Pith. sign in

REVIEW 5 major objections 5 minor 4 references

TubuleTracker: a high-fidelity shareware software to quantify angiogenesis architecture and maturity

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a new shareware tool, tubuleTracker, analyzes in vitro angiogenesis images about ten times faster than ImageJ and with counts closer to manual reviewers, and that a new vessel-circularity metric tracks network…

desk verdict The speed claim is solid and the tool is a good idea, but the accuracy claim is unsupported by the statistics and the manual gold standard is unvalidated. read the letter →

arxiv 2507.02024 v1 pith:SBYBLM7C submitted 2025-07-02 q-bio.QM cs.CVq-bio.CB

Danish Mahmood , Stephanie Buczkowski , Sahaj Shah , Autumn Anthony , Rohini Desetty , Carlo R Bartoli This is my paper
classification q-bio.QMcs.CVq-bio.CB
keywords angiogenesisendothelialcellculturetubuleTrackerImageJcomparisonanalysissoftwarevesselcircularityskeletonizationHUVECtubeformation
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

tubuleTracker is new shareware software that analyzes phase-contrast images of cultured endothelial cells to count tubules, measure length, nodes, area, and a new 'circularity' metric. The paper reports that on 54 images it processed each image in about 6 seconds, roughly ten times faster than ImageJ and far faster than the 8 minutes of manual tracing, and that its counts of tubules, lengths, and nodes landed closer to manual readings than ImageJ's did. The paper also claims that circularity and tubule area track human ratings of network maturity more consistently than the traditional metrics do. If these claims hold, automated analysis could replace slow, subjective manual scoring in routine angiogenesis assays.

What carries the argument

The core object is the image-processing pipeline: non-local means denoising, adaptive thresholding tuned by peaking the lowest-frequency FFT amplitude, Gaussian low-pass filtering in the frequency domain, Otsu thresholding, skeletonization, and graph extraction of main nodes, terminal nodes, and edges. The second new object is the circularity metric, defined for each image as the inverse of the difference between the area of the best-fit ellipse of non-populated regions and the actual non-populated area; the paper uses it as a maturity index.

What would settle it

Reanalyze the same 54 images with a fourth independent lab's manual counts and compute inter-rater agreement; if reviewer disagreement is comparable to or larger than the reported differences between methods, the accuracy ranking collapses. Alternatively, an independent dataset where maturity is fixed by a molecular marker such as CD31 or VE-cadherin junction staining would settle whether circularity tracks maturity or merely correlates with the raters' subjective scale.

Watch

Extended reading notes

Core claim

The central claim is that tubuleTracker provides faster, more accurate quantitation of in vitro angiogenesis than the commonly used ImageJ Angiogenesis Analyzer and than manual visual analysis. On 54 phase-contrast images of HUVEC networks on Geltrex, tubuleTracker took 6 ± 2 s per image versus 58 ± 4 s for ImageJ and about 8 min for trained reviewers, and its tubule count (92 ± SD) fell between the manual count (168 ± SD) and ImageJ's count (433 ± SD), while node count (77 ± SD) was not statistically different from manual (69 ± SD), whereas ImageJ's (106 ± SD) was. For maturity, vessel circularity and tubule area increased monotonically with qualitative maturity ratings from 1 to 5, whereas tubule count, length, and node count did not, so the paper argues circularity is a better architectural marker of network maturity.

Load-bearing premise

The paper treats the average of three manual reviewers' counts as the true value, but it does not report how much the reviewers agreed with each other or show that manual counts correspond to real biological tubules; if manual analysis is not a trustworthy reference, the 'more accurate than ImageJ' claim has no anchor.

Editorial extensions

If this is right

  • At 6 seconds per image, tubuleTracker makes whole-experiment batch analysis practical, which the paper notes is easy because the Python script can be called on directories of images.
  • If circularity and tubule area really track maturity ratings, researchers can substitute a quantitative output for the 1-5 qualitative scale.
  • Because tubuleTracker treats wide monolayers of undifferentiated cells as single tubules, its counts may be closer to what manual reviewers intend to mark than ImageJ's over-fragmented skeletons.
  • Shareware availability means other labs can test the algorithm on their own images without paying per-sample service fees charged by commercial alternatives such as Wimasis or AngioSys.

