Pith. sign in

REVIEW 4 major objections 4 minor 25 references

This paper shows that combining a patient's tumor pathology slide with their preoperative CT scan predicts kidney cancer recurrence better than either image type alone, and nearly as well as the standard clinical risk score.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

Multimodal fusion of CT and pathology images improves recurrence risk prediction in kidney cancer, with the best model approaching the clinical Leibovich score.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Careful feasibility benchmark for CT+WSI fusion in ccRCC recurrence, but the fusion-advantage claim rests on differences within one standard deviation without significance testing. the 4 major comments →

arxiv 2508.21581 v1 pith:HTA2FIEU submitted 2025-08-29 cs.CV

Integrating Pathology and CT Imaging for Personalized Recurrence Risk Prediction in Renal Cancer

Dani\"el Boeke , Cedrik Blommestijn , Rebecca N. Wray , Kalina Chupetlovska , Shangqi Gao , Zeyu Gao , Regina G. H. Beets-Tan , Mireia Crispin-Ortuzar
show 3 more authors
James O. Jones Wilson Silva Ines P. Machado
This is my paper
classification cs.CV
keywords clear cell renal cell carcinomarecurrence risk predictionmultimodal imaging fusionwhole-slide imagesCT imagingCox survival modelfoundation modelsLeibovich score
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 tries to establish that recurrence risk in clear cell renal cell carcinoma can be predicted from routine imaging alone, and that combining the two image types—the postoperative tumor slide and the preoperative CT—gives more accurate risk rankings than either modality on its own. If true, this matters because current practice relies on a coarse clinical score that assigns many patients the same risk and ignores imaging altogether. The authors report that pathology-based models consistently beat CT-only models, that joining the two feature vectors before prediction (intermediate fusion) beat both late fusion and unimodal models, and that the best fused model's concordance index (0.775) approaches the adjusted Leibovich score (0.805). They also find that the clinical score's apparent advantage shrinks when its tied scores are broken randomly, suggesting discretization inflates its apparent individual-level accuracy.

Core claim

The central claim is that whole-slide histopathology and preoperative CT carry complementary prognostic information for clear cell renal cell carcinoma recurrence, and that a simple fusion of their learned representations extracts it. In the authors' experiments, frozen pathology foundation-model embeddings consistently produced higher C-index values than fine-tuned CT encoders when used alone, so pathology dominates the signal; but every WSI–CT combination that concatenated the two feature vectors before the survival head matched or exceeded its WSI-only baseline, with the best result from TITAN–CONCH plus ResNet-18 (C-index 0.775 ± 0.044). The adjusted Leibovich score remained the highest

What carries the argument

The pipeline's engine is intermediate fusion by embedding concatenation inside a Cox-based survival model. Whole-slide images are compressed by frozen pathology foundation models (TITAN–CONCH or CHIEF–CTransPath) into patient-level vectors; CT volumes are cropped to the kidneys, downsampled, and encoded by fine-tuned 3D encoders (MedicalNet ResNet variants or SwinUNETR) into vectors of the same dimension. Concatenating the two vectors and feeding them through a small multilayer perceptron trained with the Cox partial-likelihood loss lets the model learn cross-modality feature interactions before risk scoring. The same setup is also run unimodally and with late fusion (a weighted average of s

Load-bearing premise

The 156-patient cohort used for all comparisons is the subset of patients whose CT scans passed a manual quality review; if the excluded 31 patients differ systematically in recurrence risk, the relative performance of the models may not generalize.

What would settle it

Take the same intermediate-fusion pipeline to an external ccRCC cohort with more recurrence events and compare, under identical nested cross-validation, fused WSI+CT against WSI-only: if the fused model does not exceed the WSI-only C-index, the paper's central claim that CT adds complementary value through fusion is refuted. Separately, if the adjusted Leibovich score keeps a large lead over learned models after randomized tie-breaking in a larger sample, the claim that discretization overstates its individualized performance would be weakened.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the reported ordering holds, future recurrence-risk tools could be built from routine imaging alone, with no extra tests beyond slides and CT scans already acquired in standard care.
  • CT's prognostic value in this setting is conditional on fusion: the same CT encoders that scored near chance alone contributed to the best fused model, so radiology should be evaluated in combination, not in isolation.
  • The adjusted Leibovich score's lead narrows substantially under random tie-breaking, implying that discrete clinical scores may overstate their ability to rank individual patients; continuous learned risk scores may be fairer comparators.
  • Because simple concatenation already improved every WSI–CT combination, more expressive fusion (cross-attention or co-learning) is a plausible next step the authors explicitly leave open.

