pith. sign in

arxiv: 2607.00802 · v1 · pith:6RZXNVSNnew · submitted 2026-07-01 · 💻 cs.MM

CellPrior-Net: Prior-Guided Nuclei Detection and Classification for H&E Whole-Slide Images

Falah Jabar , Pasquale Lombardi , Aria Torkpour , Masoud Tafavvoghi , Per Niklas Benzler Waaler , Sigve Andersen , Erna-Elise Paulsen , Mette P{\o}hl
show 5 more authors
Lill-Tove Rasmussen Busund Tom Donnem Elin Richardsen David J. Pinato Mehrdad Rakaee
This is my paper

Pith reviewed 2026-07-02 01:47 UTC · model grok-4.3

classification 💻 cs.MM
keywords nuclei detectionH&E whole-slide imageshematoxylin channel priorlightweight CNNcell classificationcomputational pathologyinference efficiencytumor microenvironment
0
0 comments X

The pith

CellPriorNet uses the hematoxylin channel as prior information in a lightweight CNN to detect and classify nuclei in H&E whole-slide images with accuracy comparable to heavier models but much lower inference time.

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

The paper introduces CellPriorNet as a pipeline that feeds the hematoxylin channel directly into a lightweight convolutional network to guide nuclei-aware feature learning. It benchmarks this on eight public and private datasets containing 10.4 million nuclei drawn from varied organs, scanners, magnifications, and staining conditions. The central result is that the method matches the detection and classification performance of existing heavier pipelines while cutting inference time substantially. A follow-on pipeline called CellQuant Net adds a quality-assessment step to skip artifact regions before running CellPriorNet. The work positions this efficiency gain as enabling routine quantitative analysis of tumor microenvironments at whole-slide scale.

Core claim

CellPriorNet is an efficient nuclei detection and classification pipeline that utilizes a lightweight convolutional neural network architecture and hematoxylin channel as prior information to enhance nuclei-aware feature learning, achieving comparable performance to state-of-the-art methods while significantly reducing inference time across eight datasets totaling 10.4 million nuclei from different organs, scanners, magnifications, and clinical centers.

What carries the argument

Hematoxylin channel prior fed into a lightweight CNN to guide nuclei-aware feature learning.

If this is right

  • Whole-slide images can be processed at scale without heavy post-processing or large model footprints.
  • CellQuant Net extends the method into an end-to-end quantification pipeline that first excludes artifact regions.
  • The approach generalizes across organs and acquisition conditions without retraining for each new scanner or staining protocol.
  • Downstream computational pathology tasks that rely on nuclei counts or types become feasible on routine clinical hardware.

Where Pith is reading between the lines

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

  • The same prior-channel idea could be tested on other stain channels or on fluorescence images to check whether the efficiency gain transfers.
  • Integration with existing digital pathology viewers might allow real-time nuclei overlays during slide review.
  • If the speed improvement holds, large retrospective cohorts could be re-analyzed for microenvironment metrics that were previously too costly to compute.

Load-bearing premise

The hematoxylin channel supplies reliable prior information that improves nuclei feature learning even when nuclei morphology, staining, scanners, organs, magnifications, and slide artifacts all vary.

What would settle it

Running the same lightweight network on the same test sets with the hematoxylin channel replaced by random noise or omitted entirely and measuring whether detection and classification metrics fall below the reported comparable levels.

read the original abstract

Accurate nuclei detection and classification in hematoxylin and eosin (H and E) whole-slide images (WSIs) is a key task in computational pathology, particularly for quantitative analysis of the tumor microenvironment. However, this task remains highly challenging due to variations in nuclei morphology, staining procedures, scanners, organs, magnifications, and WSI artifacts. In addition, many existing pipelines rely on computationally demanding architectures and post-processing procedures, making gigapixel WSI analysis time consuming. In this work, CellPriorNet (CP Net) is proposed, an efficient nuclei detection and classification pipeline that utilizes a lightweight convolutional neural network architecture and hematoxylin (H) channel as prior information to enhance nuclei-aware feature learning. Extensive benchmarking was conducted against state of the art pipelines on 8 public and private datasets (total:10.4M nuclei) obtained from different organs, scanners, magnifications, and clinical centers. Experimental results demonstrate that CP Net achieves comparable performance while significantly reducing inference time. Furthermore, CellQuant Net was introduced, an end to end nuclei quantification pipeline, that integrates a quality assessment (QA) model to exclude regions with artifacts, followed by CP-Net cell detection and classification. The pipeline is publicly available on GitHub, and provides a potentially efficient and scalable framework for downstream computational pathology applications.

