Persistent Homology as a Morphological Signature of Fibrin Networks

arXiv:2605.22145 · q-bio.OT · Submitted 2026-05-21 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Genomics Radio. Generated commentary on the latest computational biology and genomics papers.

Ines: I'm Ines.

Marcus: And I'm Marcus, here with Yuki.

Ines: We're looking at the paper: "Persistent Homology as a Morphological Signature of Fibrin Networks".

Paper summary: Ines: So, to wrap up this paper on "Persistent Homology as a Morphological Signature of Fibrin Networks," the authors are showing that persistent homology is a useful tool for understanding fibrin networks in a clinical setting <ref:2605.22145#pg1>.

Marcus: This work is considered a proof of concept that persistent homology is useful for understanding fibrin networks in a clinical setting, but they’re being careful about the limitations because of the small data set and lack of a healthy baseline to compare with means <ref:2605.22145#pg1>.

Yuki: It’s an important step towards a better understanding of the significance of fibrin network topology and its possible clinical significance in vivo >

Ines: The paper sets the stage for future research, suggesting that the goal should be to investigate if a reference range could be calculated by studying fibrin networks in plasma from healthy volunteers >

Marcus: It’s about developing a standardized analysis framework, which is what they are aiming for with this initial investigation into clot structure <ref:2605.22145#pg1>.

Conclusion: Ines: So, this paper is about using persistent homology to look at how fibrin networks change over time in patients after esophageal cancer surgery <ref:2605.22145#pg1>.

Marcus: They found that for the fibrin network structure, there wasn't really a big difference between the different treatment groups they compared <ref:2605.22145#pg9>.

Yuki: Which means that based on this analysis, the way these clots formed didn't really seem to depend on those specific anticoagulant treatments <ref:2605.22145#pg4>.

Ines: Exactly. The authors are pointing out that their method of looking at the topology—the shape of the network—doesn't show any significant change during that recovery period <ref:2605.22145#pg10>.

Marcus: They also tested how they selected the data, and it seems like their center-crop method is pretty robust, meaning those results aren't just an artifact of how they started looking at the images <ref:2605.22145#pg10>.

Yuki: But as you know, this is a small study with no healthy control group to compare against, so they’re being cautious about what they can actually claim about these findings <ref:2605.22145#pg1>

Ines: That's the main limitation. The authors are suggesting that next steps should involve looking at fibrin networks in plasma from healthy volunteers to see if they can establish some sort of baseline for comparison <ref:2605.22145#pg1>.

Marcus: So, the main point is that persistent homology is a tool we can use to characterize these physical structures, even if this specific study didn't find a major difference in the outcome they were testing <ref:2605.22145#pg9>.

Mathematical Data Science Centre, Australian National University · Department of Mathematical Sciences, Aalborg University · School of Science and Technology, University of New England · College of Science and Engineering, James Cook University · Interdisciplinary Center for Scientific Computing, Heidelberg University · Department of General and Surgical Intensive Care Medicine, Medical University of Innsbruck · Department of Clinical Medicine, Aarhus University · Department of Clinical Biocemistry, Aarhus University Hospital · ¡Institute of Science and Technology Austria · School of Mathematics and Statistics, University of Sydney

q-bio.OT

Submitted: 2026-05-21

Updated: 2026-11-18

Comments: 12 pages, 5 figures

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 77/100

The gist: Persistent Homology as a Morphological Signature of Fibrin Networks is an investigation into applying topological data analysis to high-resolution confocal microscopy images of fibrin networks from…

Key concepts

Persistent Homology
A mathematical tool used in topological data analysis to find structural features that persist across different scales of data. It helps characterize the underlying shape and connectivity of complex structures like fibrin networks by tracking how features emerge and survive as you change the resolution or scale at which you look.
Fibrin Networks
These are three-dimensional structures formed by fibrin protein strands, which act as a scaffold for blood coagulation. They are crucial for hemostasis and thrombosis, and their organization is believed to determine how stable a blood clot will be.
Topological Data Analysis (TDA)
A novel approach that uses persistent homology to capture structural features in data that remain consistent even when the data is transformed or viewed at different resolutions. It allows researchers to analyze the shape and connectivity of complex systems, such as biological networks, by focusing on their underlying topological properties.

