Enhancing Diagnostic Precision in Gastric Bleeding through Automated Lesion Segmentation: A Deep DuS-KFCM Approach

Feng Tien; Huafeng Qin; Juntao Gao; Mingkun Xu; Qun Song; Shirley Siu; Simon Fong; Wei Luo; Xian-Xian Liu; Yuanyuan Wei

REVIEW 5 major objections 5 minor 44 references

Enhancing Diagnostic Precision in Gastric Bleeding through Automated Lesion Segmentation: A Deep DuS-KFCM Approach

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a two-stage pipeline—fuzzy clustering pre-segmentation followed by DeepLabv3+ refinement—segments gastric bleeding lesions in endoscopic images with 87.95% accuracy and 96.33% specificity on the gastric bleeding and…

desk verdict Short version: plausible clinical application, but the missing split, wrong metric formulas, and inconsistent results make the headline claim untestable as written. read the letter →

arxiv 2411.14385 v2 pith:LRMYOAG4 submitted 2024-11-21 eess.IV cs.CV

Xian-Xian Liu , Mingkun Xu , Yuanyuan Wei , Huafeng Qin , Qun Song , Simon Fong , Feng Tien , Wei Luo

show 3 more authors

Juntao Gao Zhihua Zhang Shirley Siu

This is my paper · ORCID

classification eess.IVcs.CV

keywords gastricbleedingsegmentationfuzzyC-meansDeepLabv3+ResNet50GLCMtexturefeaturescoarse-to-fineendoscopicimageanalysishybridneuro-fuzzymodel

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to show that gastric bleeding lesions in endoscopic images can be segmented more accurately by combining fuzzy clustering with a deep refinement network than by either approach alone. It introduces Deep DuS-KFCM, a two-stage pipeline in which a spatial kernelized fuzzy C-means variant (DuS-KFCM) uses color and texture features to produce a coarse lesion mask, and DeepLabv3+ with a ResNet50 backbone refines that mask into the final boundary. On the gastric bleeding and gastric red-spot datasets, the authors report 87.95% accuracy and 96.33% specificity, with higher precision and IoU than FKM, GMM, FCM, and other compared methods. If the result holds, the method would give clinicians a fast, reproducible way to delineate bleeding regions from endoscopy frames without manual tracing.

What carries the argument

The machinery is a two-stage coarse-to-fine segmentation pipeline. Stage one is DuS-KFCM, a Dual Spatial Kernelized Constrained Fuzzy C-Means algorithm that fuses RGB color features with 22 GLCM texture statistics, selects discriminative features, and clusters pixels into coarse lesion and background regions. Stage two is DeepLabv3+ with a ResNet50 encoder: atrous convolutions and an atrous spatial pyramid pooling module gather multi-scale context, and a decoder fuses that context with low-level spatial features to refine the lesion boundary. The paper's argument is that the fuzzy stage handles the intensity-overlap problem and the deep stage recovers boundary precision, so the two stages carry different parts of the segmentation task.

What would settle it

Train the DeepLabv3+ refinement stage on the original images alone (no DuS-KFCM masks) using the same train/test split, and compare it on a held-out test set whose images were never used to generate clustering masks; if accuracy and specificity do not improve materially, the coarse-to-fine contribution is not supported. A second check is to recompute the clustering masks on the test set independently and verify the refinement stage is not simply reproducing training-set masks.

Watch

Extended reading notes

Core claim

The central claim is that the intensity overlap between bleeding tissue and adjacent gastric structures can be resolved by a coarse-to-fine strategy rather than by a single segmentation model. In the paper's telling, DuS-KFCM first separates likely lesion pixels using fuzzy memberships built from spatial, color, and GLCM texture information; DeepLabv3+ then takes those coarse masks and the original image and refines them into sharp, noise-resistant lesion boundaries. The authors report that the combined model reaches 87.95% accuracy and 96.33% specificity on the gastric bleeding and gastric red-spot datasets, and they present visual and quantitative comparisons against FKM, GMM, and FCM showing better boundary detection. They interpret this as evidence that the hybrid neuro-fuzzy design is well suited to subtle bleeding symptoms.

Load-bearing premise

The load-bearing premise is that the public pixel-level annotations are trustworthy and that feeding each image's own fuzzy-clustering mask into the deep refinement stage during training does not leak information or bias the measured improvement; the paper reports no ablation or independent validation that would rule this out.