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 introduces CellPrior-Net (CP Net), a lightweight CNN-based pipeline for nuclei detection and classification in H&E whole-slide images that incorporates the hematoxylin channel as prior information to improve nuclei-aware feature learning. It benchmarks the method against state-of-the-art approaches on eight public and private datasets totaling 10.4 million nuclei spanning multiple organs, scanners, magnifications, and clinical centers, claiming comparable detection/classification performance with substantially reduced inference time. The work also presents CellQuant Net, an end-to-end quantification pipeline that adds a quality-assessment model to exclude artifact regions before applying CP Net.

Significance. If the reported performance and runtime gains are reproducible, the contribution lies in delivering a practical, scalable alternative for gigapixel WSI analysis in computational pathology, potentially lowering barriers to large-scale tumor-microenvironment quantification without requiring heavy post-processing.

major comments (2)
  1. [Abstract and §4] Abstract and §4 (Experimental Results): the central claim of 'comparable performance' is stated without reference to specific per-dataset metrics (e.g., F1, precision-recall, or Panoptic Quality) or statistical significance tests against the cited SOTA baselines; without these numbers it is impossible to judge whether any observed runtime reduction comes at an acceptable accuracy cost.
  2. [§3.2 and §4.3] §3.2 (Architecture) and §4.3 (Ablation): the motivation that the H-channel prior 'enhances nuclei-aware feature learning' across staining, scanner, and magnification variations is presented as an empirical design choice, yet no ablation isolating the H-channel input versus standard RGB or other color-deconvolution channels is reported; this leaves the load-bearing role of the prior unverified.
minor comments (2)
  1. [§4] The GitHub link for CellQuant Net is mentioned but the manuscript does not specify the exact train/validation/test splits or any cross-center evaluation protocol used for the eight datasets.
  2. [Tables and Figures] Figure captions and Table 1 should explicitly list the number of nuclei per dataset rather than only the aggregate 10.4 M figure.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help strengthen the clarity and rigor of our claims. We address each major point below and will revise the manuscript to incorporate the requested details.

read point-by-point responses

  1. Referee: [Abstract and §4] Abstract and §4 (Experimental Results): the central claim of 'comparable performance' is stated without reference to specific per-dataset metrics (e.g., F1, precision-recall, or Panoptic Quality) or statistical significance tests against the cited SOTA baselines; without these numbers it is impossible to judge whether any observed runtime reduction comes at an acceptable accuracy cost.

    Authors: We agree that the abstract and §4 would benefit from explicit per-dataset metrics and statistical tests to substantiate the 'comparable performance' claim. In the revised manuscript we will expand §4 with detailed tables reporting F1, precision, recall, and Panoptic Quality for each of the eight datasets, together with statistical significance tests (paired t-tests or Wilcoxon signed-rank tests with p-values) against the cited SOTA baselines. These additions will allow direct assessment of the accuracy-runtime trade-off. revision: yes

  2. Referee: [§3.2 and §4.3] §3.2 (Architecture) and §4.3 (Ablation): the motivation that the H-channel prior 'enhances nuclei-aware feature learning' across staining, scanner, and magnification variations is presented as an empirical design choice, yet no ablation isolating the H-channel input versus standard RGB or other color-deconvolution channels is reported; this leaves the load-bearing role of the prior unverified.

    Authors: We concur that an ablation isolating the H-channel prior is required to verify its contribution. The revised manuscript will include a new ablation study in §4.3 that compares model performance (detection and classification metrics) when using the H-channel input versus standard RGB input and versus other color-deconvolution channels (e.g., eosin-only) across the same eight datasets. This will provide direct empirical support for the design choice. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper presents an empirical CNN architecture (CellPrior-Net) that ingests the hematoxylin channel as an additional input channel and reports end-to-end detection/classification metrics on 8 datasets. No equations, first-principles derivations, fitted parameters renamed as predictions, or self-citation chains appear in the abstract or described experimental design. The central claims rest on standard supervised training and benchmarking against external SOTA methods; the H-channel prior is an architectural choice whose value is measured by observed performance rather than by any internal identity or self-referential reduction. The work is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Abstract-only view limits visibility into parameters; main reliance is on the domain assumption that H-channel prior improves feature learning.