Where Pith is reading between the lines

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

  • If the fusion gain replicates in a larger cohort, it would imply that current CT encoders, pretrained mainly for segmentation, are leaving prognostic information on the table; a CT foundation model trained for survival tasks might close the gap with pathology.
  • The paper's case analysis hints that slide sampling can miss aggressive regions; a natural extension is to test multiple slides per tumor or attention-based slide selection, which the authors did not do.
  • A practical translation would be a continuous imaging-based risk score that could be thresholded flexibly for surveillance intensity, rather than locked into three clinical risk groups; this follows from the paper's tie-breaking argument but is not tested here.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes a modular deep learning framework that integrates preoperative CT and postoperative whole-slide images (WSIs) for recurrence risk prediction in clear cell renal cell carcinoma (ccRCC). Using a TCGA-KIRC subset of 156 patients, it compares unimodal WSI and CT models, late fusion, and intermediate fusion, all based on frozen foundation-model encoders plus a Cox-based MLP prediction head. The adjusted Leibovich score is used as a clinical baseline. The paper reports that WSI-based models outperform CT-only models, that intermediate fusion gives the best learned model (TITAN-CONCH + ResNet-18, C-index 0.775 ± 0.044), approaching the adjusted Leibovich score (0.805 ± 0.035), and that random tie-breaking lowers the baseline to 0.749 ± 0.043.

Significance. If the claims hold, the study is a useful feasibility demonstration of multimodal imaging integration for a clinical task where structured scores remain the standard. Strengths include the use of a public dataset, nested cross-validation with fixed outer folds, and honest reporting of fold-level standard deviations. The main limitations are the small cohort (156 patients, 40 events), manual cohort filtering, and the absence of statistical testing for model comparisons, which makes the central fusion-advantage claim suggestive rather than established. The paper's methodological framework is sound in structure but needs additional statistical rigor before the conclusions can be accepted at face value.

major comments (4)
  1. [Section 4, Table 1] The central claim that intermediate fusion improves over unimodal WSI models is not statistically supported. The best intermediate model (TITAN-CONCH + ResNet-18) achieves C-index 0.775 ± 0.044 versus 0.745 ± 0.046 for TITAN-CONCH unimodal. With n=156 and 40 events, a 0.030 difference is well within the reported fold-level standard deviations. No paired significance test (e.g., permutation over patients or outer folds, bootstrap confidence intervals) is provided. Please add such tests for all pairwise comparisons of interest, report p-values/confidence intervals, and adjust the abstract/conclusion wording accordingly if the difference is not significant.
  2. [Section 3, Experimental Protocol] The 'best model' is identified from 18 learned configurations (2 WSI encoders × 3 CT encoders × 3 fusion strategies) using the same outer folds. No multiple-comparison correction or independent model-selection procedure is described. The reported best result may partly reflect selection bias. Please report results for all configurations (already in Table 1) and additionally provide a model-selection rule, e.g., choosing the fusion strategy on inner-fold performance only, or correct for the number of comparisons when making inferential claims.
  3. [Section 3, Datasets] The manual exclusion of 31 out of 187 patients based on CT quality, contrast phase, and kidney visibility is a potential source of selection bias. No comparison of excluded versus included patients' clinical characteristics or outcomes is provided. If the exclusions correlate with outcome, the relative model performance may not generalize. Please report baseline demographics, stage, grade, event rates, and follow-up for the excluded patients, and discuss or quantitatively assess the impact of this filtering.
  4. [Sections 3 and 4 (Adjusted Leibovich score; Case A)] There is an internal inconsistency regarding necrosis. Section 3 states that necrosis was omitted from the adjusted Leibovich score 'due to its absence in the dataset,' but Section 4 Case A states that 'the pathology report noted necrosis and high Fuhrman grade.' Clarify whether structured necrosis data were unavailable despite pathology reports containing the information, and discuss how this affects the validity of the adjusted baseline. Additionally, random tie-breaking is applied only to the Leibovich baseline; if the claim is that discretization overstates baseline performance, provide the same analysis for any tied risk scores produced by the learned models (or justify their absence).
minor comments (4)
  1. [Section 4, 'Performance Analysis of Unimodal and Multimodal Strategies'] The text states that 'all WSI-CT combinations matching or exceeding their WSI-only baselines' in intermediate fusion, but Table 1 shows TITAN-CONCH + ResNet-10 intermediate fusion at 0.742 versus 0.745 unimodal TITAN-CONCH. Please correct this overstatement.
  2. [Section 2/Table 1] The naming of CT encoders is unclear: Table 1 lists 'ResNet-10' and 'ResNet-18,' but Section 2 mentions MedicalNet and SwinUNETR. Please clarify which architecture corresponds to each named model and whether ResNet-10/18 are MedicalNet variants.
  3. [Section 4, Leibovich (RT)] The 'random tie-breaking' result is reported as a single mean C-index. Please specify the number of random repetitions and the seed or variance across repetitions to make the result reproducible.
  4. [General] The paper does not mention code or feature-extraction pipeline availability. Given the public dataset, providing code would strengthen reproducibility, even if only for preprocessing and evaluation.