Terminology

Summary

Persistent Homology as a Morphological Signature of Fibrin Networks is an investigation into applying topological data analysis to high-resolution confocal microscopy images of fibrin networks from patients undergoing oesophageal cancer surgery to understand clot structure and its relationship with anticoagulant treatment. p. 1

Where the paper says it

Abstract We present an investigation of the applicability of topological data analysis (TDA) to the study of high-resolution confocal microscopy images of fibrin network structures from patients with oesophageal cancer undergoing intended curative surgery. Investigation of clot structure brings new knowledge about blood coagulation, risk of bleeding, and thrombosis in this group of patients.

Overall, we found no significant differences in fibrin network topology across the perioperative period, and no consistent differences in network structure between the standard and intervention groups <ref:2605. p. 1
Where the paper says it

Each z-stack was cropped to a centre region for analysis, the validity of which is assessed in detail. Overall, we found no significant differences in fibrin network topology across the perioperative period, and no consistent differences in network structure between the standard and intervention groups.

22145#pg14>.

Background on Fibrin Networks

Fibrin protein networks play a crucial role in hemostasis and thrombosis by providing the structural framework for blood coagulation <ref:2605.22145#pg4>. Fibrin strands are formed when fibrinogen is cleaved by thrombin <ref:2605. p. 1

Where the paper says it

Introduction Fibrin protein networks play a crucial role in hemostasis and thrombosis by providing the structural framework for blood coagulation. Fibrin strands are formed when fibrinogen is cleaved by thrombin.

22145#pg4>. The resulting fibrin monomers polymerise, adhering to each other, and form a three-dimensional network that stabilises the platelets that make up the otherwise fragile, primary clot <ref:2605. p. 1
Where the paper says it

Fibrin strands are formed when fibrinogen is cleaved by thrombin. The resulting fibrin monomers polymerise, adhering to each other, and form a three-dimensional network that stabilises the platelets that make up the otherwise fragile, primary clot.

22145#pg4>. The organisation and architecture of fibrin networks are believed to influence clot stability <ref:2605. p. 1
Where the paper says it

The fibrin network is eventually cleaved by plasmin and dissolved; a process known as fibrinolysis [17]. The organisation and architecture of fibrin networks are believed to influence clot stability.

22145#pg4>.

Methodology for Topological Analysis

Topological Data Analysis (TDA) offers a novel approach for characterising fibrin architecture by capturing structural features that persist across multiple spatial scales <ref:2605.22145#pg4>. The study uses persistent homology to analyse images of fibrin networks obtained from human blood samples <ref:2605. p. 2

Where the paper says it

The present study aims to further investigate the applicability of TDA to the analysis of fibrin networks. In this study, we use persistent homology to analyse images of fibrin networks obtained from human blood samples.

22145#pg4>.

The methodology involves several steps for data processing and analysis:

  1. Image pre-processing includes computing the center of the data by finding the median of the cumulative intensity distribution along each spatial axis, and then cropping the x and y axes to obtain a 400 × 400 µm2 window <ref:2605. p. 3

    Where the paper says it

    5 µm Z-steps. Image pre-processing For our 3D pipeline, we compute the center of the data by finding the median of the cumulative intensity distribution along each spatial axis, and then crop the x and y axes to obtain a 400 × 400 µm2 window.

    22145#pg5>.

  2. The z-stack is smoothed and downsampled along the axial direction by averaging neighboring slices with a 3-slice box filter and subsampling with stride two <ref:2605. p. 3

    Where the paper says it

    Image pre-processing For our 3D pipeline, we compute the center of the data by finding the median of the cumulative intensity distribution along each spatial axis, and then crop the x and y axes to obtain a 400 × 400 µm2 window. The z-stack is then smoothed and downsampled along the axial direction by averaging neighboring slices with a 3-slice box filter and subsampling with stride two.

    22145#pg5>.

  3. A mild Gaussian filter (σ = 1) is applied in the x and y dimensions only <ref:2605. p. 3

    Where the paper says it

    The z-stack is then smoothed and downsampled along the axial direction by averaging neighboring slices with a 3-slice box filter and subsampling with stride two. A mild Gaussian filter (σ = 1) is applied in the x and y dimensions only.