Reading between the lines

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

  • If manual analysis is the reference, the accuracy claim depends on inter-reviewer reliability; the paper does not report agreement among its three reviewers, so a natural next validation is to quantify that agreement.
  • Circularity, as defined here on vacant mesh regions, could be tested as a predictor of functional maturity markers such as lumen formation or pericyte coverage; the paper does not make that link.
  • The FFT-based threshold-tuning step suggests a general approach to skeletonization of other membrane or vascular networks, but its robustness across imaging modalities such as fluorescence, confocal, or brightfield remains untested.
  • Because the software is shareware, an independent replication on a public image set with pre-registered manual annotations would convert the relative 'closer to manual' claim into an absolute accuracy benchmark.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The manuscript presents tubuleTracker, a Python-based software tool for quantifying endothelial cell network architecture (tubule count, length, node count, area, and a proposed circularity metric) from phase-contrast images of in vitro HUVEC cultures. The authors compare tubuleTracker against manual analysis by three reviewers and against the ImageJ Angiogenesis Analyzer on 54 images, and they analyze how tubuleTracker metrics vary with subjective 1-5 'angiogenesis maturity' scores. The paper's central claims are that tubuleTracker is significantly faster and more accurate than ImageJ and manual analysis, and that vessel circularity effectively characterizes angiogenesis maturity.

Significance. The speed advantage of tubuleTracker (reported 6±2 s/image vs 58±4 s/image for ImageJ and ~8 min/image for manual analysis) is directly supported by the timing data and would be a practical benefit for high-throughput screening. The paper also offers a useful open-source/shareware contribution and a transparent algorithm description. However, the central accuracy claim rests entirely on an unvalidated manual reference standard, and the maturity claim is a post hoc, in-sample analysis. As presented, the manuscript does not substantiate its two headline conclusions; the significance of the work would be much higher if the software were validated against synthetic ground truth or an independent, externally verified data set.

major comments (5)
  1. [§2.2, §4.1] Manual analysis by three reviewers is treated as the gold standard for accuracy, but no inter-reviewer reliability is reported and the manual counts are never validated against any ground truth. The ANOVA p-values in §2.6 only establish that the three methods differ; they do not establish that tubuleTracker is closer to the manual reference. This is load-bearing for the Abstract and §4.1 conclusions that tubuleTracker is 'more accurate' than ImageJ.
  2. [§2.4] The binarization parameters (OpenCV adaptive-threshold c-value, Gaussian low-pass filter kernel radius k, and small-branch removal threshold) are tuned per image to maximize an FFT smoothness heuristic, not to match biological tubule structure. No held-out images are used to assess generalization, and no sensitivity analysis is reported. The claimed accuracy advantage over ImageJ is therefore not demonstrated for new images.
  3. [§4.1] The sentence 'the difference in tubule count reported by tubuleTracker and manual analysis was significantly higher than that reported by ImageJ and manual analysis' directly contradicts the data in §3.2 (tubuleTracker 92 vs manual 168; ImageJ 433) and the stated conclusion that tubuleTracker is more accurate. This sentence must be corrected, and the accuracy comparison should be reported as paired absolute differences with confidence intervals.
  4. [§2.5, §3.3] The claim that vessel circularity is 'especially effective' in capturing angiogenesis maturity is based on in-sample ANOVA across subjective maturity scores on the same 54 images used to define the metric. No correlation coefficient or effect size is reported, no comparison to alternative maturity metrics is provided, and the metric definition (inverse of the difference between the best-fit ellipse area and the actual non-populated area) is not justified. The conclusion is post hoc rather than demonstrated.
  5. [§2.6, §3.2] The statistical reporting is insufficient for the strength of the claims. Values are given as 'mean±SD' but no standard deviations are actually reported, and the ANOVA/Tukey results are presented without effect sizes, confidence intervals, or correction for multiple comparisons. The reader cannot assess the magnitude of the reported differences between methods or maturity groups.
minor comments (5)
  1. [Abstract, §3.2] The notation '168±SD' appears without actual standard deviation values, making the numerical results impossible to interpret; please report the measured SDs.
  2. [Throughout] Terminology is inconsistent: the software is called 'tubuleTracker', 'Tubule Tracker', and 'tubuletracker' in different places; unify the name.
  3. [§2.5, Figure 3 legend] The metric is variously called 'network circularity', 'vessel circularity', and 'circularity of meshes'; define one term and specify its units and range.
  4. [§2.4, Figure 1] The FFT-based selection of the c-value is described only qualitatively ('until the amplitude ... peaked'); provide the precise criterion or pseudocode so that the algorithm is reproducible.
  5. [§2.2, §3.3] The term 'gold-standard manual classification' in the §3.3 heading is not justified because the manual classification is not validated; consider renaming it to 'manual classification' and acknowledging this limitation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the software metrics are independently defined from the manual reference and maturity scores, so no claim reduces by construction to its inputs.