Circularity Check

0 steps flagged

No circularity: the pipeline is a standard nested cross-validation benchmark against an external clinical score, with no fitted parameter renamed as a prediction and no load-bearing self-citation.

full rationale

The paper's derivation chain is empirical rather than definitional. Patient-level CT and WSI features are produced by externally pretrained encoders (TITAN, CONCH, CHIEF, CTransPath, MedicalNet, SwinUNETR) and then fed into a Cox-based MLP. Training uses five outer folds and three inner folds: hyperparameters, including the late-fusion weight alpha, are tuned only on inner folds, and final evaluation is on held-out outer folds. The reported C-indices are therefore not forced by construction. The adjusted Leibovich baseline is computed from structured clinicopathologic variables (with necrosis omitted) and is an independent clinical comparator, not an input to the learned models. The random tie-breaking analysis only re-scores the same baseline under a different tie convention; it does not transform a fitted quantity into a prediction. The paper invokes no uniqueness theorem and no self-citation chain: all referenced foundation models are external prior work, and no load-bearing claim depends on a citation by the present authors. The lack of significance testing and the variability across CT encoders are correctness/robustness concerns, not circularity. No self-definitional, fitted-input-as-prediction, or citation-smuggling pattern is present.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The central claim rests on standard survival modeling assumptions, transferability of pretrained encoders, and a manually curated cohort. Hyperparameters are tuned in inner CV, not on the test folds, so no circularity from that source.

free parameters (2)
  • Late fusion weight alpha = 0.8-0.9 (per outer fold)
    Tuned on inner validation folds to combine WSI and CT risk scores in late fusion; not used for the main intermediate fusion result.
  • MLP hyperparameters (learning rate, weight decay, L1, hidden size, dropout, scheduler floor) = Selected via Optuna on inner folds
    Standard hyperparameter optimization; nested CV ensures no test-set leakage.
axioms (4)
  • domain assumption Cox proportional hazards assumption holds for recurrence risk
    The training loss is the Cox partial likelihood, which assumes proportional hazards. The C-index remains a valid ranking metric even under non-proportionality, but the model's calibration could be affected.
  • domain assumption Pretrained foundation models transfer to ccRCC recurrence prediction
    The paper relies on CONCH, TITAN, CTransPath, CHIEF, MedicalNet, and SwinUNETR features being useful. Results partially validate this for pathology encoders, but CT encoders from segmentation tasks may not align well with survival prediction.
  • domain assumption Manual CT quality selection is unbiased
    Exclusion of 31 patients based on CT quality and contrast phase may introduce selection bias that affects generalizability.
  • domain assumption Adjusted Leibovich score without necrosis is a valid clinical comparator
    The original Leibovich score includes necrosis; its omission may weaken the baseline, affecting the 'approaching' claim.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Integrating Pathology and CT Imaging for Personalized Recurrence Risk Prediction in Renal Cancer." pith.science (2026). https://pith.science/paper/HTA2FIEU