    22145#pg5>.

  4. Intensity values are normalised to the [0, 1] range by clipping to the 1st and 99th percentiles <ref:2605. p. 3

    Where the paper says it

    A mild Gaussian filter (σ = 1) is applied in the x and y dimensions only. Intensity values are normalised to the [0, 1] range by clipping to the 1st and 99th percentiles.

    22145#pg5>.

The images are then represented as cubical complexes, where each pixel is treated as a square, and a greyscale image defines a lowerstar filtration of X <ref:2605. p. 3

Where the paper says it

In particular, we can think of each image pixel as a square, and obtain a twodimensional cubical complex X by adding in all (possibly shared) edges and vertices. A greyscale image, defined by an intensity function on the pixels, describes a lowerstar filtration [11] of X whereby, for each value t of the intensity parameter, Xt is the smallest two-dimensional subcomplex of X that includes all pixels (squares) with greyscale intensity less than or equal to t.

22145#pg6>. From this filtered complex, persistence diagrams are obtained using the GUDHI library <ref:2605. p. 4
Where the paper says it

From this filtered complex, we obtain a persistence diagram, which we use as a representation of the structure of the corresponding fibrin network. The GUDHI library [9] was used to create the cubical complexes from the images and to compute the persistence diagrams from these.

22145#pg6>.

Statistical Comparison of Treatment Groups

The study compares the effects of two different lengths of post-surgical anticoagulant regimes on fibrin network formation <ref:2605.22145#pg4>. This comparison was performed using permutational multivariate analysis of Variance (PERMANOVA) based on 2-Wasserstein distances <ref:2605. p. 4

Where the paper says it

We use the 2-Wasserstein distance [4, 14] on persistence diagrams as a dissimilarity score between fibrin networks. Statistical Analysis In comparing different collections of persistence diagrams, we used permutational multivariate analysis of Variance (PERMANOVA) [2] based on 2-Wasserstein distances.

22145#pg8>.

The results showed no evidence of difference in composition between the two treatment groups, with an observed pseudo-F statistic of 0. p. 5

Where the paper says it

The exact p-value was 0. 8651, providing no evidence of difference in composition between the two treatment groups.

3474 and an exact p-value of 0. p. 5
Where the paper says it

3474, indicating that between-group variation was less than within-group variation. The exact p-value was 0.

8651 <ref:2605. p. 0
Where the paper says it

dk; †These authors contributed equally to this work. 1 arXiv:2605.

22145#pg9>. Furthermore, clustering by sample time provided no evidence for an observable difference across the periods <ref:2605. p. 7
Where the paper says it

Taken together, these results provide no statistical evidence of a change in composition across the periods, through the small sample size limits the power of these tests. The clustering of the four different vectorizations of persistence diagrams also provide no evidence for an observable difference between sample times.

22145#pg10>.

Validation of Topological Signal Robustness

Robustness of cropping was assessed through three experiments to validate the center-crop procedure <ref:2605.22145#pg10>.

  1. The first experiment showed that topology changes gradually with spatial displacement from the center, with a Pearson correlation of r = 0. p. 8

    Where the paper says it

    The first established that topology changes gradually with spatial displacement from the center, the second tested whether the centre crop is representative of the image as a whole, and the third assessed whether crop-level variation is neglible relative to the inter-sample signal of interest. In the first experiment, we computed persistence diagrams for crops centered at positions shifted in increments of 20 pixels in both x and y directions, ranging from −100 to 100 pixels.

    483 (p < 0. p. 8
    Where the paper says it

    4006) at the largest shift of 141 pixels, with Pearson correlation of r = 0. 483 (p < 0.

    0001).

  2. The second experiment indicated that the centre crop is systematically more topologically distinct from random crops than random crops are from each other, with a significant t-test result (t = 5. p. 8

    Where the paper says it

    106, p < 0. p. 8

    Where the paper says it

    A onesample t-test against a null ratio of 1 was significant (t = 5. 106, p < 0.

    0001), indicating that the centre crop is systematically more topologically distinct from random crops than random crops are from each other.