Editorial extensions

If this is right

Automated segmentation could give clinicians a highlighted lesion boundary from an endoscopy frame with less manual effort and less inter-observer variability.
The reported specificity of 96.33% implies few false-positive detections, which matters for screening decisions that should avoid unnecessary follow-up.
The method's applicability to both gastric bleeding and gastric red-spot images suggests the coarse-to-fine design is not tied to one lesion appearance.
Combining fuzzy clustering with a deep decoder may reduce sensitivity to noise and to intensity overlap between lesion and healthy tissue, the problem that motivated the work.

Reading between the lines

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

The soft membership maps produced by fuzzy clustering need no ground truth, so they could be reused as pseudo-labels or attention priors for semi-supervised segmentation of unlabelled endoscopy video.
The same coarse-to-fine recipe—fuzzy pre-segmentation plus atrous-convolution refinement—could transfer to other endoscopic lesion types with similar intensity overlap, but the transfer should be validated on new datasets.
A direct ablation comparing the full model to DeepLabv3+ trained on raw images would separate the clustering stage's contribution from the deep stage's, giving a cleaner estimate of where the accuracy gain comes from.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Short version: this paper is not ready for the literature. The clinical problem is real, and the rough idea of combining fuzzy clustering with DeepLabv3+ is not crazy, but the manuscript does not supply the evidence needed to evaluate it: no split, no code, no ablation, and the metric definitions in Section 3 are wrong. The headline numbers cannot be interpreted.

What is actually new: the specific combination of SKFCM with GLCM features and DeepLabv3+ refinement applied to gastric bleeding and red spots from Kvasir-SEG. That is an application of known components. There is no new theory, no new dataset, no clinical validation.

The paper does a few things well: it works with a public dataset, it reports visual comparisons against FKM/GMM/FCM, and it situates the problem in a real clinical context. The qualitative figures suggest the method can find lesion boundaries on some images. That is worth acknowledging.

The soft spots are structural. First, the protocol in Section 2.2 says DuS-KFCM pre-segmentation is computed on each training image and then used to train DeepLabv3+, and the same clustering is applied to test images before refinement. But the paper never states how training and test sets were formed, whether images from the same endoscopy session appear in both, or whether any feature selection or hyperparameters were tuned on test data. So the reported 87.95% accuracy and 96.33% specificity may simply reflect the per-image prior from clustering, not a learned refinement. There is no ablation separating DeepLabv3+ alone from Deep DuS-KFCM.

Second, the metrics are defined incorrectly. Eq. (2) gives sensitivity = TP/(TN+FN); Eq. (8) gives specificity = TN/(TN+FN). Those are not sensitivity and specificity. The abstract says 96.33% specificity but the results section never reports that number; instead it says 98% accuracy and then later 87.95% accuracy. The numbers do not cohere.

Third, no code, no hyperparameters, no error bars, and no held-out evaluation. The baselines are classic methods, not recent deep segmentation networks, so the comparison is weak. The stress-test concern is accurate: the central quantitative claim is not testable as written.

Who is this for? Someone doing a survey of fuzzy-clustering-plus-deep approaches might note it, but the paper will not be citable until the experimental protocol and metrics are fixed. A serious editor should probably send it back for major revision at best, but honestly the current form is closer to a desk reject: the missing split and broken metrics are not stylistic, they are load-bearing. I would not spend a referee slot on it as is.

Referee Report

5 major / 5 minor

Summary. This manuscript proposes Deep DuS-KFCM, a coarse-to-fine segmentation pipeline for gastric bleeding lesions in endoscopic images. The method first applies a fuzzy clustering technique (DuS-KFCM) with GLCM and color features to obtain a coarse lesion mask, then refines that mask with a DeepLabv3+ (ResNet50) network. The paper claims high accuracy (87.95%) and specificity (96.33%) on public gastric bleeding and gastric red spot datasets and states that the method outperforms contemporary segmentation approaches. The central contribution, as presented, is a hybrid neuro-fuzzy segmentation architecture.

