pith. sign in

arxiv: 2606.07368 · v1 · pith:5L65DKGOnew · submitted 2026-06-05 · 💻 cs.CV · cs.AI

Mitosis Detection in the Wild: Multi-Tumor and Context-Aware Generalization in the MIDOG 2025 Challenge

Marc Aubreville , Jonas Ammeling , Sweta Banerjee , Viktoria Weiss , Taryn A. Donovan , Robert Klopfleisch , Jiaqi Lv , Shan E Ahmed Raza
show 59 more authors
Rapha\"el Bourgade Thomas Walter Yasemin Topuz Song\"ul Varl{\i} Charles-Antoine Collins-Fekete Zhuoyan Shen Navya Sri Kelam Nitin Singhal Christian Marzahl Brian Napora Tengyou Xu Hongyan Gu Mario Vento Gennaro Percannella Norbert Ropiak Izabela Wasiak Jie Xiao Shaojun Liu Seungho Choe April Khademi Vidushi Walia Sujatha Kotte Andrew Broad Alex Wright Guillaume Balezo Esha Sadia Nasir Mostafa Jahanifar Yosuke Yamagishi Shouhei Hanaoka Mattia Sarno Francesco Tortorella Biwen Meng Jingxin Liu Sara Krauss Daniel Hieber Lavish Ramchandani Dev Kumar Das Mieko Ochi Yuan Bae Piotr Giedziun Mateusz Maniewski Vangala Govindakrishnan Saipradeep Naveen Sivadasan Leire Benito-Del-Valle Adrian Galdran Kaustubh Atey Sameer Anand Jha Adinath Dukre Imran Razzak Maxime W. Lafarge Viktor H. Koelzer Nils Porsche Nikolas Stathonikos Mitko Veta Dominik Hirling Zsanett Zs\'ofia Iv\'an Peter Horvath Katharina Breininger Christof A. Bertram
This is my paper

Pith reviewed 2026-06-27 22:11 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords mitosis detectiondomain generalizationcomputational pathologyMIDOG challengeatypical mitotic figureswhole-slide imaginghard negative regions
0
0 comments X

The pith

Mitosis detection models degrade sharply outside hotspots and across diverse tumor types in MIDOG 2025.

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

The paper presents the MIDOG 2025 challenge as a test of mitosis detection under broader biological and contextual variance than prior benchmarks. It evaluates models on 365 cases spanning 12 tumor types from multiple species and scanners, requiring detection not only in hotspots but also in random tissue regions and areas with many hard negatives, plus separate classification of atypical mitotic figures. Top submissions reached an F1 of 0.740 on detection and 0.908 balanced accuracy on atypical figures, yet most models showed large drops in challenging regions where false-positive rates tripled and results differed markedly by tumor type. Ensembling produced small average gains while test-time augmentation did not. The work positions this multi-context evaluation as a closer proxy for clinical whole-slide reliability.

Core claim

MIDOG 2025 establishes that current mitosis detectors remain unreliable in the wild: performance holds in conventional hotspots but declines substantially in random and hard-negative regions and varies across the 12 tumor types, with the curated multi-tumor, multi-scanner test set intended to expose these limitations more realistically than hotspot-only benchmarks.

What carries the argument

The multi-contextual evaluation protocol that mandates mitosis detection across hotspots, random tissue areas, and hard-negative regions together with atypical mitotic figure classification on a 365-case test set spanning 12 tumor types.

If this is right

  • False-positive rates triple in regions rich in hard negatives compared with hotspot evaluation.
  • Detection accuracy differs substantially across tumor types, exposing blind spots for rare or highly pleomorphic cases.
  • Simple ensembling raises average F1 by 1.5 points and balanced accuracy by 1.3 points; test-time augmentation yields no meaningful change.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Models may require targeted handling of pleomorphic mitotic figures to close the observed tumor-type gaps.
  • Clinical pipelines might still need per-tumor or per-scanner calibration even after training on this diverse set.
  • The absence of benefit from test-time augmentation suggests that certain standard robustness tricks transfer poorly to this histology task.

Load-bearing premise

The 365-case collection across 12 tumor types and multiple scanners captures enough of the biological and contextual variation that occurs in routine clinical whole-slide analysis.

What would settle it

A submitted model that maintains its hotspot F1 score without increase in false positives when tested on the challenging ROIs and shows no performance gaps across all 12 tumor types.

Figures

Figures reproduced from arXiv: 2606.07368 by Adinath Dukre, Adrian Galdran, Alex Wright, Andrew Broad, April Khademi, Biwen Meng, Brian Napora, Charles-Antoine Collins-Fekete, Christian Marzahl, Christof A. Bertram, Daniel Hieber, Dev Kumar Das, Dominik Hirling, Esha Sadia Nasir, Francesco Tortorella, Gennaro Percannella, Guillaume Balezo, Hongyan Gu, Imran Razzak, Izabela Wasiak, Jiaqi Lv, Jie Xiao, Jingxin Liu, Jonas Ammeling, Katharina Breininger, Kaustubh Atey, Lavish Ramchandani, Leire Benito-Del-Valle, Marc Aubreville, Mario Vento, Mateusz Maniewski, Mattia Sarno, Maxime W. Lafarge, Mieko Ochi, Mitko Veta, Mostafa Jahanifar, Naveen Sivadasan, Navya Sri Kelam, Nikolas Stathonikos, Nils Porsche, Nitin Singhal, Norbert Ropiak, Peter Horvath, Piotr Giedziun, Rapha\"el Bourgade, Robert Klopfleisch, Sameer Anand Jha, Sara Krauss, Seungho Choe, Shan E Ahmed Raza, Shaojun Liu, Shouhei Hanaoka, Song\"ul Varl{\i}, Sujatha Kotte, Sweta Banerjee, Taryn A. Donovan, Tengyou Xu, Thomas Walter, Vangala Govindakrishnan Saipradeep, Vidushi Walia, Viktor H. Koelzer, Viktoria Weiss, Yasemin Topuz, Yosuke Yamagishi, Yuan Bae, Zhuoyan Shen, Zsanett Zs\'ofia Iv\'an.

