pith. sign in

arxiv: 2605.18522 · v1 · pith:6W2ZE4JRnew · submitted 2026-05-18 · 💻 cs.CV · cs.AI· cs.LG

Beyond Morphology: Quantifying the Diagnostic Power of Color Features in Cancer Classification

Farnaz Kheiri , Shahryar Rahnamayan , Masoud Makrehchi This is my paper

Pith reviewed 2026-05-20 11:16 UTC · model grok-4.3

classification 💻 cs.CV cs.AIcs.LG
keywords color featureshistopathologycancer classificationbenign versus malignantRGB histogramsHSV color spacemachine learningpre-screening
0
0 comments X

The pith

Color features alone can classify benign versus malignant tissue with up to 89 percent accuracy.

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

This paper tests whether raw color information in histopathology images carries enough signal to distinguish cancer cases without any shape or structural details. The authors extract only statistical color moments and binned histograms from RGB and HSV channels, then feed these into ordinary machine learning classifiers across ten different experimental setups. Performance reaches as high as 89 percent on binary benign-malignant tasks and stays well above random guessing, which the work links to broad color shifts tied to malignancy. The results point to these lightweight color measures as a practical first filter that could flag obvious cases before heavier models are applied.

Core claim

The authors show that global color moments and discretized RGB and HSV histograms, deliberately stripped of all morphological cues, still produce accuracies up to 89 percent when used to separate benign from malignant histopathology samples. This outcome holds across multiple datasets and classical classifiers, and the paper attributes the signal to consistent chromatic changes associated with malignancy rather than to tissue architecture.

What carries the argument

Statistical color moments together with discretized RGB and HSV color histograms that capture only global intensity distributions.

If this is right

  • Simple color features can serve as an effective pre-screening step that identifies samples with strong chromatic signs of malignancy.
  • These lightweight models could reduce the load on more complex deep-learning systems by triaging obvious cases first.
  • Raw color distributions encode a non-random diagnostic signal that works reliably in binary benign-malignant decisions.
  • The approach remains effective across a range of experimental settings without requiring structural cues.

Where Pith is reading between the lines

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

  • Color-based triage might prove especially useful in resource-limited settings where full deep models are expensive to run.
  • The findings suggest that future pipelines could benefit from explicit color normalization steps to isolate the malignancy signal more cleanly.
  • Similar global color measures could be tested on other tissue types or imaging modalities to see whether chromatic shifts are a general marker of pathology.

Load-bearing premise

The measured color statistics and histograms contain no hidden shape information and the color differences truly reflect malignancy rather than staining or scanner differences.

What would settle it

Re-running the identical color-feature classifiers on a set of images that have been normalized for uniform staining and scanner calibration and observing whether accuracy falls to chance levels.

Figures

Figures reproduced from arXiv: 2605.18522 by Farnaz Kheiri, Masoud Makrehchi, Shahryar Rahnamayan.

Figure 1
Figure 1. Figure 1: The end-to-end workflow for histopathological image acquisition and data preparation. The process begins [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Workflow for the Color Moments feature extraction and classification. An input H&E histopathology patch [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
read the original abstract

In histopathology, human experts primarily rely on color as a means of enhancing contrast to interpret tissue morphology, whereas machine vision models process color as raw statistical information. This distinction raises a fundamental question: to what extent can pixel intensity alone, independent of structural and morphological cues, support cancer classification? To address this question, we systematically evaluated the standalone discriminative power of global color features while deliberately excluding all morphological information. Specifically, we extracted statistical color moments and discretized RGB and HSV color histograms, and assessed their performance across ten diverse experimental settings using classical machine learning classifiers. Our results demonstrate that color features alone can achieve strong performance in binary diagnostic tasks (e.g., benign versus malignant), with classification accuracies reaching up to 89%. This performance is likely attributable to global chromatic shifts associated with malignancy. Importantly, these simple color-based representations consistently outperformed random baselines by a substantial margin, indicating that raw color distributions encode a non-random and diagnostically relevant signal for cancer detection. Consequently, this study suggests that simple, computationally efficient color features can serve as an effective pre-screening tool. By identifying samples with strong chromatic indicators of malignancy, these lightweight models could function as a first-pass triage system, reducing the computational burden on complex deep learning architectures.