axioms (1)
  • domain assumption Hematoxylin channel serves as effective prior information to enhance nuclei-aware feature learning in H&E images despite variations in staining and scanners
    Invoked in abstract as the core design choice for the architecture.
invented entities (1)
  • CellPrior-Net (CP Net) no independent evidence
    purpose: Lightweight CNN pipeline for nuclei detection and classification
    New model name and architecture introduced in the work.

pith-pipeline@v0.9.1-grok · 5833 in / 1237 out tokens · 35474 ms · 2026-07-02T01:47:15.748229+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

38 extracted references · 36 canonical work pages · 1 internal anchor

  1. [1]

    Deep Learning Model for Predicting Immunotherapy Response in Advanced Non−Small Cell Lung Cancer,

    M. Rakaee et al., “Deep Learning Model for Predicting Immunotherapy Response in Advanced Non−Small Cell Lung Cancer,” JAMA Oncol., vol. 11, no. 2, Feb. 2025, doi: 10.1001/jamaoncol.2024.5356

    work page doi:10.1001/jamaoncol.2024.5356 2025

  2. [2]

    Ancestry-Associated Performance Variability of Open -Source AI Models for EGFR Prediction in Lung Cancer,

    M. Rakaee et al., “Ancestry-Associated Performance Variability of Open -Source AI Models for EGFR Prediction in Lung Cancer,” JAMA Oncol., vol. 12, no. 4, Apr. 2026, doi: 10.1001/jamaoncol.2025.6430

    work page doi:10.1001/jamaoncol.2025.6430 2026

  3. [4]

    arXiv:2408.00738 (2024)

    E. Zimmermann et al., “Virchow2: Scaling Self-Supervised Mixed Magnification Models in Pathology,” ArXiv PrePrint, vol. 1, no. 1, Nov. 2024, doi: https://arxiv.org/abs/2408.00738

    work page arXiv 2024

  4. [5]

    A visual -language Foundation Model for Computational Pathology,

    M. Y. Lu et al. , “A visual -language Foundation Model for Computational Pathology,” Nat. Med. , vol. 30, no. 3, Mar. 2024, doi: 10.1038/s41591-024-02856-4

    work page doi:10.1038/s41591-024-02856-4 2024

  5. [6]

    Towards a General-Purpose Foundation Model for Computational Pathology,

    R. J. Chen et al., “Towards a General-Purpose Foundation Model for Computational Pathology,” Nat. Med., vol. 30, no. 3, Mar. 2024, doi: 10.1038/s41591-024-02857-3

    work page doi:10.1038/s41591-024-02857-3 2024

  6. [7]

    A Whole-Slide Foundation Model for Digital Pathology from Real-World Data,

    H. Xu et al., “A Whole-Slide Foundation Model for Digital Pathology from Real-World Data,” Nature, vol. 630, no. 8015, Jun. 2024, doi: 10.1038/s41586-024-07441-w

    work page doi:10.1038/s41586-024-07441-w 2024

  7. [8]

    Phikon-v2, A Large and Public Feature Extractor for Biomarker Prediction,

    A. Filiot, P. Jacob, A. Mac Kain, and C. Saillard, “Phikon-v2, A Large and Public Feature Extractor for Biomarker Prediction,” ArXiv PrePrint, vol. 1, no. 1, Sep. 2024, doi: https://arxiv.org/abs/2409.09173

    work page arXiv 2024

  8. [9]

    H-optimus-0,

    C. and J. R. and L.-L. F. and M. Z. and C. D. and D. E. and V. J.-P. Saillard, “H-optimus-0,” Github. Accessed: Jun. 03, 2026. [Online]. Available: https://github.com/bioptimus/releases/tree/main/models/h-optimus/v0

    2026

  9. [10]

    Visual Language Pretrained Multiple Instance Zero-Shot Transfer for Histopathology Images,

    M. Y. Lu et al., “Visual Language Pretrained Multiple Instance Zero-Shot Transfer for Histopathology Images,” in 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition , Vancouver, BC, Canada: IEEE, Jun. 2023. doi: 10.1109/CVPR52729.2023.01893

    work page doi:10.1109/cvpr52729.2023.01893 2023

  10. [11]