Figure 1
Figure 1. Figure 1: Domains of the test set of the MIDOG 2025 challenge. For each domain, both the abbreviation chosen as well as the full name is given. Shown are four random samples from within the hotspot regions of interest. to chromosome segregation errors [29, 11]. This asymmetric cell division can lead to daughter cells with an abnormal number of chromosomes (aneuploidy), a relevant mecha￾nism how tumor cells can accum… view at source ↗
Figure 2
Figure 2. Figure 2: Region types in track 1 of the MIDOG 2025 chal￾lenge. Left panel shows hotspot regions, middle panel shows random regions and right panel shows challenging regions. Green squares indicate PHH3-confirmed mitotic figures, yellow circles indicate false detections by our reference model. Peer review. Each submitted preprint was reviewed inde￾pendently by at least two expert reviewers, who assessed contribution… view at source ↗
Figure 3
Figure 3. Figure 3: Violin plot showing the distribution of mitotic figures (MFs) in the MIDOG 2025 test set for track 1, with tumor types being sorted by median hotspot MF count. a) The hotspot regions of interest (ROIs) were selected by a pathologist in the most mitotically active tumor region, whereas the random ROI b) regions were sampled from the tissue area of the whole slide image. The challenging regions (c), selected… view at source ↗
Figure 4
Figure 4. Figure 4: Class distribution for all tumor types of the track 2 test set. For the tumor type abbreviations, please consult [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Post-challenge analysis workflow. Model parameter count and inference time were established using modification of the submitted docker containers. Additionally, we ablated the containers from using test time augmentation and ensembling to investigate the effects of both methods. Moreover, we investigated the rank stability of the chal￾lenge, as influenced by the region type (overall, hotspot, ran￾dom, chal… view at source ↗
Figure 6
Figure 6. Figure 6: a) Rank and b) 𝐹1 scores for all participants across the area types of the track 1 test set. c) shows correlations (Pearson 𝑟) with 95% confidence intervals (CI) between the performance of all participants across the various area types (hots.=hotspot, chall.=challenging, rand.=random). The performance in hotspot and challenging areas had a non-significant correlation, all others were found to be highly sig… view at source ↗
Figure 7
Figure 7. Figure 7: Inference time vs. main metric for track 1 (a) and track 2 (b) of the MIDOG 2025 challenge. Inference time was determined on a Linux workstation with an NVIDIA RTX 4090. Bubble size indicates max. VRAM usage of container, averaged over 50 cases. Datasets. All approaches used the challenge’s training set [110]. The majority (12/20) also used the smaller breast cancer atypical mitosis dataset AMi-Br [20]. Du… view at source ↗
Figure 8
Figure 8. Figure 8: Ablation study for test-time augmentation (TTA) and ensembling. Shown are only participants that utilized either TTA or ensembling in track one (top) and track two (bottom). 4.5. Ablation studies Our component analysis of TTA or ensembling reveals a mixed benefit of both methods for both tracks ( [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
read the original abstract

Automated mitosis detection is a well-established task in computational pathology. While previous benchmarks focused on scanner-induced domain shift, clinical "real-world" application requires models to be robust across the vast variance to be expected in the histological landscape. The MItosis DOmain Generalization (MIDOG) 2025 challenge was designed to evaluate algorithmic performance across unprecedented biological and contextual diversity. We curated a test dataset of 365 cases, encompassing 12 distinct human, canine and feline tumor types, digitized across multiple scanning platforms. Moving beyond hand-selected hotspots, the challenge required detection also in random tissue areas (representative of the whole slide detection situation) and challenging areas (areas rich in hard negatives). In the second track, we introduced the classification of atypical mitotic figures (AMFs). There were 18 teams submitting to the detection track, with F1 scores ranging up to 0.740. In the AMF detection track, we had 21 submissions with balanced accuracy values up to 0.908. Our analysis reveals that while most models perform reliably in traditional hotspots, significant performance degradation occurs in challenging ROIs, where false positive rates tripled. Furthermore, performance varied significantly across the 12 tumor types, highlighting "blind spots" in current state-of-the-art architectures when encountering rare or highly pleomorphic malignancies. Moreover, we evaluated the effectiveness of ensembling and found a mean increases of 1.5 and 1.3 percentage points in F1 score and balanced accuracy, respectively. In contrast, TTA showed no relevant improvement. MIDOG 2025 demonstrates that "in the wild" mitosis detection remains a significant hurdle. The transition from hotspot-only evaluation to a multi-contextual framework provides a more realistic proxy for clinical reliability.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript reports results from the MIDOG 2025 challenge on mitosis detection. It describes curation of a 365-case test set spanning 12 tumor types (human, canine, feline) and multiple scanners, with evaluation in hotspots, random tissue areas, and hard-negative ROIs. A second track addresses atypical mitotic figure classification. Aggregated results from 18 detection-track and 21 AMF-track submissions show peak F1 of 0.740 and balanced accuracy of 0.908; performance degrades markedly in challenging ROIs (FP rates triple) and varies across tumor types. Ensembling yields modest gains (+1.5 pp F1, +1.3 pp balanced accuracy) while test-time augmentation does not. The central claim is that in-the-wild mitosis detection remains a significant hurdle and that the multi-contextual protocol supplies a more realistic clinical proxy.