full rationale

tubuleTracker's central claims are empirical comparisons, not derivations. The accuracy comparison (tubule count, length, node count) uses manual analysis as a reference standard, but the software's outputs are produced by an independent image-processing pipeline (denoising, adaptive thresholding, FFT filtering, skeletonization); they are not fitted to the manual counts. The maturity analysis compares software-computed metrics, including the novel 'circularity' metric, against human 1-5 maturity ratings. The circularity metric is defined geometrically from the best-fit ellipse of non-populated regions (Section 2.5), while the maturity score is a qualitative human judgment based on narrative criteria (Section 2.2). Although the maturity criteria and the circularity metric are both related to the amount of cell-free area, they are not mathematically identical, so the observed significant ANOVA is an empirical association rather than a tautology. The FFT-based thresholding heuristic (Section 2.4) is an image-processing parameter selection, not a parameter fitted to the outcome variables or to the manual gold standard. There are no load-bearing self-citations. The lack of inter-reviewer reliability for the manual gold standard and the in-sample nature of the maturity validation are important correctness and generalizability concerns, but they are not circularity in the sense of a prediction being equivalent to its inputs by construction.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claims rely on per-image parameter fitting, a subjective manual gold standard, and a new metric defined and evaluated on the same 54 images.

free parameters (3)
  • OpenCV adaptive thresholding c-value = adjusted per image until FFT lowest-frequency amplitude peaks (exact value not reported)
    In Section 2.4, the c-value is increased per image to maximize smoothness of the binary; this is a per-image fit that influences all downstream metrics.
  • Gaussian low-pass filter kernel radius k = selected such that the amplitude on the kernel perimeter equals k/3; value not reported
    Section 2.4 defines this rule, which is ambiguous and not independently justified; it determines the final binary and thus the skeleton metrics.
  • Small branch removal threshold = not specified
    Section 2.4 mentions removing small terminating node artifacts (pseudo-nodes) but gives no size threshold; the node count depends on this choice.
assumptions (3)
  • domain assumption Manual counts by three reviewers are an unbiased gold standard
    Section 2.2 states manual analysis served as the control for comparison; no inter-reviewer reliability is reported.
  • ad hoc to paper The FFT amplitude heuristic selects biologically accurate binary images
    Section 2.4 uses a peak in the lowest-frequency amplitude as the criterion for choosing c-value; no evidence links this criterion to tubule identity.
  • ad hoc to paper The new circularity definition captures network maturity
    Section 2.5 defines circularity as the inverse of the difference between ellipse area and non-populated area; this unbounded, non-standard metric is correlated with maturity only on the same dataset.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TubuleTracker: a high-fidelity shareware software to quantify angiogenesis architecture and maturity." pith.science (2026). https://pith.science/paper/SBYBLM7C

@misc{pith2026250702024,
  author       = {Pith},
  title        = {Pith review of: TubuleTracker: a high-fidelity shareware software to quantify angiogenesis architecture and maturity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SBYBLM7C}},
  note         = {Machine review of arXiv:2507.02024}
}
read the original abstract