    Hibou: A Family of Foundational Vision Transformers for Pathology,

    D. Nechaev, A. Pchelnikov, and E. Ivanova, “Hibou: A Family of Foundational Vision Transformers for Pathology,” ArXiv PrePrint, vol. 1, no. 1, Aug. 2024, doi: https://arxiv.org/abs/2406.05074

    work page arXiv 2024

  11. [12]

    Transformer-Based Unsupervised Contrastive Learning for Histopathological Image Classification,

    X. Wang et al., “Transformer-Based Unsupervised Contrastive Learning for Histopathological Image Classification,” Med. Image Anal., vol. 81, no. 1, Oct. 2022, doi: 10.1016/j.media.2022.102559

    work page doi:10.1016/j.media.2022.102559 2022

  12. [13]

    A Clinical Benchmark of Public Self -Supervised Pathology Foundation Models,

    G. Campanella et al., “A Clinical Benchmark of Public Self -Supervised Pathology Foundation Models,” Nat. Commun., vol. 16, no. 1, Apr. 2025, doi: 10.1038/s41467-025-58796-1

    work page doi:10.1038/s41467-025-58796-1 2025

  13. [14]

    Benchmarking Pathology Foundation Models: Adaptation Strategies and Scenarios,

    J. Lee, J. Lim, K. Byeon, and J. T. Kwak, “Benchmarking Pathology Foundation Models: Adaptation Strategies and Scenarios,” Comput. Biol. Med., vol. 190, no. 1, May 2025, doi: 10.1016/j.compbiomed.2025.110031

    work page doi:10.1016/j.compbiomed.2025.110031 2025

  14. [15]

    A Multimodal Generative AI Copilot for Human Pathology,

    M. Y. Lu et al. , “A Multimodal Generative AI Copilot for Human Pathology,” Nature, vol. 634, no. 8033, Oct. 2024, doi: 10.1038/s41586-024-07618-3

    work page doi:10.1038/s41586-024-07618-3 2024

  15. [16]

    A multimodal and temporal foundation model for virtual patient representations at healthcare system scale

    A. Zhang et al., “A Multimodal and Temporal Foundation Model for Virtual Patient Representations at Healthcare System Scale,” ArXiv PrePrint, vol. 1, no. 1, Apr. 2026, doi: https://arxiv.org/abs/2604.18570

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  16. [17]

    Marks, U

    M. Marks et al. , “CellSAM: a Foundation Model for Cell Segmentation,” Nat. Methods , vol. 22, no. 12, Dec. 2025, doi: 10.1038/s41592-025-02879-w

    work page doi:10.1038/s41592-025-02879-w 2025

  17. [18]

    Horst, T

    F. Hörst et al., “CellViT: Vision Transformers for Precise Cell Segmentation and Classification,” Med. Image Anal., vol. 94, no. 1, May 2024, doi: 10.1016/j.media.2024.103143

    work page doi:10.1016/j.media.2024.103143 2024

  18. [19]

    CellViT++: Energy-Efficient and Adaptive Cell Segmentation and Classification Using Foundation Models,

    F. Hörst, M. Rempe, H. Becker, L. Heine, J. Keyl, and J. Kleesiek, “CellViT++: Energy-Efficient and Adaptive Cell Segmentation and Classification Using Foundation Models,” Comput. Methods Programs Biomed. , vol. 277, no. 1, Apr. 2026, doi: 10.1016/j.cmpb.2025.109206

    work page doi:10.1016/j.cmpb.2025.109206 2026

  19. [20]

    LKCell: Efficient Cell Nuclei Instance Segmentation with Large Convolution Kernels,

    Z. Cui, J. Yao, L. Zeng, J. Yang, W. Liu, and X. Wang, “LKCell: Efficient Cell Nuclei Instance Segmentation with Large Convolution Kernels,” ArXiv Prepr, vol. 1, no. 1, Jul. 2024, doi: https://arxiv.org/abs/2407.18054 12

    work page arXiv 2024

  20. [21]

    Cell Nuclei Detection and Classification in Whole Slide Images with Transformers,

    O. Pina, E. Dorca, and V. Vilaplana, “Cell Nuclei Detection and Classification in Whole Slide Images with Transformers,” ArXiv Prepr, vol. 1, Feb. 2025, doi: https://arxiv.org/abs/2502.06307

    work page arXiv 2025

  21. [23]