Significance. If the test-set representativeness holds, the work is significant for shifting mitosis-detection benchmarks from scanner-only domain shift to biological and contextual diversity, exposing model blind spots on rare or pleomorphic tumors. The scale of participation, explicit comparison of ensembling versus TTA, and move beyond hotspot-only evaluation constitute concrete strengths that can guide future architecture design.

major comments (2)
  1. [Abstract] Abstract (dataset curation paragraph): The headline claim that performance degradation demonstrates a general clinical barrier rests on the 365-case set being representative of clinical whole-slide variance in mitotic density, pleomorphism, and hard-negative prevalence; however, no quantitative anchor (Wasserstein distance, KS tests, or prevalence statistics) against a broader clinical reference cohort is supplied.
  2. [Abstract] Abstract (results paragraph): The reported FP-rate tripling and tumor-type gaps are presented as evidence of generalization failure, yet the manuscript supplies neither per-tumor-type case counts nor per-ROI mitotic-density statistics, preventing assessment of whether the observed drops are driven by a few atypical cases or reflect systematic limitations.
minor comments (2)
  1. The manuscript would benefit from reporting mean and standard deviation of F1 and balanced-accuracy scores across all submissions rather than only the maximum values, to contextualize the range 0–0.740.
  2. Clarify whether the 365 cases include any overlap with prior MIDOG editions or whether all material is newly acquired for this challenge.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract (dataset curation paragraph): The headline claim that performance degradation demonstrates a general clinical barrier rests on the 365-case set being representative of clinical whole-slide variance in mitotic density, pleomorphism, and hard-negative prevalence; however, no quantitative anchor (Wasserstein distance, KS tests, or prevalence statistics) against a broader clinical reference cohort is supplied.

    Authors: The test set was curated to span 12 tumor types across species and multiple scanners to capture substantial biological and contextual diversity, based on expert selection rather than formal distributional matching to an external reference cohort. No such large-scale annotated reference cohort was available for computing Wasserstein distances or KS tests. We will revise the abstract to replace the phrasing 'general clinical barrier' with 'significant challenge under diverse conditions' and add a clarifying sentence on the diversity-driven curation approach. revision: partial

  2. Referee: [Abstract] Abstract (results paragraph): The reported FP-rate tripling and tumor-type gaps are presented as evidence of generalization failure, yet the manuscript supplies neither per-tumor-type case counts nor per-ROI mitotic-density statistics, preventing assessment of whether the observed drops are driven by a few atypical cases or reflect systematic limitations.

    Authors: The full manuscript (Section 3.1 and Table 1) reports the number of cases per tumor type. Per-ROI mitotic density statistics are not provided because ROIs were expert-selected into qualitative categories (hotspots, random tissue, hard-negative) without per-ROI density quantification. The performance degradation and tumor-type variation are observed consistently across the set rather than driven by outliers. We will add a short reference to the case distribution in the abstract if space allows. revision: partial

Circularity Check

0 steps flagged

Empirical challenge report with no derivations or self-referential predictions

full rationale

The paper is a challenge summary reporting F1 scores and balanced accuracies from 18-21 external team submissions on a fixed test set of 365 cases. No equations, fitted parameters, or predictions are derived within the manuscript; all quantitative results originate from participant algorithms evaluated on the provided data. The central claim (performance degradation in non-hotspot ROIs and across tumor types) is a direct empirical observation, not a reduction to any self-defined input or self-citation chain. No load-bearing steps match the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on the domain assumption that the selected 365 cases and ROIs represent clinical 'in the wild' variance; no free parameters, invented entities, or mathematical axioms are present.

axioms (1)
  • domain assumption The curated 365-case test set across 12 tumor types and multiple scanners captures the biological and contextual diversity of real-world histology slides.
    Invoked in the abstract description of dataset design and the claim that the framework is a realistic proxy.

pith-pipeline@v0.9.1-grok · 6206 in / 1167 out tokens · 17732 ms · 2026-06-27T22:11:04.938883+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

135 extracted references · 41 canonical work pages · 6 internal anchors

  1. [1]

    Ammeling, J., Aubreville, M., Banerjee, S., Bertram, C.A., Breininger, K., Hirling, D., Horvath, P., Stathonikos, N., Veta, M.,

  2. [2]

    URL:https: //doi.org/10.5281/zenodo.15077361, doi:10.5281/zenodo.15077361

    Mitosis domain generalization challenge 2025. URL:https: //doi.org/10.5281/zenodo.15077361, doi:10.5281/zenodo.15077361

    work page doi:10.5281/zenodo.15077361 2025

  3. [3]

    Atey, K., Jha, S.A., Bala, G., Sethi, A., 2026. Mix, Align, Distil: Reliable Cross-Domain Atypical Mitosis Classification, in: Aubre- ville,M.,Bertram,C.A.(Eds.),MitoticFigureDetectionandAtypia Classification in Whole Slide Images, Springer. pp. 137–143

    2026

  4. [4]

    A completely annotated whole slide imagedatasetofcaninebreastcancertoaidhumanbreastcancerre- search.Scientificdata7:417,1–10.doi:10.1038/s41597-020-00756-z

    Aubreville, M., Bertram, C.A., Donovan, T.A., Marzahl, C., Maier, A., Klopfleisch, R., 2020a. A completely annotated whole slide imagedatasetofcaninebreastcancertoaidhumanbreastcancerre- search.Scientificdata7:417,1–10.doi:10.1038/s41597-020-00756-z

    work page doi:10.1038/s41597-020-00756-z

  5. [5]

    Deep learning algorithms out-perform veterinary pathologists in detecting the mitotically most active tumor region

    Aubreville, M., Bertram, C.A., Marzahl, C., Gurtner, C., Dettwiler, M.,Schmidt,A.,Bartenschlager,F.,Merz,S.,Fragoso,M.,Kershaw, O., et al., 2020b. Deep learning algorithms out-perform veterinary pathologists in detecting the mitotically most active tumor region. ScientificReports10:16447,1–11.doi:10.1038/s41598-020-73246-2

    work page doi:10.1038/s41598-020-73246-2

  6. [6]

