Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

PanTS: The Pancreatic Tumor Segmentation Dataset

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read PanTS, a 36,390-scan, 145-center pancreatic CT dataset, claims its 16x larger tumor-annotation count is what makes AI detect, localize, and segment pancreatic tumors better than AI trained on existing public datasets.

desk verdict A genuinely large and useful pancreatic CT dataset with real benchmark results, but the claim that gains are 'directly attributable' to annotation scale is confounded and should be rewritten. read the letter →

arxiv 2507.01291 v1 pith:YKSPZLU6 submitted 2025-07-02 eess.IV cs.CV

Wenxuan Li , Xinze Zhou , Qi Chen , Tianyu Lin , Pedro R. A. S. Bassi , Szymon Plotka , Jaroslaw B. Cwikla , Xiaoxi Chen
show 10 more authors
Chen Ye Zheren Zhu Kai Ding Heng Li Kang Wang Yang Yang Yucheng Tang Daguang Xu Alan L. Yuille Zongwei Zhou
This is my paper
classification eess.IVcs.CV
keywords pancreatictumorsegmentationCTdatasetvoxel-wiseannotationsscaleout-of-distributiongeneralizationanatomicalcontextsupervisionpancreasheadbodyandtailmedicalimagingbenchmark
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

PanTS is a pancreatic CT dataset of 36,390 scans from 145 medical centers, carrying expert-validated, voxel-level labels for pancreatic tumors, the pancreas head/body/tail, and 24 surrounding anatomical structures — over 993,000 annotated structures in all. The paper's central claim is that AI models trained on PanTS detect, localize, and segment pancreatic tumors substantially better than models trained on existing public datasets, and that this improvement is directly caused by the $16\times$ larger number of labeled tumors, not by better model architecture. Evidence for the claim: a fixed nnU-Net trained on PanTS's 9,901 training scans reaches 0.959 detection AUC on scans from unseen hospitals, versus 0.810 and 0.819 for models trained on the 281-scan and 2,238-scan public sets, and the same model ranks first on the official MSD-Pancreas leaderboard. Adding the 24 anatomy labels as training targets further lifts tumor segmentation by +10.3% Dice, suggesting anatomical context helps as well. If these claims hold, labeling more tumors — rather than designing smarter models — is the binding constraint on making pancreatic cancer AI work in new hospitals.

What carries the argument

The central object is the dataset itself: 36,390 CT scans from 145 medical centers, of which 9,901 are released for training and 26,489 are reserved for third-party evaluation, with expert-validated voxel-wise annotations covering 28 classes — pancreatic tumor, pancreas head/body/tail, and 24 surrounding vascular, skeletal, abdominal, and thoracic structures — totaling over 993,000 annotated structures. The argument's engine is the comparison design: an identical nnU-Net (a self-configuring segmentation network) is trained on three datasets of increasing tumor-annotation count, isolating dataset scale as the variable of interest, and evaluated both on an out-of-distribution PanTS test set (via ROC-AUC for detection) and on the official MSD-Pancreas test set (via DSC and NSD for segmentation). The second mechanism is anatomical context: adding the 24 surrounding structures as training labels is argued to act as an implicit regularizer that sharpens tumor boundaries, supported by the +10.3% DSC gain of the 28-class model over the 2-class model and by t-SNE feature plots showing better separation of tumor features. Inter-annotator agreement (median DSC 86.1% across 300 re-annotated scans, with a 20% agreement threshold triggering expert review) is the quality-control instrument that underwrites the reliability of the labels that drive these gains.

What would settle it

Retrain the same nnU-Net on a size-matched subset of PanTS (2,238 scans, matching PANORAMA) with tumor labels only, and measure detection AUC on the PanTS test set: if AUC stays near 0.82, annotation count alone does not explain the jump to 0.959, while if AUC climbs well above 0.82, scale survives as the driver. A complementary check relabels all tumor types in PANORAMA as 'tumor' instead of background; if its AUC then rises toward 0.96, labeling protocol, not raw count, is the decisive factor.

Watch

Extended reading notes

Core claim

The central claim is that scaling up voxel-wise tumor annotations is the dominant factor in improving AI performance for pancreatic tumor detection, localization, and segmentation, and that a dataset combining that scale with anatomy-aware labels is what current public resources lack. The paper supports this by training an identical nnU-Net on three pancreatic CT datasets of increasing size — MSD-Pancreas (281 scans), PANORAMA (2,238 scans), and PanTS's 9,901 training scans — and reporting that detection AUC on a held-out set from unseen hospitals rises from 0.810 to 0.819 to 0.959, with the largest jump arriving with the largest annotation count. It further reports that a PanTS-trained nnU-Net ranks first on the official MSD-Pancreas leaderboard, beating prior entries by at least +4.9% DSC and +3.1% NSD in tumor segmentation, and that expanding training from a 2-class labeling (tumor and pancreas only) to a 28-class labeling (tumor, pancreas head/body/tail, and 24 surrounding structures) improves tumor DSC by +10.3% and NSD by +9.7%, which it interprets as the surrounding structures acting as an implicit spatial regularizer. The paper's conclusion is that these gains are directly attributable to the $16\times$ larger-scale tumor annotations and indirectly supported by the 24 additional structures, making PanTS a new benchmark and a design template for large medical-imaging datasets.