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 evaluates the standalone discriminative power of global color features—specifically statistical color moments and discretized RGB/HSV histograms—for binary cancer classification tasks in histopathology. By design, all morphological and structural cues are excluded. The authors report that these features achieve accuracies up to 89% across ten diverse experimental settings with classical machine-learning classifiers, substantially outperforming random baselines, and attribute the signal to global chromatic shifts associated with malignancy. They conclude that such lightweight color representations can serve as an effective pre-screening triage tool.

Significance. If the reported color signal proves robust to acquisition confounds, the result would demonstrate that simple, computationally cheap global chromatic statistics carry substantial diagnostic information independent of morphology. This could support efficient first-pass screening pipelines that reduce load on heavier deep-learning models. The direct empirical evaluation on held-out data and consistent outperformance of random baselines are positive aspects; however, the absence of controls for staining and scanner variation limits the strength of the biological attribution.

major comments (2)
  1. [Abstract] Abstract (paragraph on deliberate exclusion of morphological cues and the ten experimental settings): The central attribution of performance to 'global chromatic shifts associated with malignancy' lacks supporting controls. No color normalization, stain deconvolution, multi-scanner stratification, or batch-effect mitigation is described. Because histopathology datasets routinely exhibit precisely these global chromatic variations from staining intensity and scanner calibration, the observed separation up to 89% is equally consistent with domain shift; this directly undermines the claim that the signal is diagnostically relevant rather than artifactual.
  2. [Abstract] Abstract and results sections: Concrete accuracy figures (up to 89%) and claims of outperformance are presented without dataset sizes, number of images or patients per setting, cross-validation protocol, or error bars. These omissions make it impossible to evaluate whether the reported margins over random baselines are statistically reliable or sensitive to particular data partitions.
minor comments (2)
  1. Clarify the precise list of classifiers employed and any hyper-parameter selection procedure; the current description leaves the experimental pipeline under-specified.
  2. Add a limitations paragraph explicitly discussing the risk of staining/scanner confounds and how future work could address it (e.g., via Macenko normalization or multi-center cohorts).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback on our manuscript. We address each major comment below and have revised the manuscript accordingly to improve clarity and acknowledge limitations where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract (paragraph on deliberate exclusion of morphological cues and the ten experimental settings): The central attribution of performance to 'global chromatic shifts associated with malignancy' lacks supporting controls. No color normalization, stain deconvolution, multi-scanner stratification, or batch-effect mitigation is described. Because histopathology datasets routinely exhibit precisely these global chromatic variations from staining intensity and scanner calibration, the observed separation up to 89% is equally consistent with domain shift; this directly undermines the claim that the signal is diagnostically relevant rather than artifactual.

    Authors: We agree that the absence of explicit color normalization or stain deconvolution represents a limitation in attributing the signal exclusively to biological factors. The ten experimental settings draw from multiple public histopathology datasets that inherently include staining and scanner variations, and the consistent outperformance of random baselines across these settings suggests the color signal is not solely an artifact of any single acquisition protocol. Nevertheless, we have revised the abstract and added a dedicated limitations paragraph in the discussion to explicitly note that future work should incorporate stain normalization (e.g., Macenko or Vahadane methods) and multi-scanner stratification to further isolate biological chromatic shifts from technical domain effects. We maintain that the practical utility for lightweight pre-screening holds regardless of the precise source of the chromatic signal. revision: partial

  2. Referee: [Abstract] Abstract and results sections: Concrete accuracy figures (up to 89%) and claims of outperformance are presented without dataset sizes, number of images or patients per setting, cross-validation protocol, or error bars. These omissions make it impossible to evaluate whether the reported margins over random baselines are statistically reliable or sensitive to particular data partitions.

    Authors: The abstract is constrained by length, but the full manuscript already details the dataset composition (image and patient counts per setting), the 5-fold cross-validation protocol, and reports mean accuracy with standard deviation across folds in both text and figures. To address the referee's concern, we have expanded the abstract with a concise statement on dataset scale and cross-validation, and we ensure all numerical claims in the results section are accompanied by error bars and patient-level stratification details. revision: yes

Circularity Check

0 steps flagged

No circularity; empirical ML evaluation on held-out data

full rationale

The paper reports classification accuracies from training classical ML models on explicitly extracted global color moments and histograms, evaluated across ten experimental settings on held-out data. No equations, derivations, fitted parameters later called predictions, or self-citations appear in the provided text or abstract. Central claims rest on direct empirical performance against random baselines rather than any reduction to inputs by construction. This is a standard self-contained empirical study whose results can be externally verified or falsified on the same datasets.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard image-processing assumptions plus the domain premise that global color shifts track malignancy independently of morphology; no new physical entities are postulated and the few free choices are conventional feature-engineering decisions.