    Mitosis domain generalization in histopathology images—the MIDOG challenge

    Aubreville, M., Stathonikos, N., Bertram, C.A., Klopfleisch, R., Ter Hoeve, N., Ciompi, F., Wilm, F., Marzahl, C., Donovan, T.A., Maier, A., et al., 2023a. Mitosis domain generalization in histopathology images—the MIDOG challenge. Medical Image Analysis 84, 102699

  7. [7]

    Domaingeneralizationacrosstumortypes,laboratories, and species—insights from the 2022 edition of the mitosis domain generalization challenge

    Aubreville, M., Stathonikos, N., Donovan, T.A., Klopfleisch, R., Ammeling, J., Ganz, J., Wilm, F., Veta, M., Jabari, S., Eckstein, M., etal.,2024. Domaingeneralizationacrosstumortypes,laboratories, and species—insights from the 2022 edition of the mitosis domain generalization challenge. Medical Image Analysis 94, 103155

    2024

  8. [8]

    A comprehensive multi-domain dataset for mitotic figure detection

    Aubreville,M.,Wilm,F.,Stathonikos,N.,Breininger,K.,Donovan, T.A., Jabari, S., Veta, M., Ganz, J., Ammeling, J., Van Diest, P.J., et al., 2023b. A comprehensive multi-domain dataset for mitotic figure detection. Scientific data 10, 484

  9. [9]

    Balezo,G.,Bourgade,R.,Feki,H.,Monnier,L.,Blons,M.,Blondel, A., Decencière, E., Planas, A.P., Walter, T., 2026. Efficient Fine- Tuning of DINOv3 Pretrained on Natural Images for Atypical Mi- toticFigureClassification,in:Aubreville,M.,Bertram,C.A.(Eds.), Mitotic Figure Detection and Atypia Classification in Whole Slide Images, Springer. pp. 15–25

    2026

  10. [10]

    Mitosis Domain Generalization (MIDOG) Challenge 2025 Baselines, in: Aubreville, M., Bertram, C.A

    Banerjee, S., Ammeling, J., Weiss, V., Donovan, T.A., Klopfleisch, R., Bertram, C.A., Breininger, K., Aubreville, M., 2026a. Mitosis Domain Generalization (MIDOG) Challenge 2025 Baselines, in: Aubreville, M., Bertram, C.A. (Eds.), Mitosis Domain Generaliza- tion (MIDOG) Challenge 2025 Baselines, Springer. pp. 1–14

    2025

  11. [12]

    Bench- marking deep learning and vision foundation models for atyp- ical vs

    Banerjee, S., Weiss, V., Donovan, T.A., Fick, R.H., Conrad, T., Ammeling, J., Porsche, N., Klopfleisch, R., Kaltenecker, C.C., Breininger, K., Aubreville, M., Bertram, C.A., 2026c. Bench- marking deep learning and vision foundation models for atyp- ical vs. normal mitosis classification with cross-dataset evalua- tion. Machine Learning for Biomedical Imag...

    2026

  12. [13]

    org/10.59275/j.melba.2026-6c1g

    URL:https://melba-journal.org/2026:006, doi:https://doi. org/10.59275/j.melba.2026-6c1g

    work page doi:10.59275/j.melba.2026-6c1g 2026

  13. [14]

    Is Syn- thetic Image Augmentation Useful for Imbalanced Classification Problems? Case-Study on the MIDOG2025 Atypical Cell Detec- tion Competition, in: Aubreville, M., Bertram, C.A

    Benito-Del-Valle, L., Moreno-Sánchez, P.A., Eguskiza, I., Vitoria, I., Picón, A., López-Saratxaga, C., Galdran, A., 2026. Is Syn- thetic Image Augmentation Useful for Imbalanced Classification Problems? Case-Study on the MIDOG2025 Atypical Cell Detec- tion Competition, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classific...

    2026

  14. [15]

    Computer-assisted mitotic count using a deep learning–based algorithm improves interobserver reproducibility and accuracy

    Bertram,C.A.,Aubreville,M.,Donovan,T.A.,Bartel,A.,Wilm,F., Marzahl, C., Assenmacher, C.A., Becker, K., Bennett, M., Corner, S., et al., 2022. Computer-assisted mitotic count using a deep learning–based algorithm improves interobserver reproducibility and accuracy. Veterinary pathology 59, 211–226. doi:10.1177/ 03009858211067478

    2022

  15. [16]

    Computerized calculation of mitotic count distribution in canine cutaneous mast cell tumor sections: mitotic countisareadependent

    Bertram,C.A.,Aubreville,M.,Gurtner,C.,Bartel,A.,Corner,S.M., Dettwiler, M., Kershaw, O., Noland, E.L., Schmidt, A., Sledge, D.G., et al., 2020a. Computerized calculation of mitotic count distribution in canine cutaneous mast cell tumor sections: mitotic countisareadependent. Veterinarypathology57,214–226. doi:10. 1177/0300985819890686

  16. [17]

    A large-scale dataset for mitotic figure assessment on whole slide images of canine cutaneous mast cell tumor

    Bertram, C.A., Aubreville, M., Marzahl, C., Maier, A., Klopfleisch, R., 2019. A large-scale dataset for mitotic figure assessment on whole slide images of canine cutaneous mast cell tumor. Scientific data 6, 1–9. doi:10.1038/s41597-019-0290-4

    work page doi:10.1038/s41597-019-0290-4 2019

  17. [18]

    Atyp- ical mitotic figures are prognostically meaningful for canine cuta- neous mast cell tumors