Significance. If the reported results were reproducible and the metric definitions correct, the proposed combination of fuzzy clustering priors with a deep refinement network could be a practically relevant contribution to endoscopic lesion segmentation. However, the manuscript as written does not support this claim: the sensitivity and specificity formulas are mathematically wrong, the experimental protocol permits information leakage between the clustering and the deep refinement stages, the reported numbers are inconsistent across sections, and no reproducibility details (splits, hyperparameters, error bars) are provided. The clinical motivation is sound, but the evidence presented is not sufficient to assess the method's actual performance.

major comments (5)

[Section 3, Eqs. (2) and (8)] The definitions of sensitivity and specificity are incorrect. Eq. (2) defines sensitivity as TP/(TN+FN), which omits true positives from the denominator; the standard definition is TP/(TP+FN). Eq. (8) defines specificity as TN/(TN+FN), using false negatives instead of false positives; the standard definition is TN/(TN+FP). Because all reported performance values in Section 4 rely on these metrics, the quantitative results are not interpretable and any comparison to literature values is invalid.
[Section 2.2 and Figure 1(c)] The experimental protocol applies the DuS-KFCM clustering algorithm to the same images that are later used for training and testing the deep network, and the test-time pipeline feeds the clustering mask into the final classification stage. The paper does not report an ablation separating DeepLabv3+ alone from the full pipeline, nor does it provide any held-out split or cross-validation details. Consequently, the reported accuracy of 87.95% and specificity of 96.33% cannot be attributed to the proposed refinement; they may be driven by the strong per-image prior supplied by the clustering step.
[Section 4 and abstract] The quantitative claims are internally inconsistent. The abstract reports an accuracy of 87.95% and a specificity of 96.33%; Section 4 reports 'a remarkable accuracy of 98%' in endoscopy gastric images; and the conclusion reports accuracy values of 87.95% and 79.72% and precision values of 86.69% and 62.82%. The specificity value from the abstract does not appear anywhere in the results section. These contradictions make the central quantitative claim unverifiable as written.
[Section 4] The experimental setup is not reproducible. The paper specifies no train/validation/test split sizes, no hyperparameters for DeepLabv3+ training (optimizer, learning rate, batch size, number of epochs), no details on the CFS feature selection procedure, no values for the fuzzy clustering parameters (fuzzifier m, number of clusters), and no error bars or statistical significance tests. Without these details, the reported 'unprecedented accuracy' cannot be independently verified or reproduced.
[Sections 1 and 4] The claim that the method outperforms 'contemporary segmentation methods' is not supported by the comparisons presented. The baselines used are FKM, GMM, and FCM, which are classical clustering methods, not modern deep-learning segmentation approaches. The paper makes no comparison to U-Net, DeepLab variants, or other published segmentation methods on the same datasets, and it does not cite or benchmark against known Kvasir-SEG results. The superiority claim is therefore unsubstantiated.

minor comments (5)

[Throughout] The method name is inconsistently written as both 'Dus-KFCM' and 'DuS-KFCM'; please standardize the spelling in the abstract, main text, and figures.
[Eq. (1)] Equation (1) uses 'Si ∩ Gi' without defining how the intersection is computed across multiple clusters or how overlapping regions are handled; please provide a precise mathematical definition.
[Section 3] The text states that specificity is 'expressed in (9)' but the equation itself is numbered (8); the cross-reference is incorrect.
[References] Reference [38] is incomplete, giving only a short author list and a DOI to an arXiv preprint without a title or full author list; reference [6] cites a general cancer statistics webpage for a specific survival-rate claim, which should be supported by an epidemiological source.
[Figure 3 caption] The caption refers to 'Synthetic Pseudo-color Images,' but the relationship between these synthetic images and the real gastric bleeding/red spot datasets is never explained in the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation identified; reported issues are metric/protocol errors, not self-referential reductions.

full rationale

The Deep DuS-KFCM pipeline is a two-stage system: an unsupervised fuzzy clustering (DuS-KFCM) produces candidate lesion masks, and a DeepLabv3+ network refines them. The final segmentation is scored against public ground-truth masks. No equation in the paper defines the final output as a function of the same ground-truth labels used for fitting, nor does any fitted parameter reappear as the reported prediction. The training description (Section 2.2) says the clustering segmentations are used to train the deep model, but the reported accuracy is measured against ground truth, not against the clustering masks, so the central result is not equal to its own input by construction. The paper contains no load-bearing self-citations: reference [23], cited for the DuS-KFCM name, is by different authors. The serious problems are experimental: no train/test split details, no ablation isolating the deep refinement, incorrect metric formulas (Eqs. 2 and 8), and the abstract's 96.33% specificity is absent from the results. These are correctness and reproducibility flaws, not instances of a derivation reducing to its assumptions. Therefore, no circular step can be specifically exhibited, and the circularity score is 0.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new physical entities, but the central performance claim rests on several unreported free parameters and two unvalidated domain assumptions, which are listed above.

free parameters (5)