@misc{pith2026250821581,
  author       = {Pith},
  title        = {Pith review of: Integrating Pathology and CT Imaging for Personalized Recurrence Risk Prediction in Renal Cancer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HTA2FIEU}},
  note         = {Machine review of arXiv:2508.21581}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Recurrence risk estimation in clear cell renal cell carcinoma (ccRCC) is essential for guiding postoperative surveillance and treatment. The Leibovich score remains widely used for stratifying distant recurrence risk but offers limited patient-level resolution and excludes imaging information. This study evaluates multimodal recurrence prediction by integrating preoperative computed tomography (CT) and postoperative histopathology whole-slide images (WSIs). A modular deep learning framework with pretrained encoders and Cox-based survival modeling was tested across unimodal, late fusion, and intermediate fusion setups. In a real-world ccRCC cohort, WSI-based models consistently outperformed CT-only models, underscoring the prognostic strength of pathology. Intermediate fusion further improved performance, with the best model (TITAN-CONCH with ResNet-18) approaching the adjusted Leibovich score. Random tie-breaking narrowed the gap between the clinical baseline and learned models, suggesting discretization may overstate individualized performance. Using simple embedding concatenation, radiology added value primarily through fusion. These findings demonstrate the feasibility of foundation model-based multimodal integration for personalized ccRCC risk prediction. Future work should explore more expressive fusion strategies, larger multimodal datasets, and general-purpose CT encoders to better match pathology modeling capacity.

Figures

Figures reproduced from arXiv: 2508.21581 by Cedrik Blommestijn, Dani\"el Boeke, Ines P. Machado, James O. Jones, Kalina Chupetlovska, Mireia Crispin-Ortuzar, Rebecca N. Wray, Regina G. H. Beets-Tan, Shangqi Gao, Wilson Silva, Zeyu Gao.

Figure 1
Figure 1. Figure 1: Overview of the model pipeline. Preoperative CT and resected pathology WSIs are independently preprocessed and encoded into patient-level feature vectors. These vectors are used in unimodal survival models or combined using intermediate fusion via embedding concatenation or late fusion strategies for recurrence risk prediction. Manual segmentations, with kidneys in green and tumors in orange, were not used… view at source ↗
Figure 2
Figure 2. Figure 2: Case examples illustrating predicted recurrence risk from the best-performing model, intermediate fusion of TITAN–CONCH and ResNet-18, compared to the ad￾justed Leibovich score. Risk scores are shown as percentiles within the full patient cohort. All patients experienced recurrence. Bars indicate each method’s estimated relative risk. Kidney tissue is delineated in green and tumor tissue in orange on CT. m… view at source ↗

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

25 extracted references · 8 canonical work pages

  1. [1]

    Koyama, M.: Optuna: A Next-generation Hyperparameter Optimization Framework (2019)

    Akiba, T., Sano, S., Yanase, T., Ohta, T. Koyama, M.: Optuna: A Next-generation Hyperparameter Optimization Framework (2019). https://doi.org/10.48550/arXiv.1907.10902

    Pith/arXiv arXiv doi:10.48550/arxiv.1907.10902 2019

  2. [2]

    https://doi.org/10.7937/K9/TCIA.2016.V6PBVTDR

    Akin, O., et al.: The Cancer Genome Atlas Kidney Re- nal Clear Cell Carcinoma Collection (TCGA-KIRC) (2016). https://doi.org/10.7937/K9/TCIA.2016.V6PBVTDR

    doi:10.7937/k9/tcia.2016.v6pbvtdr 2016

  3. [3]

    https://doi.org/10.7937/JC8X-9874

    Baid, U., et al.: RSNA-ASNR-MICCAI-BraTS-2021 (2023). https://doi.org/10.7937/JC8X-9874

    work page doi:10.7937/jc8x-9874 2021

  4. [4]

    Uro- logic Oncology: Seminars and Original Investigations 39(7), 438.e11–438.e21 (Jul 2021)