    Bertram, C.A., Bartel, A., Donovan, T.A., Kiupel, M., 2023. Atyp- ical mitotic figures are prognostically meaningful for canine cuta- neous mast cell tumors. Veterinary Sciences 11, 5

    2023

  18. [19]

    Mitotic activity: A systematic literature review of the assessment methodology and prognostic value in canine tumors

    Bertram, C.A., Donovan, T.A., Bartel, A., 2024a. Mitotic activity: A systematic literature review of the assessment methodology and prognostic value in canine tumors. Veterinary pathology 61, 752–

  19. [20]

    doi:10.1177/03009858241239565

    work page doi:10.1177/03009858241239565

  20. [21]

    Mitotic activity: A systematic literature review of the assessment methodology and prognostic value in feline tumors

    Bertram, C.A., Donovan, T.A., Bartel, A., 2024b. Mitotic activity: A systematic literature review of the assessment methodology and prognostic value in feline tumors. Veterinary Pathology 61, 743–

  21. [22]

    doi:10.1177/03009858241239566. M. Aubreville et al.:Preprint submitted to ElsevierPage 14 of 18 The MIDOG 2025 Challenge

    work page doi:10.1177/03009858241239566 2025

  22. [23]

    Bertram, C.A., Veta, M., Marzahl, C., Stathonikos, N., Maier, A., Klopfleisch, R., Aubreville, M., 2020b. Are pathologist-defined la- bels reproducible? comparison of the tupac16 mitotic figure dataset with an alternative set of labels, in: Interpretable and Annotation- EfficientLearningforMedicalImageComputing.Springer,pp.204–

  23. [24]

    doi:10.1007/978-3-030-61166-8_22

    work page doi:10.1007/978-3-030-61166-8_22

  24. [25]

    Bertram, C.A., Weiss, V., Donovan, T.A., Banerjee, S., Conrad, T., Ammeling, J., Klopfleisch, R., Kaltenecker, C., Aubreville, M.,

  25. [26]

    Histologic dataset of normal and atypical mitotic figures on human breast cancer (ami-br), in: BVM Workshop, Springer. pp. 113–118

  26. [27]

    Robust Pan-Cancer Mitotic Figure Detection with YOLOv12, in: Aubre- ville,M.,Bertram,C.A.(Eds.),MitoticFigureDetectionandAtypia Classification in Whole Slide Images, Springer

    Bourgade, R., Balezo, G., Monier, L., Feki, H., Blons, M., Blondel, A., Loussouarn, D., Vincent-Salomon, A., Walter, T., 2026. Robust Pan-Cancer Mitotic Figure Detection with YOLOv12, in: Aubre- ville,M.,Bertram,C.A.(Eds.),MitoticFigureDetectionandAtypia Classification in Whole Slide Images, Springer. pp. 26–36

    2026

  27. [28]

    MIDOG 2025: Mitotic Figure Detection with Attention-Guided False Pos- itive Correction, in: Aubreville, M., Bertram, C.A

    Broad, A., Keighley, J., Godson, L., Wright, A., 2026. MIDOG 2025: Mitotic Figure Detection with Attention-Guided False Pos- itive Correction, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classification in Whole Slide Images, Springer. pp. 87–90

    2026

  28. [29]

    End-to-end object detection with transformers

    Carion, N., Massa, F., Synnaeve, G., Usunier, N., Kirillov, A., Zagoruyko,S.,2020. End-to-endobjectdetectionwithtransformers, in:Europeanconferenceoncomputervision,Springer.pp.213–229. doi:10.1007/978-3-030-58452-8_13

    work page doi:10.1007/978-3-030-58452-8_13 2020

  29. [30]

    Encoder-decoder with atrous separable convolution for semantic imagesegmentation,in:ProceedingsoftheEuropeanconferenceon computer vision (ECCV), pp

    Chen, L.C., Zhu, Y., Papandreou, G., Schroff, F., Adam, H., 2018. Encoder-decoder with atrous separable convolution for semantic imagesegmentation,in:ProceedingsoftheEuropeanconferenceon computer vision (ECCV), pp. 801–818

    2018

  30. [31]

    Towards a general-purpose foundation model for computational pathology

    Chen, R.J., Ding, T., Lu, M.Y., Williamson, D.F., Jaume, G., Song, A.H., Chen, B., Zhang, A., Shao, D., Shaban, M., et al., 2024. Towards a general-purpose foundation model for computational pathology. Nature medicine 30, 850–862

    2024

  31. [32]

    Teacher-Student Model for Detecting and Classifying Mitosis in the MIDOG 2025 Challenge, in: Aubreville, M., Bertram, C.A

    Choe, S., Qin, X., Shafique, A., Dy, A., Done, S., Androutsos, D., Khademi, A., 2026. Teacher-Student Model for Detecting and Classifying Mitosis in the MIDOG 2025 Challenge, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classification in Whole Slide Images, Springer. pp. 207–214

    2026

  32. [33]

    Groupequivariantconvolutionalnet- works, in: Proceedings of the International Conference on Machine Learning (ICML), pp

    Cohen,T.,Welling,M.,2016. Groupequivariantconvolutionalnet- works, in: Proceedings of the International Conference on Machine Learning (ICML), pp. 2990–2999

    2016

  33. [34]

    2022 , publisher =

    Detlefsen, N.S., Borovec, J., Schock, J., Jha, A.H., Koker, T., Di Liello, L., Stancl, D., Quan, C., Grechkin, M., Falcon, W., 2022. Torchmetrics - measuring reproducibility in pytorch. Journal of Open Source Software 7, 4101. URL:https://doi.org/10.21105/ joss.04101, doi:10.21105/joss.04101

    work page doi:10.21105/joss.04101 2022

  34. [35]