Fuzzy clustering fuzzifier m
Standard FCM-family algorithms require a fuzzifier m; the paper never reports its value, yet the coarse segmentation depends on it.
Number of clusters k
Segmentation accuracy in Eq. (1) sums over k clusters; k is not specified for GB or GRS images.
GLCM feature subset
22 GLCM statistics are computed and then 'optimized by CFS'; the selected features and threshold are not reported.
DeepLabv3+ training hyperparameters
No learning rate, epochs, batch size, optimizer, or loss weights are given.
Spatial intensity profile parameters
The spatial intensity profiles used in stage 1 are described qualitatively, with no kernel width or neighborhood size.

assumptions (3)

domain assumption Ground-truth masks in the public Kvasir-SEG and red spot images are accurate pixel-level labels.
All reported metrics compare predictions to these masks; any label error propagates into every score. Invoked in Section 4 and Figure 3.
ad hoc to paper The DuS-KFCM coarse masks are suitable supervision for the deep refinement network.
Section 2.2 says the clustering output is used to train the model, but no evidence is given that clustering noise does not bias the deep network.
standard math Standard fuzzy clustering and DeepLabv3+ components behave as in their original publications.
The pipeline assumes these off-the-shelf components work as published, but original references are only partially cited.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Enhancing Diagnostic Precision in Gastric Bleeding through Automated Lesion Segmentation: A Deep DuS-KFCM Approach." pith.science (2026). https://pith.science/paper/LRMYOAG4

@misc{pith2026241114385,
  author       = {Pith},
  title        = {Pith review of: Enhancing Diagnostic Precision in Gastric Bleeding through Automated Lesion Segmentation: A Deep DuS-KFCM Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LRMYOAG4}},
  note         = {Machine review of arXiv:2411.14385}
}

read the original abstract

Timely and precise classification and segmentation of gastric bleeding in endoscopic imagery are pivotal for the rapid diagnosis and intervention of gastric complications, which is critical in life-saving medical procedures. Traditional methods grapple with the challenge posed by the indistinguishable intensity values of bleeding tissues adjacent to other gastric structures. Our study seeks to revolutionize this domain by introducing a novel deep learning model, the Dual Spatial Kernelized Constrained Fuzzy C-Means (Deep DuS-KFCM) clustering algorithm. This Hybrid Neuro-Fuzzy system synergizes Neural Networks with Fuzzy Logic to offer a highly precise and efficient identification of bleeding regions. Implementing a two-fold coarse-to-fine strategy for segmentation, this model initially employs the Spatial Kernelized Fuzzy C-Means (SKFCM) algorithm enhanced with spatial intensity profiles and subsequently harnesses the state-of-the-art DeepLabv3+ with ResNet50 architecture to refine the segmentation output. Through extensive experiments across mainstream gastric bleeding and red spots datasets, our Deep DuS-KFCM model demonstrated unprecedented accuracy rates of 87.95%, coupled with a specificity of 96.33%, outperforming contemporary segmentation methods. The findings underscore the model's robustness against noise and its outstanding segmentation capabilities, particularly for identifying subtle bleeding symptoms, thereby presenting a significant leap forward in medical image processing.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 41 canonical work pages

[1]

Bertram, C. A. et al. Computer-assisted mitotic count using a deep learning-based algorithm improves interobserver reproducibility and accuracy. Vet. Pathol. 59, 211–226 (2022)

work page 2022
[2]

Huang, H.-Y. et al. Classification of skin cancer using novel hyperspectral imaging engineering via YOLOv5. J. Clin. Med. 12, 1134 (2023)

work page 2023
[3]

& Jemal, A

Siegel, R., Ma, J., Zou, Z. & Jemal, A. Cancer statistics. CA Cancer J. Clin. 64, 9–29 (2014)

work page 2014
[4]

https://www.cancer.org/cancer/pancreatic-cancer/about/key- statistics.html, (2018)

Cancer statistics. https://www.cancer.org/cancer/pancreatic-cancer/about/key- statistics.html, (2018)

work page 2018
[5]

Sung, H. et al. Global cancer statistics 2020: Globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin. 71, 209–249 (2021)

work page 2021
[6]

& Maaten, L

Misra, I. & Maaten, L. V. D. Self-Supervised Learning of Pretext-Invariant Rep- resentations. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 6707–6717 (2020)

work page 2020
[7]

Lambin, P. et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur. J. Cancer 48, 441–446 (2012)

work page 2012
[8]

Eilaghi, A. et al. CT texture features are associated with overall survival in pan- creatic ductal adenocarcinoma – a quantitative analysis. BMC Med. Imaging 17, 38 (2017)

work page 2017

Show all 44 references

[9]