    One Model is All You Need: Multi -Task Learning Enables Simultaneous Histology Image Segmentation and Classification,

    S. Graham et al., “One Model is All You Need: Multi -Task Learning Enables Simultaneous Histology Image Segmentation and Classification,” Med. Image Anal., vol. 83, no. 1, Jan. 2023, doi: 10.1016/j.media.2022.102685

    work page doi:10.1016/j.media.2022.102685 2023

  22. [24]

    PointNu -Net: Keypoint-Assisted Convolutional Neural Network for Simultaneous Multi-Tissue Histology Nuclei Segmentation and Classification,

    K. Yao, K. Huang, J. Sun, and A. Hussain, “PointNu -Net: Keypoint-Assisted Convolutional Neural Network for Simultaneous Multi-Tissue Histology Nuclei Segmentation and Classification,” IEEE Trans. Emerg. Top. Comput. Intell. , vol. 8, no. 1, Feb. 2024, doi: 10.1109/TETCI.2023.3281864

    work page doi:10.1109/tetci.2023.3281864 2024

  23. [25]

    TSFD-Net: Tissue Specific Feature Distillation Network for Nuclei Segmentation and Classification,

    T. Ilyas, Z. I. Mannan, A. Khan, S. Azam, H. Kim, and F. De Boer, “TSFD-Net: Tissue Specific Feature Distillation Network for Nuclei Segmentation and Classification,” Neural Networks, vol. 151, no. 1, Jul. 2022, doi: 10.1016/j.neunet.2022.02.020

    work page doi:10.1016/j.neunet.2022.02.020 2022

  24. [26]

    Fully Automatic Content-Aware Tiling Pipeline for Pathology Whole Slide Images,

    F. Jabar et al., “Fully Automatic Content-Aware Tiling Pipeline for Pathology Whole Slide Images,” Intell. Based. Med., vol. 12, no. 1, 2025, doi: 10.1016/j.ibmed.2025.100318

    work page doi:10.1016/j.ibmed.2025.100318 2025

  25. [27]

    PANDA -PLUS-Bench: A Clinical Benchmark for Evaluating the Robustness of AI Foundation Models in Prostate Cancer Diagnosis,

    J. L. Ebbert and D. Della Corte, “PANDA -PLUS-Bench: A Clinical Benchmark for Evaluating the Robustness of AI Foundation Models in Prostate Cancer Diagnosis,” AI in Medicine, vol. 1, no. 2, May 2026, doi: 10.3390/aimed1020014

    work page doi:10.3390/aimed1020014 2026

  26. [28]

    GrandQC: A Comprehensive Solution to Quality Control Problem in Digital Pathology,

    Z. Weng et al., “GrandQC: A Comprehensive Solution to Quality Control Problem in Digital Pathology,” Nat. Commun., vol. 15, no. 1, Dec. 2024, doi: 10.1038/s41467-024-54769-y

    work page doi:10.1038/s41467-024-54769-y 2024

  27. [29]

    NuLite: Lightweight and Fast Model for Nuclei Instance Segmentation and Cassification,

    C. Tommasino, C. Russo, and A. M. Rinaldi, “NuLite: Lightweight and Fast Model for Nuclei Instance Segmentation and Cassification,” Biomed. Signal Process. Control, vol. 114, no. 1, Apr. 2026, doi: 10.1016/j.bspc.2025.109333

    work page doi:10.1016/j.bspc.2025.109333 2026

  28. [30]

    Segment Anything Model (SAM) for Medical Image Segmentation: A Preliminary Review,

    L. Zhang, X. Deng, and Y. Lu, “Segment Anything Model (SAM) for Medical Image Segmentation: A Preliminary Review,” in 2023 IEEE International Conference on Bioinformatics and Biomedici , IEEE, Dec. 2023. doi: 10.1109/BIBM58861.2023.10386032

    work page doi:10.1109/bibm58861.2023.10386032 2023

  29. [31]

    Lizard: A Large -Scale Dataset for Colonic Nuclear Instance Segmentation and Classification,

    S. Graham et al. , “Lizard: A Large -Scale Dataset for Colonic Nuclear Instance Segmentation and Classification,” in 2021 IEEE/CVF International . Conference on Computer Vision Workshops, Montreal, BC, Canada: IEEE, Oct. 2021. doi: 10.1109/ICCVW54120.2021.00082

    work page doi:10.1109/iccvw54120.2021.00082 2021

  30. [32]