    Mitotic figures—normal, atypical, and imposters: A guide to identification

    Donovan, T.A., Moore, F.M., Bertram, C.A., Luong, R., Bolfa, P., Klopfleisch,R.,Tvedten,H.,Salas,E.N.,Whitley,D.B.,Aubreville, M., et al., 2021. Mitotic figures—normal, atypical, and imposters: A guide to identification. Veterinary pathology 58, 243–257

    2021

  35. [36]

    Stain-Aware Augmentation and Hybrid Loss for Domain-Generalized Atypical Mitosis Classification, in: Aubreville, M., Bertram, C.A

    Dukre, A., Deria, A., Xie, Y., Razzak, I., 2026. Stain-Aware Augmentation and Hybrid Loss for Domain-Generalized Atypical Mitosis Classification, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classification in Whole Slide Images, Springer. pp. 159–165

    2026

  36. [37]

    Why is the winner the best?, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, pp

    Eisenmann, M., Reinke, A., Weru, V., Tizabi, M.D., Isensee, F., Adler, T.J., Ali, S., Andrearczyk, V., Aubreville, M., Baid, U., et al., 2023. Why is the winner the best?, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, pp. 19955–19966

    2023

  37. [38]

    AutoGluon-Tabular: Robust and Accurate AutoML for Structured Data

    Erickson, N., Mueller, J., Shirkov, A., Zhang, H., Larroy, P., Li, M., Smola, A., 2020. Autogluon-tabular: Robust and accurate automl for structured data. arXiv preprint arXiv:2003.06505

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  38. [39]

    Protocolfortheexamination of resection specimens from patients with invasive carcinoma of the breast

    Fitzgibbons,P.L.,Connolly,J.L.,2023. Protocolfortheexamination of resection specimens from patients with invasive carcinoma of the breast. CAP guidelines 4.8.1.0. URL:https://www.cap.org/ cancerprotocols

    2023

  39. [40]

    Pannuke: an open pan-cancer histology dataset for nuclei instance segmentation and classification, in: European congress on digital pathology, Springer

    Gamper,J.,AlemiKoohbanani,N.,Benet,K.,Khuram,A.,Rajpoot, N., 2019. Pannuke: an open pan-cancer histology dataset for nuclei instance segmentation and classification, in: European congress on digital pathology, Springer. pp. 11–19

    2019

  40. [41]

    Ganz,J.,Marzahl,C.,Ammeling,J.,Rosbach,E.,Richter,B.,Puget, C., Denk, D., Demeter, E.A., Tăbăran, F.A., Wasinger, G., et al.,

  41. [42]

    Scientificreports 14, 26273

    Information mismatch in phh3-assisted mitosis annotation leadstointerpretationshiftsinh&eslideanalysis. Scientificreports 14, 26273

  42. [43]

    RF-DETR for Robust Mitotic Figure Detection: A MIDOG 2025 Track 1 Approach, in: Aubreville, M., Bertram, C.A

    Giedziun, P., Sotysik, J., Górczany, M., Ropiak, N., Przymus, M., Krajewski,P.,Kwicień,J.,Bartczak,A.,Wasiak,I.,Maniewski,M., 2026a. RF-DETR for Robust Mitotic Figure Detection: A MIDOG 2025 Track 1 Approach, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classification in Whole Slide Images, Springer. pp. 99–104

    2025

  43. [44]

    FoundationModel-DrivenClassificationofAtypicalMitotic FigureswithDomain-AwareTrainingStrategies,in:Aubreville,M., Bertram, C.A

    Giedziun, P., Sotysik, J., Górczany, M., Ropiak, N., Przymus, M., Krajewski,P.,Kwiecień,J.,Bartczak,A.,Wasiak,I.,Maniewski,M., 2026b. FoundationModel-DrivenClassificationofAtypicalMitotic FigureswithDomain-AwareTrainingStrategies,in:Aubreville,M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classifi- cation in Whole Slide Images, Springer. p...

  44. [45]

    Classification of chromosome segregation errors in cancer

    Gisselsson, D., 2008. Classification of chromosome segregation errors in cancer. Chromosoma 117, 511–519. doi:10.1007/ s00412-008-0169-1

    2008

  45. [46]

    Hover-Net: Simultaneous Segmentation and Classification of Nuclei in Multi-tissue Histology Images,

    Graham,S.,Vu,Q.D.,Raza,S.E.A.,Azam,A.,Tsang,Y.W.,Kwak, J.T., Rajpoot, N., 2019. Hover-net: Simultaneous segmentation and classification of nuclei in multi-tissue histology images. Medical Image Analysis 58, 101563. doi:10.1016/j.media.2019.101563

    work page doi:10.1016/j.media.2019.101563 2019

  46. [47]

    Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp

    He, K., Zhang, X., Ren, S., Sun, J., 2016. Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778. doi:10.1109/ CVPR.2016.90

    2016

  47. [48]

    Hendzel, M.J., Wei, Y., Mancini, M.A., Van Hooser, A., Ranalli, T., Brinkley, B., Bazett-Jones, D.P., Allis, C.D., 1997. Mitosis- specific phosphorylation of histone h3 initiates primarily within pericentromeric heterochromatin during g2 and spreads in an or- dered fashion coincident with mitotic chromosome condensation. Chromosoma 106, 348–360. doi:10.10...

    work page doi:10.1007/s004120050256 1997

  48. [49]

    , year = 2017, month = jul, pages =

    Huang, G., Liu, Z., Van Der Maaten, L., Weinberger, K.Q., 2017. Densely connected convolutional networks, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 4700–4708. doi:10.1109/CVPR.2017.243

    work page doi:10.1109/cvpr.2017.243 2017

  49. [50]

    Isensee, F., Jaeger, P.F., Kohl, S.A., Petersen, J., Maier-Hein, K.H.,

  50. [51]