free parameters (2)
  • histogram bin count
    Discretization of RGB and HSV spaces requires choosing the number of bins; this choice affects feature resolution and was not stated as fixed a priori.
  • classifier hyper-parameters
    Classical ML models were used across ten settings; any tuning of regularization or kernel parameters constitutes fitted values.
axioms (1)
  • domain assumption Global chromatic shifts in histopathology images are associated with malignancy rather than staining or imaging artifacts.
    Invoked when the authors attribute performance to 'global chromatic shifts associated with malignancy' without further controls.

pith-pipeline@v0.9.0 · 5760 in / 1414 out tokens · 45963 ms · 2026-05-20T11:16:59.848329+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    Image analysis in histopathology and cytopathol- ogy: from early days to current perspectives.Journal of Imaging, 10(10):252, 2024

    Tibor Mezei, Melinda Kolcsár, András Joó, and Simona Gurzu. Image analysis in histopathology and cytopathol- ogy: from early days to current perspectives.Journal of Imaging, 10(10):252, 2024

    work page 2024

  2. [2]

    Application of histochemical stains in anatomical research: A brief overview of the methods.Translational Research in Anatomy, 35:100294, 2024

    Michał Golberg, Józef Kobos, Edward Clarke, Armand Bajaka, Anna Sm˛ edra, Krzysztof Balawender, Agata Wawrzyniak, Michał Seneczko, Stanisław Orkisz, and Andrzej˙Zytkowski. Application of histochemical stains in anatomical research: A brief overview of the methods.Translational Research in Anatomy, 35:100294, 2024

    work page 2024

  3. [3]

    Histological stains in the past, present, and future.Cureus, 13(10), 2021

    Arslaan Javaeed, Shanza Qamar, Sundus Ali, Mir Ahmad Talha Mustafa, Areeba Nusrat, and Sanniya Khan Ghauri. Histological stains in the past, present, and future.Cureus, 13(10), 2021

    work page 2021

  4. [4]

    Quantitative assessment of h&e staining for pathology: development and clinical evaluation of a novel system.Diagnostic Pathology, 19(1):42, 2024

    Catriona Dunn, David Brettle, Martin Cockroft, Elizabeth Keating, Craig Revie, and Darren Treanor. Quantitative assessment of h&e staining for pathology: development and clinical evaluation of a novel system.Diagnostic Pathology, 19(1):42, 2024

    work page 2024

  5. [5]

    Review the cancer genome atlas (tcga): an immeasurable source of knowledge.Contemporary Oncology/Współczesna Onkologia, 2015(1):68–77, 2015

    Katarzyna Tomczak, Patrycja Czerwi´nska, and Maciej Wiznerowicz. Review the cancer genome atlas (tcga): an immeasurable source of knowledge.Contemporary Oncology/Współczesna Onkologia, 2015(1):68–77, 2015

    work page 2015

  6. [6]

    Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer

    Babak Ehteshami Bejnordi, Mitko Veta, Paul Johannes Van Diest, Bram Van Ginneken, Nico Karssemeijer, Geert Litjens, Jeroen AWM Van Der Laak, Meyke Hermsen, Quirine F Manson, Maschenka Balkenhol, et al. Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer. Jama, 318(22):2199–2210, 2017

    work page 2017

  7. [7]

    Peter Bandi, Oscar Geessink, Quirine Manson, Marcory Van Dijk, Maschenka Balkenhol, Meyke Hermsen, Babak Ehteshami Bejnordi, Byungjae Lee, Kyunghyun Paeng, Aoxiao Zhong, et al. From detection of individual metastases to classification of lymph node status at the patient level: the camelyon17 challenge.IEEE transactions on medical imaging, 38(2):550–560, 2018

    work page 2018

  8. [8]

    Rotation equivariant CNNs for digital pathology

    Bastiaan S Veeling, Jasper Linmans, Jim Winkens, Taco Cohen, and Max Welling. Rotation equivariant CNNs for digital pathology. June 2018

    work page 2018

  9. [9]

