauer, and Johannes C

    Laurin Lux, Alexander H. Berger, Alexander Weers, Nico Stucki, Daniel Rueckert, Ulrich Bauer, and Johannes C. Paet- zold. Topograph: An efficient graph-based framework for strictly topology preserving image segmentation, 2025. 2

    2025

  19. [27]

    Jun Ma, Jianan Chen, Matthew Ng, Rui Huang, Yu Li, Chen Li, Xiaoping Yang, and Anne L. Martel. Loss odyssey in medical image segmentation. Medical Image Analysis , 71: 102035, 2021. 1

    2021

  20. [28]

    Tizabi, Florian Buettner, Evangelia Christodoulou, Ben Glocker, Fabian Isensee, Jens Kleesiek, Michal Kozubek, Mauricio Reyes, Michael A

    Lena Maier-Hein, Annika Reinke, Patrick Godau, Minu D. Tizabi, Florian Buettner, Evangelia Christodoulou, Ben Glocker, Fabian Isensee, Jens Kleesiek, Michal Kozubek, Mauricio Reyes, Michael A. Riegler, Manuel Wiesen- farth, A. Emre Kavur, Carole H. Sudre, Michael Baum- gartner...

    2024

  21. [29]

    napari: a multi-dimensional image viewer for python, 2019

    napari contributors. napari: a multi-dimensional image viewer for python, 2019. 2

    2019

  22. [30]

    Topology-aware loss for aorta and great vessel segmentation in computed tomography im- ages, 2024

    Seher Ozcelik, Sinan Unver, Ilke Ali Gurses, Rustu Turkay, and Cigdem Gunduz-Demir. Topology-aware loss for aorta and great vessel segmentation in computed tomography im- ages, 2024. 4

    2024

  23. [31]

    The axon initial segment and the main- tenance of neuronal polarity

    Matthew Rasband. The axon initial segment and the main- tenance of neuronal polarity. Nature reviews. Neuroscience, 11:552–62, 2010. 1

    2010

  24. [32]

    Imaging of the axon initial segment

    Jessica Di Re, Cihan Kayasandik, Gonzalo Botello-Lins, Demetrio Labate, and Fernanda Laezza. Imaging of the axon initial segment. Current Protocols in Neuroscience , 89(1): e78, 2019. 2

    2019

  25. [33]

    U-net: Convolutional networks for biomedical image segmentation,

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation,

  26. [34]

    Fiji: an open-source platform for biological- image analysis

    Johannes Schindelin, Ignacio Arganda-Carreras, Erwin Frise, Verena Kaynig, Mark Longair, Tobias Pietzsch, Stephan Preibisch, Curtis Rueden, Stephan Saalfeld, Ben- jamin Schmid, Jean-Yves Tinevez, Daniel James White, V olker Hartenstein, Kevin Eliceiri, Pavel Tomancak, and Al- ...

    2012

  27. [35]

    GitHub repos- itory for clDice: Topology-aware Loss Function for Tubu- lar Structure Segmentation

    Suprosanna Shit, Johannes C Paetzold, Anjany Sekuboyina, Ivan Ezhov, Alexander Unger, Andrey Zhylka, Josien PW Pluim, Ulrich Bauer, and Bjoern H Menze. GitHub repos- itory for clDice: Topology-aware Loss Function for Tubu- lar Structure Segmentation . https://github.com/ jocpa...

    2020

  28. [36]

    cldice-a novel topology-preserving loss function for tubular structure seg- mentation

    Suprosanna Shit, Johannes C Paetzold, Anjany Sekuboyina, Ivan Ezhov, Alexander Unger, Andrey Zhylka, Josien PW Pluim, Ulrich Bauer, and Bjoern H Menze. cldice-a novel topology-preserving loss function for tubular structure seg- mentation. In Proceedings of the IEEE/CVF Confere...

    2021

  29. [37]

    Paetzold, and Ul- rich Bauer

    Nico Stucki, Vincent B ¨urgin, Johannes C. Paetzold, and Ul- rich Bauer. Efficient betti matching enables topology-aware 3d segmentation via persistent homology, 2024. 2

    2024

  30. [38]

    Live imaging of excitable axonal microdomains in ankyrin-g-gfp mice

    Christian Thome, Jan Maximilian Janssen, Seda Karabulut, Claudio Acuna, Elisa D’Este, Stella J Soyka, Konrad Baum, Michael Bock, Nadja Lehmann, Johannes Roos, Niko- las A Stevens, Masashi Hasegawa, Dan A Ganea, Chlo ´e M Benoit, Jan Gr ¨undemann, Lia Min, Kalynn M Bird, Chris-...

    2024

  31. [39]

    Protein-retention expansion microscopy of cells and tissues labeled using standard fluorescent proteins and antibodies

    Paul Tillberg, Fei Chen, Kiryl Piatkevich, Yongxin Zhao, Chih-Chieh Yu, Brian English, Linyi Gao, Anthony Mar- torell, Ho-Jun Suk, Fumiaki Yoshida, Ellen DeGennaro, Douglas Roossien, Guanyu Gong, Uthpala Seneviratne, Steven Tannenbaum, Robert Desimone, Dawen Cai, and Ed- ward ...

    2016

  32. [40]

    Efficient com- putation of persistent homology for cubical data

    Hubert Wagner, Chao Chen, and Erald Vuc ¸ini. Efficient com- putation of persistent homology for cubical data. 2012. 2

    2012

  33. [41]

    App2: Automatic tracing of 3d neuron morphology based on hierarchical pruning of gray-weighted image distance-trees

    Hang Xiao and Hanchuan Peng. App2: Automatic tracing of 3d neuron morphology based on hierarchical pruning of gray-weighted image distance-trees. Bioinformatics (Oxford, England), 29, 2013. 2

    2013

  34. [42]

    Cp-loss: Connectivity-preserving loss for road curb detec- tion in autonomous driving with aerial images

    Zhenhua Xu, Yuxiang Sun, Lujia Wang, and Ming Liu. Cp-loss: Connectivity-preserving loss for road curb detec- tion in autonomous driving with aerial images. CoRR, abs/2107.11920, 2021. 2

    2021 arXiv

  35. [43]

    Angstmann, Yuhuang Wu, Holly Stefen, Esmeralda Pari ´c, Thomas Fath, and Paul M.G

    Zhuang Xu, Christopher N. Angstmann, Yuhuang Wu, Holly Stefen, Esmeralda Pari ´c, Thomas Fath, and Paul M.G. Curmi. Location of the axon initial segment assembly can be predicted from neuronal shape. iScience, 27(3):109264,

  36. [44]

    Yermakov, Domenica E

    Leonid M. Yermakov, Domenica E. Drouet, Ryan B. Griggs, Khalid M. Elased, and Keiichiro Susuki. Type 2 diabetes leads to axon initial segment shortening in db/db mice.Fron- tiers in Cellular Neuroscience, V olume 12 - 2018, 2018. 5

    2018

  37. [45]

    Axon initial seg- ments: Diverse and dynamic neuronal compartments

    Takeshi Yoshimura and Matthew Rasband. Axon initial seg- ments: Diverse and dynamic neuronal compartments. Cur- rent opinion in neurobiology, 27C:96–102, 2014. 1

    2014

  38. [46]

    Topological deep learning: a review of an emerging paradigm

    Ali Zia, Abdelwahed Khamis, James Nichols, Usman Bashir Tayab, Zeeshan Hayder, Vivien Rolland, Eric Stone, and Lars Petersson. Topological deep learning: a review of an emerging paradigm. Artificial Intelligence Review , 57(4),

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.