    Nature methods 18, 203–211

    nnu-net: a self-configuring method for deep learning-based biomedical image segmentation. Nature methods 18, 203–211

  51. [52]

    A subphase-labeled mitotic dataset for ai-powered cell division analy- sis

    Ivan, Z.Z., Hirling, D., Grexa, I., Ammeling, J., Molnar, C., Micsik, T., Dobra, K., Kuthi, L., Sukosd, F., Fillinger, J., et al., 2026. A subphase-labeled mitotic dataset for ai-powered cell division analy- sis. Scientific data

    2026

  52. [53]

    Mitosis subtyping dataset

    Jahanifar, M., 2025. Mitosis subtyping dataset. URL:https://doi. org/10.5281/zenodo.15390543, doi:10.5281/zenodo.15390543

    work page doi:10.5281/zenodo.15390543 2025

  53. [54]

    Pan-cancerprofilingofmitotictopol- ogy & mitotic errors: Insights into prognosis, genomic alterations, and immune landscape

    Jahanifar, M., Dawood, M., Zamanitajeddin, N., Shephard, A., Chohan,B.S.,Bertram,C.A.,Wahab,N.,Eastwood,M.,Aubreville, M.,Raza,S.E.A.,etal.,2025. Pan-cancerprofilingofmitotictopol- ogy & mitotic errors: Insights into prognosis, genomic alterations, and immune landscape. medRxiv , 2025–06

    2025

  54. [55]

    Distinct mitotic segregation errors mediate chromo- somal instability in aggressive urothelial cancers

    Jin, Y., Stewénius, Y., Lindgren, D., Frigyesi, A., Calcagnile, O., Jonson, T., Edqvist, A., Larsson, N., Lundberg, L.M., Chebil, G., et al., 2007. Distinct mitotic segregation errors mediate chromo- somal instability in aggressive urothelial cancers. Clinical cancer research 13, 1703–1712

    2007

  55. [56]

    Tripolar mitosis in human cells and embryos: occurrence, pathophysiology and medical implications

    Kalatova, B., Jesenska, R., Hlinka, D., Dudas, M., 2015. Tripolar mitosis in human cells and embryos: occurrence, pathophysiology and medical implications. Acta histochemica 117, 111–125

    2015

  56. [57]

    Kather, J.N., Krisam, J., Charoentong, P., Luedde, T., Herpel, E., Weis, C.A., Gaiser, T., Marx, A., Valous, N.A., Ferber, D., et al.,

  57. [58]

    Predicting survival from colorectal cancer histology slides using deep learning: A retrospective multicenter study. PLoS M. Aubreville et al.:Preprint submitted to ElsevierPage 15 of 18 The MIDOG 2025 Challenge medicine 16, e1002730

    2025

  58. [59]

    Atypical mitotic figureclassificationinmidog2025usinglora-enhancedunimodels

    Kelam, N.S., Bonthu, S., Singhai, N., 2025. Atypical mitotic figureclassificationinmidog2025usinglora-enhancedunimodels. doi:https://doi.org/10.5281/zenodo.17020013

    work page doi:10.5281/zenodo.17020013 2025

  59. [60]

    Ensemble YOLO Framework for Multi-Domain Mitotic Figure Detection in Histopathology Images, in: Aubreville, M., Bertram, C.A

    Kelam, N.S., Parekh, A., Bonthu, S., Singhal, N., 2026. Ensemble YOLO Framework for Multi-Domain Mitotic Figure Detection in Histopathology Images, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classification in Whole Slide Images, Springer. pp. 105–111

    2026

  60. [61]

    YOLOv11: An Overview of the Key Architectural Enhancements

    Khanam, R., Hussain, M., 2024. Yolov11: An overview of the key architectural enhancements. arXiv preprint arXiv:2410.17725

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  61. [62]

    Proposal of a 2-tier histologic grading system for canine cutaneous mastcelltumorstomoreaccuratelypredictbiologicalbehavior

    Kiupel, M., Webster, J., Bailey, K., Best, S., DeLay, J., Detrisac, C., Fitzgerald, S., Gamble, D., Ginn, P., Goldschmidt, M., et al., 2011. Proposal of a 2-tier histologic grading system for canine cutaneous mastcelltumorstomoreaccuratelypredictbiologicalbehavior. Vet. Pathol. 48, 147–155. doi:10.1177/0300985810386469

    work page doi:10.1177/0300985810386469 2011

  62. [63]

    MIDOG 2025 Track 2: A Deep Learning Model for Classification of Atypical and Normal Mitotic Figures under Class and Hardness Imbalances, in: Aubreville, M., Bertram, C.A

    Kotte, S., Saipradeep, V.G., Walia, V., Nandagopal, D., Joseph, T., Sivadasan, N., Lali, B.S., 2026. MIDOG 2025 Track 2: A Deep Learning Model for Classification of Atypical and Normal Mitotic Figures under Class and Hardness Imbalances, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classifi- cation in Whole Slide Images, S...

    2026

  63. [64]

    Deep Learning Meets Morphology: A Hybrid Approach for Mitotic Figure Classification, in: Aubreville, M., Bertram, C.A

    Krauss, S., Spieß, E., Hieber, D., Kramer, F., Schobel, J., Müller, D., 2026. Deep Learning Meets Morphology: A Hybrid Approach for Mitotic Figure Classification, in: Aubreville, M., Bertram, C.A. (Eds.),MitoticFigureDetectionandAtypiaClassificationinWhole Slide Images, Springer. pp. 37–46

    2026

  64. [65]

    The hungarian method for the assignment problem

    Kuhn, H.W., 1955. The hungarian method for the assignment problem. Naval research logistics quarterly 2, 83–97

    1955

  65. [66]

    Roto-translation equivariant convolutional networks: Application to histopathologyimageanalysis