    A dataset for breast cancer histopatho- logical image classification.Ieee transactions on biomedical engineering, 63(7):1455–1462, 2015

    Fabio A Spanhol, Luiz S Oliveira, Caroline Petitjean, and Laurent Heutte. A dataset for breast cancer histopatho- logical image classification.Ieee transactions on biomedical engineering, 63(7):1455–1462, 2015

    work page 2015

  10. [10]

    Digital pathology: advantages, limitations and emerging perspectives.Journal of clinical medicine, 9(11):3697, 2020

    Stephan W Jahn, Markus Plass, and Farid Moinfar. Digital pathology: advantages, limitations and emerging perspectives.Journal of clinical medicine, 9(11):3697, 2020

    work page 2020

  11. [11]

    Image analysis and machine learning in digital pathology: Challenges and opportunities.Medical image analysis, 33:170–175, 2016

    Anant Madabhushi and George Lee. Image analysis and machine learning in digital pathology: Challenges and opportunities.Medical image analysis, 33:170–175, 2016

    work page 2016

  12. [12]

    Investigation on potential bias factors in histopathology datasets.Scientific Reports, 15(1):11349, 2025

    Farnaz Kheiri, Shahryar Rahnamayan, Masoud Makrehchi, and Azam Asilian Bidgoli. Investigation on potential bias factors in histopathology datasets.Scientific Reports, 15(1):11349, 2025

    work page 2025

  13. [13]

    Classification and mutation prediction from non–small cell lung cancer histopathology images using deep learning.Nature medicine, 24(10):1559–1567, 2018

    Nicolas Coudray, Paolo Santiago Ocampo, Theodore Sakellaropoulos, Navneet Narula, Matija Snuderl, David Fenyö, Andre L Moreira, Narges Razavian, and Aristotelis Tsirigos. Classification and mutation prediction from non–small cell lung cancer histopathology images using deep learning.Nature medicine, 24(10):1559–1567, 2018

    work page 2018

  14. [14]

    Histopatho- logical image analysis: A review.IEEE reviews in biomedical engineering, 2:147–171, 2009

    Metin N Gurcan, Laura E Boucheron, Ali Can, Anant Madabhushi, Nasir M Rajpoot, and Bulent Yener. Histopatho- logical image analysis: A review.IEEE reviews in biomedical engineering, 2:147–171, 2009

    work page 2009

  15. [15]

    Quantifying the effects of data augmentation and stain color normalization in convolutional neural networks for computational pathology.Medical image analysis, 58:101544, 2019

    David Tellez, Geert Litjens, Péter Bándi, Wouter Bulten, John-Melle Bokhorst, Francesco Ciompi, and Jeroen Van Der Laak. Quantifying the effects of data augmentation and stain color normalization in convolutional neural networks for computational pathology.Medical image analysis, 58:101544, 2019

    work page 2019

  16. [16]

    Feature selection-driven bias deduction in histopathology images: Tackling site-specific influences

    Farnaz Kheiri, Azam Asilian Bidgoli, Masoud Makrehchi, and Shahryar Rahnamayan. Feature selection-driven bias deduction in histopathology images: Tackling site-specific influences. In2024 IEEE Congress on Evolutionary Computation (CEC), pages 1–8. IEEE, 2024

    work page 2024

  17. [17]

    The impact of site-specific digital histology signatures on deep learning model accuracy and bias.Nature communications, 12(1):4423, 2021

    Frederick M Howard, James Dolezal, Sara Kochanny, Jefree Schulte, Heather Chen, Lara Heij, Dezheng Huo, Rita Nanda, Olufunmilayo I Olopade, Jakob N Kather, et al. The impact of site-specific digital histology signatures on deep learning model accuracy and bias.Nature communications, 12(1):4423, 2021

    work page 2021

  18. [18]

    Color transfer between images.IEEE Computer graphics and applications, 21(5):34–41, 2002

    Erik Reinhard, Michael Adhikhmin, Bruce Gooch, and Peter Shirley. Color transfer between images.IEEE Computer graphics and applications, 21(5):34–41, 2002. 10 Beyond Morphology: Quantifying the Diagnostic Power of Color Features in Cancer Classification

    work page 2002

  19. [19]