    NuCLS: A scalable Crowdsourcing Approach and Dataset for Nucleus Classification and Segmentation in Breast Cancer,

    M. Amgad et al., “NuCLS: A scalable Crowdsourcing Approach and Dataset for Nucleus Classification and Segmentation in Breast Cancer,” Gigascience, vol. 11, no. 1, May 2022, doi: 10.1093/gigascience/giac037

    work page doi:10.1093/gigascience/giac037 2022

  31. [33]

    A Panoptic Segmentation Dataset and Deep - Learning Approach for Explainable Scoring of Tumor -Infiltrating Lymphocytes,

    S. Liu, M. Amgad, D. More, M. A. Rathore, R. Salgado, and L. A. D. Cooper, “A Panoptic Segmentation Dataset and Deep - Learning Approach for Explainable Scoring of Tumor -Infiltrating Lymphocytes,” NPJ Breast Cancer, vol. 10, no. 1, Jun. 2024, doi: 10.1038/s41523-024-00663-1

    work page doi:10.1038/s41523-024-00663-1 2024

  32. [34]

    Restaining-Based Annotation for Cancer Histology Segmentation to Overcome Annotation-Related Limitations Among Pathologists,

    D. Komura et al., “Restaining-Based Annotation for Cancer Histology Segmentation to Overcome Annotation-Related Limitations Among Pathologists,” Patterns, vol. 4, no. 2, Feb. 2023, doi: 10.1016/j.patter.2023.100688

    work page doi:10.1016/j.patter.2023.100688 2023

  33. [35]

    PanNuke Dataset Extension, Insights and Baselines,

    J. Gamper et al. , “PanNuke Dataset Extension, Insights and Baselines,” ArXiv Prep , vol. 1, no. 1, Apr. 2020. doi: https://arxiv.org/abs/2003.10778

    work page arXiv 2020

  34. [36]

    Machine Learning-Based Immune Phenotypes Correlate with STK11/KEAP1 Co -mutations and Prognosis in Resectable NSCLC: a Sub-study of the TNM-I Trial,

    M. Rakaee et al., “Machine Learning-Based Immune Phenotypes Correlate with STK11/KEAP1 Co -mutations and Prognosis in Resectable NSCLC: a Sub-study of the TNM-I Trial,” Annals of Oncology, vol. 34, no. 7, Jul. 2023, doi: 10.1016/j.annonc.2023.04.005

    work page doi:10.1016/j.annonc.2023.04.005 2023

  35. [37]

    A Method for Normalizing Histology Slides for Quantitative Analysis,

    M. Macenko et al. , “A Method for Normalizing Histology Slides for Quantitative Analysis,” in 2009 IEEE International Symposium on Biomedical Imaging: From Nano to Macro, Boston, MA, USA: IEEE, Jun. 2009. doi: 10.1109/ISBI.2009.5193250

    work page doi:10.1109/isbi.2009.5193250 2009

  36. [38]

    UniRepLKNet: A Universal Perception Large -Kernel ConvNet for Audio, Video, Point Cloud, Time -Series and Image Recognition,

    X. Ding et al., “UniRepLKNet: A Universal Perception Large -Kernel ConvNet for Audio, Video, Point Cloud, Time -Series and Image Recognition,” in 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA: IEEE, Jun. 2024. doi: 10.1109/CVPR52733.2024.00527

    work page doi:10.1109/cvpr52733.2024.00527 2024

  37. [39]

    Nearest Neighbors Analysis,

    Scikit-learn Developers, “Nearest Neighbors Analysis,” Scikit -learn Documentation , June . 2026, doi: https://scikit- learn.org/stable/modules/generated/sklearn.neighbors.NearestNeighbors.html Supplementary Material ConSeP [1]: This dataset consists of 41 image tiles collected from colon cancer patients at the University Hospitals Coventry and Warwickshi...

    2026

  38. [40]

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

    S. Graham et al., “Hover-Net: Simultaneous Segmentation and Classification of Nuclei in Multi-tissue Histology Images,” Med. Image Anal., vol. 58, no. 1, Dec. 2019, doi: 10.1016/j.media.2019.101563 [2] S. Graham et al., “Lizard: A Large-Scale Dataset for Colonic Nuclear Instance Segmentation and Classification,” in 2021 IEEE/CVF International Conference o...

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