& Waston, D

Al-Kadi, O. & Waston, D. Texture analysis of aggressive and nonaggressive lung tumor CE CT images. IEEE Trans. Biomed. Eng. 55, 1822–1830 (2008). 12

2008
[10]

M., Shanmugam, K

Haralick, R. M., Shanmugam, K. & Dinstein, I. Textural features for image classification. IEEE Trans. Syst. Man Cybern. 3, 610–621 (1973)

1973
[11]

& King, R

Amadasun, M. & King, R. Textural features corresponding to textural properties. IEEE Trans. Syst., Man, Cybern. 19, 1264–1274 (1989)

1989
[12]

Texture analysis using gray level run lengths

Galloway, M. Texture analysis using gray level run lengths. Computer Graph. Image Process. 4, 172–179 (1975)

1975
[13]

Thibault, G. et al. Texture indexes and gray level size zone matrix. Application to cell nuclei classification. In10th International Conference On Pattern Recognition and Information Processing. 140–145, (Minsk, Belarus, 2009)

2009
[14]

Artificial intelligence for multimodal data integration in oncology

Lipkova, J. Artificial intelligence for multimodal data integration in oncology. Cancer Cell 40, 1095–1110 (2022)

2022
[15]

Bashir, R. M. S., Qaiser, T., Raza, S. E. A. & Rajpoot, N. M. HydraMix-Net: a deep multi-task semi-supervised learning approach for cell detection and clas- sification. In Interpretable and Annotation-Efficient Learning for Medical Image Computing (eds Cardoso, J. et al.) 164–...

2020 doi
[16]

& Ricardo, C

Gustavo, C., Iuri, A. & Ricardo, C. A self-organizing map-based method for multi-label classification. IJCNN, 4291–4298 (2017)

2017
[17]

Deep feature transfer learning in combination with traditional fea- tures predicts survival among patients with lung adenocarcinoma

Paul, R. Deep feature transfer learning in combination with traditional fea- tures predicts survival among patients with lung adenocarcinoma. Tomography 2, 388–395 (2016)

2016
[18]

Liu, L. et al. Application of texture analysis based on apparent diffusion coefficient maps in discriminating different stages of rectal cancer. J. Magn. Reson Imaging 45, 1798–1808 (2017)

2017
[19]

R., Skamene, S

Valli` eres, M., Freeman, C. R., Skamene, S. R. & El Naqa, I. A radiomics model from joint FDG-PET and MRI texture features for the prediction of lung metas- tases in soft-tissue sarcomas of the extremities. Phys. Med. Biol. 60, 5471–5496 (2015)

2015
[20]

Doubeni, C. A. Precision screening for colorectal cancer: promise and challenges. Ann. Intern. Med. 163, 390–391 (2015)

2015
[21]

Hao, Q. et al. Fusing multiple deep models for in vivo human brain hyperspectral image classification to identify glioblastoma tumor. IEEE Trans. Instrum. Meas. 70, 4007314 (2021)

2021
[22]

& Qian, W

Sun, W., Zheng, B. & Qian, W. Automatic feature learning using multichannel ROI based on deep structured algorithms for computerized lung cancer diagnosis. 13 Computers Biol. Med. 89, 530–539 (2017)

2017
[23]

Self-supervised learning for medical image classification: a system- atic review and implementation guidelines

Huang, S.-C. Self-supervised learning for medical image classification: a system- atic review and implementation guidelines. npj Digital Med. 6, 74 (2023)

2023
[24]

Wang, C. et al. Non-invasive measurement using deep learning algorithm based on multi-source features fusion to predict PD-L1 expression and survival in NSCLC. Front. Immunol. 13, 828560 (2022)

2022
[25]

Noorbakhsh, J. et al. Deep learning-based cross-classifications reveal conserved spatial behaviors within tumor histological images. Nat. Commun. 11, 6367 (2020)

2020
[26]

Tomita, N. et al. Attention-Based Deep Neural Networks for Detection of Can- cerous and Precancerous Esophagus Tissue on Histopathological Slides. JAMA Netw. Open 2, e1914645 (2019)

2019
[27]

Xu, H. L. et al. Artificial intelligence performance in image-based ovarian can- cer identification: A systematic review and meta-analysis. EClinicalMedicine 53, 101662 (2022)

2022
[28]

Coudray, N. et al. Classification and mutation prediction from non–small cell lung cancer histopathology images using deep learning. Nat. Med. 24, 1559–1567 (2018)

2018
[29]