    A method for normalizing histology slides for quantitative analysis

    Marc Macenko, Marc Niethammer, James S Marron, David Borland, John T Woosley, Xiaojun Guan, Charles Schmitt, and Nancy E Thomas. A method for normalizing histology slides for quantitative analysis. In2009 IEEE international symposium on biomedical imaging: from nano to macro, pages 1107–1110. IEEE, 2009

    work page 2009

  20. [20]

    Structure-preserving color normalization and sparse stain separation for histological images.IEEE transactions on medical imaging, 35(8):1962–1971, 2016

    Abhishek Vahadane, Tingying Peng, Amit Sethi, Shadi Albarqouni, Lichao Wang, Maximilian Baust, Katja Steiger, Anna Melissa Schlitter, Irene Esposito, and Nassir Navab. Structure-preserving color normalization and sparse stain separation for histological images.IEEE transactions on medical imaging, 35(8):1962–1971, 2016

    work page 1962

  21. [21]

    Stain normalization of histopathology images using generative adversarial networks

    Farhad Ghazvinian Zanjani, Svitlana Zinger, Babak Ehteshami Bejnordi, Jeroen AWM Van Der Laak, and Peter HN de With. Stain normalization of histopathology images using generative adversarial networks. In2018 IEEE 15th International symposium on biomedical imaging (ISBI 2018), pages 573–577. IEEE, 2018

    work page 2018

  22. [22]

    Staingan: Stain style transfer for digital histological images

    M Tarek Shaban, Christoph Baur, Nassir Navab, and Shadi Albarqouni. Staingan: Stain style transfer for digital histological images. In2019 Ieee 16th international symposium on biomedical imaging (Isbi 2019), pages 953–956. IEEE, 2019

    work page 2019

  23. [23]

    Stain san: simultaneous augmentation and normalization for histopathology images.Journal of Medical Imaging, 11(4):044006–044006, 2024

    Taebin Kim, Yao Li, Benjamin C Calhoun, Aatish Thennavan, Lisa A Carey, W Fraser Symmans, Melissa A Troester, Charles M Perou, and JS Marron. Stain san: simultaneous augmentation and normalization for histopathology images.Journal of Medical Imaging, 11(4):044006–044006, 2024

    work page 2024

  24. [24]

    Semi-supervised adversarial learning for stain normalisation in histopathology images

    Cong Cong, Sidong Liu, Antonio Di Ieva, Maurice Pagnucco, Shlomo Berkovsky, and Yang Song. Semi-supervised adversarial learning for stain normalisation in histopathology images. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 581–591. Springer, 2021

    work page 2021

  25. [25]

    Stain normalization using sparse autoencoders (stanosa): application to digital pathology.Computerized Medical Imaging and Graphics, 57:50–61, 2017

    Andrew Janowczyk, Ajay Basavanhally, and Anant Madabhushi. Stain normalization using sparse autoencoders (stanosa): application to digital pathology.Computerized Medical Imaging and Graphics, 57:50–61, 2017

    work page 2017

  26. [26]

    Data-driven color augmentation for h&e stained images in computational pathology.Journal of Pathology Informatics, 14:100183, 2023

    Niccolò Marini, Sebastian Otalora, Marek Wodzinski, Selene Tomassini, Aldo Franco Dragoni, Stephane Marchand-Maillet, Juan Pedro Dominguez Morales, Lourdes Duran-Lopez, Simona Vatrano, Henning Müller, et al. Data-driven color augmentation for h&e stained images in computational pathology.Journal of Pathology Informatics, 14:100183, 2023

    work page 2023

  27. [27]

    Impact of color augmentation and tissue type in deep learning for hematoxylin and eosin image super resolution.Journal of Pathology Informatics, 13:100148, 2022

    Cyrus Manuel, Philip Zehnder, Sertan Kaya, Ruth Sullivan, and Fangyao Hu. Impact of color augmentation and tissue type in deep learning for hematoxylin and eosin image super resolution.Journal of Pathology Informatics, 13:100148, 2022

    work page 2022

  28. [28]

    Multifeature prostate cancer diagnosis and gleason grading of histological images.IEEE transactions on medical imaging, 26(10):1366–1378, 2007

    Ali Tabesh, Mikhail Teverovskiy, Ho-Yuen Pang, Vinay P Kumar, David Verbel, Angeliki Kotsianti, and Olivier Saidi. Multifeature prostate cancer diagnosis and gleason grading of histological images.IEEE transactions on medical imaging, 26(10):1366–1378, 2007

    work page 2007

  29. [29]