    Lafarge, M.W., Bekkers, E.J., Pluim, J.P., Duits, R., Veta, M., 2021. Roto-translation equivariant convolutional networks: Application to histopathologyimageanalysis. MedicalImageAnalysis68,101849

    2021

  66. [67]

    Sequential Hard Mining: A Data-Centric Approach for Mitosis Detection, in: Aubreville, M., Bertram, C.A

    Lafarge, M.W., Koelzer, V.H., 2026. Sequential Hard Mining: A Data-Centric Approach for Mitosis Detection, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classi- fication in Whole Slide Images, Springer. pp. 215–219

    2026

  67. [68]

    The characteristics and clinical significance of atypicalmitosisinbreastcancer

    Lashen, A., Toss, M.S., Alsaleem, M., Green, A.R., Mongan, N.P., Rakha, E., 2022. The characteristics and clinical significance of atypicalmitosisinbreastcancer. ModernPathology35,1341–1348

    2022

  68. [69]

    Adeepactivelearningframeworkformitoticfiguredetec- tion with minimal manual annotation and labelling

    Liu, E., Lin, A., Kakodkar, P., Zhao, Y., Wang, B., Ling, C., Zhang, Q.,2025. Adeepactivelearningframeworkformitoticfiguredetec- tion with minimal manual annotation and labelling. Histopathology 87, 536–547

    2025

  69. [70]

    Ef- ficientvit:Memoryefficientvisiontransformerwithcascadedgroup attention,in:ProceedingsoftheIEEE/CVFconferenceoncomputer vision and pattern recognition, pp

    Liu, X., Peng, H., Zheng, N., Yang, Y., Hu, H., Yuan, Y., 2023. Ef- ficientvit:Memoryefficientvisiontransformerwithcascadedgroup attention,in:ProceedingsoftheIEEE/CVFconferenceoncomputer vision and pattern recognition, pp. 14420–14430

    2023

  70. [71]

    Detecting cancer metastases on gigapixel pathol- ogy images

    Liu, Y., Gadepalli, K., Norouzi, M., Dahl, G.E., Kohlberger, T., Boyko, A., Venugopalan, S., Timofeev, A., Nelson, P.Q., Corrado, G.S., et al., 2017. Detecting cancer metastases on gigapixel pathol- ogy images. arXiv preprint arXiv:1703.02442

    work page arXiv 2017

  71. [72]

    Liu, Z., Mao, H., Wu, C.Y., Feichtenhofer, C., Darrell, T., Xie, S.,

  72. [73]

    11976– 11986

    A convnet for the 2020s, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 11976– 11986

  73. [74]

    KongNet: A Multi-headed Deep Learning Model for Detection and Classification of Nuclei in Histopathology Images

    Lv, J., Nasir, E.S., Xu, K., Jahanifar, M., Chohan, B.S., Elhaminia, B., Raza, S.E.A., 2026. Kongnet: A multi-headed deep learning model for detection and classification of nuclei in histopathology images. URL:https://arxiv.org/abs/2510.23559,arXiv:2510.23559

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  74. [75]

    Rtmdet: An empirical study of designing real- time object detectors

    Lyu, C., Zhang, W., Huang, H., Zhou, Y., Wang, Y., Liu, Y., Zhang, S., Chen, K., 2022. Rtmdet: An empirical study of designing real- time object detectors. arXiv preprint arXiv:2212.07784

    work page arXiv 2022

  75. [76]

    Mitotic fig- ure recognition: Agreement among pathologists and computerized detector

    Malon, C., Brachtel, E., Cosatto,E., Graf, H.P., Kurata, A., Kuroda, M., Meyer, J.S., Saito, A., Wu, S., Yagi, Y., 2012. Mitotic fig- ure recognition: Agreement among pathologists and computerized detector. Analytical Cellular Pathology 35, 97–100. doi:10.3233/ ACP-2011-0029

    2012

  76. [77]

    Marzahl, C., Aubreville, M., Bertram, C.A., Maier, J., Bergler, C., Kröger, C., Voigt, J., Breininger, K., Klopfleisch, R., Maier, A.,

  77. [78]

    Scientific Reports 11:4343, 1–10

    EXACT: a collaboration toolset for algorithm-aided anno- tation of images with annotation version control. Scientific Reports 11:4343, 1–10. doi:10.1038/s41598-021-83827-4

    work page doi:10.1038/s41598-021-83827-4

  78. [79]

    A bag of tricks for real-time Mitotic Figure detection, in: Aubreville, M., Bertram, C.A

    Marzahl, C., Napora, B., 2026. A bag of tricks for real-time Mitotic Figure detection, in: Aubreville, M., Bertram, C.A. (Eds.), Mitotic Figure Detection and Atypia Classification in Whole Slide Images, Springer. pp. 91–98

    2026

  79. [80]

    Mitotic index and multipolar mitosis in routine histologic sectionsasprognosticmarkersofpancreaticcancers:aclinicopatho- logical study

    Matsuda, Y., Yoshimura, H., Ishiwata, T., Sumiyoshi, H., Mat- sushita, A., Nakamura, Y., Aida, J., Uchida, E., Takubo, K., Arai, T., 2016. Mitotic index and multipolar mitosis in routine histologic sectionsasprognosticmarkersofpancreaticcancers:aclinicopatho- logical study. Pancreatology 16, 127–132

    2016

  80. [81]

    Evaluation of prognostic factors for dogs with primarylungtumors:67cases(1985-1992).JournaloftheAmerican Veterinary Medical Association 211, 1422–1427

    McNiel, E., Ogilvie, G., Powers, B., Hutchison, J., Salman, M., Withrow, S., 1997. Evaluation of prognostic factors for dogs with primarylungtumors:67cases(1985-1992).JournaloftheAmerican Veterinary Medical Association 211, 1422–1427

    1997

Showing first 80 references.