Wang, S. et al. ConvPath: a software tool for lung adenocarcinoma digital patho- logical image analysis aided by a convolutional neural network. EBioMedicine 50, 103–110 (2019)

2019
[30]

Girschik, J. et al. Precision in setting cancer prevention priorities: synthesis of data, literature, and expert opinion. Front. Public Health 5, 125 (2017)

2017
[31]

Florimbi, G. et al. Accelerating the K-nearest neighbors filtering algorithm to optimize the real-time classification of human brain tumor in hyperspectral images. Sensors 18, 2314 (2018)

2018
[32]

Wang, Y. et al. PARP inhibitors in gastric cancer: beacon of hope. J. Exp. Clin. Cancer Res. 40, 211 (2021)

2021
[33]

& Sun, J

He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In Proc. IEEE conference on computer vision and pattern recognition , 770–778 (2016)

2016
[34]

Wang, X. et al. A prognostic and predictive computational pathology image sig- nature for added benefit of adjuvant chemotherapy in early stage non-small-cell lung cancer. EBioMedicine 69, 103481 (2021). 14

2021
[35]

Choi, S. et al. Deep learning model improves tumor-infiltrating lymphocyte eval- uation and therapeutic response prediction in breast cancer. NPJ Breast Cancer 9, 71 (2023)

2023
[36]

A threshold selection method from gray-level histograms

Otsu, N. A threshold selection method from gray-level histograms. IEEE Trans- actions on Systems, Man, and Cybernetics 9, 62–66 (1979)

1979
[37]

Dietterich, T. G. Ensemble methods in machine learning. in Multiple Classifier Systems. 1–15 (Springer Nature, 2000)

2000
[38]

Wang, H. et al. In Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Preprint at https://doi.org/10.48550/arXiv.1910.01279 (2019)

2019 doi arXiv
[39]

& Sun, J

He K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 770-778 (2016)

2016
[40]

& Ishikawa, S

Komura, D. & Ishikawa, S. Machine learning methods for histopathological image analysis. Comput Struct. Biotechnol. J. 16, 34–42 (2018)

2018
[41]

Camps-Valls, G. et al. Kernel-based methods for hyperspectral image classifica- tion. IEEE Trans. Geosci. Remote Sens. 43, 1351–1362 (2005). Urbanos, G. et al. Supervised Machine

2005
[42]

Sensors 2021 21, 3827 (2021)

Learning Methods and Hyperspectral Imaging Techniques Jointly Applied for Brain Cancer Classification. Sensors 2021 21, 3827 (2021)

2021
[43]

& Adam, H

Chen, L.-C., Papandreou, G., Shroff, F. & Adam, H. Rethink- ing atrous convolution for semantic image segmentation. arXiv https://doi.org/10.48550/arXiv.1706.05587 (2017)

2017 doi arXiv
[44]

Y., Sater, H

Lu, M. Y., Sater, H. A. & Mahmood, F. Multiplex computational pathology for treatment response predication. Cancer Cell 39, 1053–1055 (2021). 15

2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Bertram, C. A. et al. Computer-assisted mitotic count using a deep learning-based algorithm improves interobserver reproducibility and accuracy. Vet. Pathol. 59, 211–226 (2022)

work page 2022

[2] [2]

Huang, H.-Y. et al. Classification of skin cancer using novel hyperspectral imaging engineering via YOLOv5. J. Clin. Med. 12, 1134 (2023)

work page 2023

[3] [3]

& Jemal, A

Siegel, R., Ma, J., Zou, Z. & Jemal, A. Cancer statistics. CA Cancer J. Clin. 64, 9–29 (2014)

work page 2014

[4] [4]

https://www.cancer.org/cancer/pancreatic-cancer/about/key- statistics.html, (2018)

Cancer statistics. https://www.cancer.org/cancer/pancreatic-cancer/about/key- statistics.html, (2018)

work page 2018

[5] [5]

Sung, H. et al. Global cancer statistics 2020: Globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin. 71, 209–249 (2021)

work page 2021

[6] [6]

& Maaten, L

Misra, I. & Maaten, L. V. D. Self-Supervised Learning of Pretext-Invariant Rep- resentations. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 6707–6717 (2020)

work page 2020

[7] [7]

Lambin, P. et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur. J. Cancer 48, 441–446 (2012)

work page 2012

[8] [8]

Eilaghi, A. et al. CT texture features are associated with overall survival in pan- creatic ductal adenocarcinoma – a quantitative analysis. BMC Med. Imaging 17, 38 (2017)

work page 2017