    MA Aswathy and M Jagannath. An svm approach towards breast cancer classification from h&e-stained histopathology images based on integrated features.Medical & biological engineering & computing, 59(9):1773– 1783, 2021

    work page 2021

  30. [30]

    Analysis of feature extraction and classification methods on histopathological images for diagnosing invasive ductal carcinoma

    Elvira Sukma Wahyuni, Vera Giyaning Tiyas, and Suatmi Murnani. Analysis of feature extraction and classification methods on histopathological images for diagnosing invasive ductal carcinoma. In2022 9th International Conference on Information Technology, Computer, and Electrical Engineering (ICITACEE), pages 263–268. IEEE, 2022

    work page 2022

  31. [31]

    Kushangi Atrey, Bikesh Kumar Singh, and Narendra Kuber Bodhey. Multi-feature classification of breast cancer histopathology images: An experimental investigation in machine learning and deep learning paradigm.Brazilian Archives of Biology and Technology, 66:e23220297, 2023

    work page 2023

  32. [32]

    Agaba Ameh Joseph, Mohammed Abdullahi, Sahalu Balarabe Junaidu, Hayatu Hassan Ibrahim, and Haruna Chiroma. Improved multi-classification of breast cancer histopathological images using handcrafted features and deep neural network (dense layer).Intelligent Systems with Applications, 14:200066, 2022

    work page 2022

  33. [33]

    A hybrid approach for classification breast cancer histopathology images.Frontiers in scientific research and technology, 3(1):1–10, 2022

    Amr H Hassan, Mohammed ElSyed Wahed, Mohammed Ail Atiea, and Mohammed Saleh Metwally. A hybrid approach for classification breast cancer histopathology images.Frontiers in scientific research and technology, 3(1):1–10, 2022

    work page 2022

  34. [34]

    Medmnist v2-a large-scale lightweight benchmark for 2d and 3d biomedical image classification.Scientific Data, 10(1):41, 2023

    Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, and Bingbing Ni. Medmnist v2-a large-scale lightweight benchmark for 2d and 3d biomedical image classification.Scientific Data, 10(1):41, 2023

    work page 2023

  35. [35]

    Lunghist700: A dataset of histological images for deep learning in pulmonary pathology.Scientific Data, 11(1):1088, 2024

    Jorge Diosdado, Pere Gilabert, Santi Seguí, and Henar Borrego. Lunghist700: A dataset of histological images for deep learning in pulmonary pathology.Scientific Data, 11(1):1088, 2024

    work page 2024

  36. [36]

    Angel Cruz-Roa, Ajay Basavanhally, Fabio González, Hannah Gilmore, Michael Feldman, Shridar Ganesan, Natalie Shih, John Tomaszewski, and Anant Madabhushi. Automatic detection of invasive ductal carcinoma in 11 Beyond Morphology: Quantifying the Diagnostic Power of Color Features in Cancer Classification whole slide images with convolutional neural network...

    work page 2014

  37. [37]

    Similarity of color images

    Markus Andreas Stricker and Markus Orengo. Similarity of color images. InStorage and retrieval for image and video databases III, volume 2420, pages 381–392. SPiE, 1995

    work page 1995

  38. [38]

    K-nearest neighbor.Scholarpedia, 4(2):1883, 2009

    Leif E Peterson. K-nearest neighbor.Scholarpedia, 4(2):1883, 2009

    work page 2009

  39. [39]

    Hearst, Susan T Dumais, Edgar Osuna, John Platt, and Bernhard Scholkopf

    Marti A. Hearst, Susan T Dumais, Edgar Osuna, John Platt, and Bernhard Scholkopf. Support vector machines. IEEE Intelligent Systems and their applications, 13(4):18–28, 1998

    work page 1998

  40. [40]

    Random forests.Machine learning, 45(1):5–32, 2001

    Leo Breiman. Random forests.Machine learning, 45(1):5–32, 2001

    work page 2001

  41. [41]

    Xgboost: extreme gradient boosting.R package version 0.4-2, 1(4):1–4, 2015

    Tianqi Chen, Tong He, Michael Benesty, Vadim Khotilovich, Yuan Tang, Hyunsu Cho, Kailong Chen, Rory Mitchell, Ignacio Cano, Tianyi Zhou, et al. Xgboost: extreme gradient boosting.R package version 0.4-2, 1(4):1–4, 2015. 12

    work page 2015