Background: In vitro endothelial cell culture is widely used to study angiogenesis. Histomicrographic images of cell networks are often analyzed manually, a process that is time-consuming and subjective. Automated tools like ImageJ (NIH) can assist, but are often slow and inaccurate. Additionally, as endothelial networks grow more complex, traditional architectural metrics may not fully reflect network maturity. To address these limitations, we developed tubuleTracker, a software tool that quantifies endothelial network architecture and maturity rapidly and objectively. Methods: Human umbilical vein endothelial cells were cultured in an extracellular matrix, and 54 images were acquired using phase contrast microscopy. Each image was analyzed manually by three independent reviewers, and by both ImageJ and tubuleTracker. Key metrics included tubule count, total length, node count, tubule area, and vessel circularity. In parallel, trained scientists rated each image for angiogenesis maturity on a 1-5 scale (1 = most mature). Results: Analysis time per image differed significantly: manual (8 min), ImageJ (58+/-4 s), and tubuleTracker (6+/-2 s) (p<0.0001). Significant differences were also found in tubule count (manual 168+/-SD, tubuleTracker 92+/-SD, ImageJ 433+/-SD), length, and node count (all p<0.0001). tubuleTracker's metrics varied significantly across angiogenesis maturity scores, including tubule count, length, node count, area, and circularity (all p<0.0001). Conclusions: tubuleTracker was faster and more consistent than both manual and ImageJ-based analysis. Vessel circularity proved especially effective in capturing angiogenesis maturity. tubuleTracker is available as free shareware for the biomedical research community.

Figures

Figures reproduced from arXiv: 2507.02024 by the authors.

Figure 1
Figure 1. A) raw image B) adaptive thresholding with c-value 1 C) adaptive thresholding with c-value 5 D) magnitude of lowest frequency of FFT with increasing c-values. Red line indicated c-value used for downstream processing E) FFT Magnitude spectrum of adaptive thresholding input with c-value 4 F) adaptive thresholding input with c-value 4 G)Gaussian Kernelled FFT Magnitude spectrum of adaptive thresholding input with c-va… view at source ↗
Figure 2
Figure 2. Image processing time of manual analysis (red line), ImageJ, and tubuleTracker [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages

  1. [2]

    Endothelial Tube Formation Assay

    Staton, C. A., Lewis, C. & Bicknell, R. Angiogenesis Assays: A Critical Appraisal of Current Techniques. Angiogenesis Assays: A Critical Appraisal of Current Techniques (2007). doi:10.1002/9780470029350. [4] Montoya-Zegarra, J. A. et al. AutoTube: a novel software for the automated morphometric analysis of vascular networks in tissues. Angiogenesis 22, 22...

    work page doi:10.1002/9780470029350 2007

  2. [3]

    on" pixels while the background becomes

    Results 3.1 tubuleTracker is faster at processing images than imageJ and manual visual analysis Figure 2 shows that the mean processing time of tubuleTracker for each image was 6±2 s/image. The mean processing time for ImageJ for each image was 58±4 s/image. The mean time taken by trained investigators for manual analysis of the images was 8 min/image. Tu...

    work page 2007

  3. [16]

    Endothelial Tube Formation Assay

    Tsuji-Tamura, K. & Ogawa, M. FOXO1 promotes endothelial cell elongation and angiogenesis by up-regulating the phosphorylation of myosin light chain 2. Angiogenesis 1, 1–23 (2023). [17] Huuskes, B. M., DeBuque, R. J., Kerr, P . G., Samuel, C. S. & Ricardo, S. D. The use of live cell imaging and automated image analysis to assist with determining optimal pa...

    work page 2023

  4. [28]

    Arnaoutova, I., George, J., Kleinman, H. K. & Benton, G. The endothelial cell tube formation assay on basement membrane turns 20: state of the science and the art. Angiogenesis 12, 267–274 (2009). [29] Udiger, D. R. et al. Cell-Based Strain Remodeling of a Nonfibrous Matrix as an Organizing Principle for Vasculogenesis. CellReports 32, 108015 (2020). [30] ...

    work page 2009

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.