    Blackmur, J.P., et al.: Leibovich score is the optimal clinico-pathological system associated with recurrence of non-metastatic clear cell renal cell carcinoma. Uro- logic Oncology: Seminars and Original Investigations 39(7), 438.e11–438.e21 (Jul 2021). https://doi.org/10.1016/j.urolonc.2021.04.007

    work page doi:10.1016/j.urolonc.2021.04.007 2021

  5. [5]

    Montorsi, F.: Renal cancer

    Capitanio, U. Montorsi, F.: Renal cancer. The Lancet 387(10021), 894–906 (Feb 2016). https://doi.org/10.1016/S0140-6736(15)00046-X

    work page doi:10.1016/s0140-6736(15)00046-x 2016

  6. [6]

    Zheng, Y.: Med3D: Transfer Learning for 3D Medical Image Analysis (2019)

    Chen, S., Ma, K. Zheng, Y.: Med3D: Transfer Learning for 3D Medical Image Analysis (2019). https://doi.org/10.48550/arXiv.1904.00625

    Pith/arXiv arXiv doi:10.48550/arxiv.1904.00625 2019

  7. [7]

    Journal of the Royal Statis- tical Society Series B: Statistical Methodology 34(2), 187–202 (Jan 1972)

    Cox, D.R.: Regression Models and Life-Tables. Journal of the Royal Statis- tical Society Series B: Statistical Methodology 34(2), 187–202 (Jan 1972). https://doi.org/10.1111/j.2517-6161.1972.tb00899.x

    arXiv 1972

  8. [8]

    Journal of Open Source Software 4(40), 1317 (Aug 2019)

    Davidson-Pilon, C.: Lifelines: Survival analysis in Python. Journal of Open Source Software 4(40), 1317 (Aug 2019). https://doi.org/10.21105/joss.01317

    doi:10.21105/joss.01317 2019

  9. [9]

    https://doi.org/10.48550/arXiv.2411.19666

    Ding, T., et al.: Multimodal Whole Slide Foundation Model for Pathology (Nov 2024). https://doi.org/10.48550/arXiv.2411.19666

    Pith/arXiv arXiv doi:10.48550/arxiv.2411.19666 2024

  10. [10]

    The Lancet Digital Health 5(8), e515–e524 (Aug 2023)

    Gui, C.P., et al.: Multimodal recurrence scoring system for prediction of clear cell renal cell carcinoma outcome: A discovery and validation study. The Lancet Digital Health 5(8), e515–e524 (Aug 2023). https://doi.org/10.1016/S2589- 7500(23)00095-X

    doi:10.1016/s2589- 2023

  11. [11]

    McNeil, B.J.: The meaning and use of the area under a receiver operating characteristic (ROC) curve

    Hanley, J.A. McNeil, B.J.: The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143(1), 29–36 (Apr 1982). https://doi.org/10.1148/radiology.143.1.7063747

    doi:10.1148/radiology.143.1.7063747 1982

  12. [12]

    JAMA: The Jour- nal of the American Medical Association 247(18), 2543 (May 1982)

    Harrell, F.E.: Evaluating the Yield of Medical Tests. JAMA: The Jour- nal of the American Medical Association 247(18), 2543 (May 1982). https://doi.org/10.1001/jama.1982.03320430047030

    work page arXiv 1982

  13. [13]

    Xu, D.: Swin UNETR: Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images (2022)

    Hatamizadeh, A., Nath, V., Tang, Y., Yang, D., Roth, H. Xu, D.: Swin UNETR: Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images (2022). https://doi.org/10.48550/arXiv.2201.01266

    Pith/arXiv arXiv doi:10.48550/arxiv.2201.01266 2022

  14. [14]

    https://doi.org/10.48550/arXiv.2307.01984

    Heller, N., et al.: The KiTS21 Challenge: Automatic segmentation of kid- neys, renal tumors, and renal cysts in corticomedullary-phase CT (2023). https://doi.org/10.48550/arXiv.2307.01984

    Pith/arXiv arXiv doi:10.48550/arxiv.2307.01984 2023

  15. [15]

    https://doi.org/10.48550/arXiv.1809.10486

    Isensee, F., et al.: nnU-Net: Self-adapting Framework for U-Net-Based Medical Image Segmentation (2018). https://doi.org/10.48550/arXiv.1809.10486