    106, p < 0.0001).

  3. The third experiment confirmed that inter-sample distances were significantly larger than intra-sample variation, with a paired t-test showing inter-sample distances approximately twice the magnitude of intra-sample variation (t = −21. p. 8

    Where the paper says it

    0001), with inter-sample distances approximately twice the magnitude of intra-sample variation. These results validate the analytical approach, that the topological signal distinguishing individuals is substantially larger than the noise introduced by the choice of crop.

    820, p < 0. p. 8
    Where the paper says it

    A paired t-test confirmed that inter-sample distances were significantly larger (t = −21. 820, p < 0.

    0001).

Conclusion and Future Directions

This work is considered a proof of concept that persistent homology is a useful tool for understanding fibrin networks in a clinical setting. The authors note the small size of the data set and lack of a healthy baseline to compare with means, limiting significant conclusions from this study. Future research should investigate if a reference range could be calculated by studying fibrin networks in plasma from healthy volunteers. p. 1

Where the paper says it

The tightness and density of the networks might influence mechanical properties of the clot, as well as the efficacy and speed of fibrinolysis and therefore clot longevity [19]. Several methods have been developed for use in research to investigate fibrin networks, but quantifying the complexity of fibrin networks remains challenging due to their heterogeneous, three-dimensional nature [15].

The study is an important step towards a better understanding of the significance of fibrin network topology and its possible clinical significance in vivo. p. 2
Where the paper says it

In this study, we use persistent homology to analyse images of fibrin networks obtained from human blood samples. Furthermore, we compare the possible effects of two different lengths of post-surgical anticoagulant regimes on fibrin network formation.

Strengths and Limitations

Strengths include this being the first publication using confocal microscopy to capture fibrin network structures in cancer patients, allowing for a longitudinal analysis of clot architecture over time. The dataset enables a comparative assessment of how different treatment strategies influence fibrin network topology throughout the post-surgical period. p. 2

Where the paper says it

In this study, we use persistent homology to analyse images of fibrin networks obtained from human blood samples. Furthermore, we compare the possible effects of two different lengths of post-surgical anticoagulant regimes on fibrin network formation.

A limitation is the small size of the data set and a lack of a healthy baseline to compare with means, which prevents making significant conclusions from this study. The authors emphasize that this work should be undertaken in order to investigate if a reference range could be calculated, with the goal of developing a standardised analysis framework. p. 0
Where the paper says it

at; tuagyl@clin. dk; †These authors contributed equally to this work.

Acknowledgements and Funding

The authors thank Dr. Dietmar Fries for graciously hosting TG in his lab and sharing his expertise with confocal imaging. p. 3

Where the paper says it

Samples and imaging Blood samples were collected at four time points: the morning before surgery (baseline), immediately after surgery, one day post-surgery, and thirty days post-surgery. All blood samples were centrifuged for 25 minutes at 3000 g and stored at -80 degrees Celsius until analysis.

The work was funded by Aarhus University, Aarhus University Hospital, Oda and Hans Svenningsens’ Foundation and Danish Society for Thrombosis and Haemostasis (The Steen Husted Grant). p. 2
Where the paper says it

Adult patients with oesophageal cancer referred for intended curative surgery at Aarhus University Hospital, Aarhus, Denmark, were included between September 2021 and April 2024. Patients were randomly assigned to receive either 10 or 30 days of prophylactic anticoagulant treatment with the low molecular weight heparin Fragmin (Pfizer, USA) aimed at managing coagulation and thrombosis risk.

TB acknowledges support by an Australian Government Research Training Program (RTP) Scholarship, and the Danish Data Science Academy (DDSA) research visit Grant ID 2025-5524. p. 0
Where the paper says it

Persistent Homology as a Morphological Signature of Fibrin Networks Thomas Burnett1† , Theresa Reinhold 2† , Bea Bleile3 , Sophie Raynor4 , Freya Jensen5 , Martin Hermann6 , Tua Gyldenholm7,8, Yossi Bokor Bleile9,10* 1Mathematical Data Science Centre, Australian National University, Canberra, Australia. 2Department of Mathematical Sciences, Aalborg University, Aalborg Denmark.

The author has applied a CC BY public copyright license to any author-accepted manuscript version arising from this submission. p. 0
Where the paper says it

10School of Mathematics and Statistics, University of Sydney, Sydney, Australia. *Corresponding author(s).

References

The references cited in the paper are listed at the end of the document. The authors provided a list of 18 references, including works on TDA, persistence diagrams, and fibrin networks. p. 2

Where the paper says it

network structure in different clinical presentations, including patients treated with anticoagulation, remains poorly characterised. In [3], the authors used persistent homology on 3D greyscale images of fibrin networks in porcine blood samples to understand the effects of dilution on the persistent homology of the networks.

The most relevant references include those discussing persistence diagrams as morphological signatures of cells and Wasserstein stability for persistence diagrams. p. 4
Where the paper says it

K-means clustering is then used to partition the samples into k distinct groups, where k is determined with elbow plots. We use the 2-Wasserstein distance [4, 14] on persistence diagrams as a dissimilarity score between fibrin networks.

The paper also references studies on fibrin networks in clinical contexts. p. 1
Where the paper says it

However, there are few analyses that investigate fibrinolysis [15]. The geometry of fibrin networks may play an important role in the stability and durability of the formed blood clot, but few studies have attempted to investigate this [17].

The list of references is extensive, covering topics from mathematical data science to clinical medicine. p. 0
Where the paper says it

Persistent Homology as a Morphological Signature of Fibrin Networks Thomas Burnett1† , Theresa Reinhold 2† , Bea Bleile3 , Sophie Raynor4 , Freya Jensen5 , Martin Hermann6 , Tua Gyldenholm7,8, Yossi Bokor Bleile9,10* 1Mathematical Data Science Centre, Australian National University, Canberra, Australia. 2Department of Mathematical Sciences, Aalborg University, Aalborg Denmark.

The paper includes a reference to the TOP-RCT study protocol. p. 2
Where the paper says it

The four new contributions of this study are: 1) the application of persistent homology to a longitudinal study to compare the effects of two different lengths of postsurgical anticoagulant regimes on fibrin network formation, 2) a modified methodology which allows for the 3-dimensional structure of the fibrin network to be quantified, 3) an investigation of the homogeneity of the fibrin network within individual samples to assess the applicability of sampling methods, and 4) the introduction of additional existing methods within TDA to the study of fibrin. Materials and methods The samples used for the present study were collected as part of the Thromboprophylaxis in Oesophageal Cancer Patients Randomised Controlled Trial (TOP-RCT) study [10].

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Develop an AI system capable of quantifying fibrin network topology using persistent homology to assess clot stability and dysregulation in patients undergoing oesophageal cancer surgery. This system will use a modified methodology that allows for the 3D structure of fibrin networks to be quantified, as described in the paper.

  2. Create an AI system that can compare the effects of different lengths of post-surgical anticoagulant regimes on fibrin network formation by analyzing topological characteristics across time points. This leverages the application of persistent homology to a longitudinal study to identify differences in clot architecture that might reflect dysregulation of hemostasis.

  3. Implement an AI system for assessing sampling methods by investigating the homogeneity of fibrin networks within individual samples using TDA. This addresses the contribution of an investigation of the homogeneity of the fibrin network within individual samples to assess the applicability of sampling methods.

  4. Build an AI system that can detect topological differences between patient groups, even when considering noise from debris, by utilizing robust cropping procedures validated in three experiments. This incorporates the finding that the topological signal distinguishing individuals is substantially larger than the noise introduced by the choice of crop.

Abstract

We present an investigation of the applicability of topological data analysis (TDA) to the study of high-resolution confocal microscopy images of fibrin network structures from patients with oesophageal cancer undergoing intended curative surgery. Investigation of clot structure brings new knowledge about blood coagulation, risk of bleeding, and thrombosis in this group of patients. Images of fibrin network formation in the collected blood samples were captured by confocal microscopy and three-dimensional z-stacks were analysed. Each z-stack was cropped to a centre region for analysis, the validity of which is assessed in detail. Overall, we found no significant differences in fibrin network topology across the perioperative period, and no consistent differences in network structure between the standard and intervention groups.

Sources

Related papers