    Pith/arXiv arXiv doi:10.48550/arxiv.1809.10486 2018

  16. [16]

    Kluger, Y.: Deep- Surv: Personalized treatment recommender system using a Cox proportional haz- ards deep neural network

    Katzman, J.L., Shaham, U., Cloninger, A., Bates, J., Jiang, T. Kluger, Y.: Deep- Surv: Personalized treatment recommender system using a Cox proportional haz- ards deep neural network. BMC Medical Research Methodology 18(1), 24 (Dec 2018). https://doi.org/10.1186/s12874-018-0482-1 12 D. Boeke et al

    doi:10.1186/s12874-018-0482-1 2018

  17. [17]

    Cancer 97(7), 1663–1671 (Apr 2003)

    Leibovich, B.C., et al.: Prediction of progression after radical nephrec- tomy for patients with clear cell renal cell carcinoma: A stratification tool for prospective clinical trials. Cancer 97(7), 1663–1671 (Apr 2003). https://doi.org/10.1002/cncr.11234

    work page doi:10.1002/cncr.11234 2003

  18. [18]

    European Urology 73(5), 772–780 (May 2018)

    Leibovich, B.C., et al.: Predicting Oncologic Outcomes in Renal Cell Carcinoma After Surgery. European Urology 73(5), 772–780 (May 2018). https://doi.org/10.1016/j.eururo.2018.01.005

    work page doi:10.1016/j.eururo.2018.01.005 2018

  19. [19]

    Nature Medicine 30(3), 863–874 (Mar 2024)

    Lu, M.Y., et al.: A visual-language foundation model for computational pathology. Nature Medicine 30(3), 863–874 (Mar 2024). https://doi.org/10.1038/s41591-024- 02856-4

    doi:10.1038/s41591-024- 2024

  20. [20]

    npj Precision Oncology 8(1), 45 (Feb 2024)

    Margue, G., et al.: UroPredict: Machine learning model on real-world data for pre- diction of kidney cancer recurrence (UroCCR-120). npj Precision Oncology 8(1), 45 (Feb 2024). https://doi.org/10.1038/s41698-024-00532-x

    work page doi:10.1038/s41698-024-00532-x 2024

  21. [21]

    Nature Genetics 45(10), 1113–1120 (Oct 2013)

    The Cancer Genome Atlas Research Network, et al.: The Cancer Genome At- las Pan-Cancer analysis project. Nature Genetics 45(10), 1113–1120 (Oct 2013). https://doi.org/10.1038/ng.2764

    doi:10.1038/ng.2764 2013

  22. [22]

    https://doi.org/10.48550/arXiv.2501.16652

    Vaidya, A., et al.: Molecular-driven Foundation Model for Oncologic Pathology (Jan 2025). https://doi.org/10.48550/arXiv.2501.16652

    Pith/arXiv arXiv doi:10.48550/arxiv.2501.16652 2025

  23. [23]

    Medical Image Analysis 81, 102559 (Oct 2022)

    Wang, X., et al.: Transformer-based unsupervised contrastive learning for histopathological image classification. Medical Image Analysis 81, 102559 (Oct 2022). https://doi.org/10.1016/j.media.2022.102559

    work page arXiv 2022

  24. [24]

    Nature 634(8035), 970–978 (Oct 2024)

    Wang, X., et al.: A pathology foundation model for cancer diagno- sis and prognosis prediction. Nature 634(8035), 970–978 (Oct 2024). https://doi.org/10.1038/s41586-024-07894-z

    doi:10.1038/s41586-024-07894-z 2024

  25. [25]

    Mahmood, F.: Accelerating Data Processing and Benchmarking of AI Models for Pathology (Feb 2025)

    Zhang, A., Jaume, G., Vaidya, A., Ding, T. Mahmood, F.: Accelerating Data Processing and Benchmarking of AI Models for Pathology (Feb 2025). https://doi.org/10.48550/arXiv.2502.06750

    Pith/arXiv arXiv doi:10.48550/arxiv.2502.06750 2025

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.