Load-bearing premise

The load-bearing premise is that the three compared training sets differ mainly in how many tumors are labeled, yet they also differ in which tumor types are labeled (one set labels only pancreatic ductal adenocarcinoma), in scanner mix and contrast phases, and in the simultaneous addition of 24 extra structure labels — and the training set actually used in the experiment is 4.4x larger than the prior set, not the $16\times$ figure the abstract credits.

Editorial extensions

If this is right

  • If scale of labeled tumors is the main driver, then dataset construction — not architecture search — becomes the highest-leverage investment for pancreatic cancer AI, and the 26,489-scan reserved test set provides a stable target for measuring it.
  • Models trained on PanTS should transfer better to new hospitals: the test set comes from three centers entirely unseen during training, yet detection AUC reaches 0.959, supporting the claim of strong out-of-distribution generalization.
  • Training with surrounding anatomy labels as auxiliary targets should reduce false positives in tumor segmentation, since the 28-class model outperformed the 2-class model by +10.3% DSC and +9.7% NSD on the same test data.
  • Adding PanTS as a training resource should raise the bar on the MSD-Pancreas benchmark, where the released baseline already ranks first with at least +4.9% DSC and +3.1% NSD margins over prior entries.
  • The standardized pancreas head/body/tail partition gives researchers a consistent way to report tumor location, which is needed for staging and resectability assessment.

Reading between the lines

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

  • The AUC jump from 0.819 to 0.959 coincides with two simultaneous changes — a 4.4x increase in training scans and a switch to labeling every tumor type rather than only pancreatic ductal adenocarcinoma — so the paper's causal attribution to annotation scale is not fully isolated; a size-matched, protocol-matched ablation would be needed to separate the two.
  • If the causal story transfers, the PanTS recipe — massive tumor labeling plus auxiliary anatomy targets — should generalize to other subtle, low-prevalence tumors in CT (for example ovarian or small-bowel lesions), where small label counts currently cap model performance.
  • The 24 structure labels and the head/body/tail partition give PanTS potential use beyond tumor segmentation, such as supervised targets for anatomy-aware report generation or radiotherapy contouring, which the paper lists as future directions rather than benchmarked outcomes.
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.

Referee Report

3 major / 5 minor

Summary. The manuscript introduces PanTS, a large-scale multi-institutional CT dataset for pancreatic tumor analysis, comprising 36,390 scans from 145 medical centers with voxel-wise annotations of pancreatic tumors, pancreas head/body/tail, and 24 surrounding anatomical structures, together with per-scan metadata. The authors release 9,901 training scans, retain 26,489 test scans for third-party evaluation, and provide nnU-Net baselines. The central claims are that models trained on PanTS significantly outperform models trained on existing public datasets in tumor detection, localization, and segmentation, and that these gains are 'directly attributable' to the 16x larger-scale tumor annotations, with the 24 additional anatomical structures providing indirect support. The paper also reports that a PanTS-trained nnU-Net is top-1 on the official MSD-Pancreas leaderboard.

Significance. If the claims hold, PanTS is a substantial community resource: it is the largest pancreatic CT dataset to date, includes a multi-center held-out test set, provides a reproducible nnU-Net baseline, reports an inter-annotator agreement study, and has third-party MSD leaderboard support. The dataset release, the detailed annotation protocol, and the external leaderboard evaluation are concrete strengths. However, the paper's causal attribution of the performance gains to annotation scale is not established by the reported experiments, because the scale comparison changes multiple variables simultaneously. The 2-class versus 28-class comparison in Section 5 is better controlled and provides credible evidence that anatomical context helps, but it does not rescue the scale-attribution claim. The dataset itself is valuable even if the causal claim is weakened.

major comments (3)
  1. [Abstract and §4, Fig. 5A] The abstract's claim that the gains are 'directly attributable to the 16x larger-scale tumor annotations' is not supported by the experiment in Fig. 5A, because the three compared datasets differ not only in the number of annotated tumors but also in label protocol and tumor-type coverage. As footnote 4 admits, PANORAMA annotates only PDAC and treats all other tumors and healthy pancreases as 'Normal', whereas PanTS labels all tumor types. The AUC jump from 0.819 to 0.959 could therefore reflect the removal of label ambiguity or a change in label semantics, rather than the increase in annotation count alone. The paper should either present a controlled comparison that isolates scale (e.g., subsets of PanTS matched in protocol and composition) or explicitly weaken the attribution to a combined effect of scale, label completeness, and diversity.
  2. [Abstract, §2.2, and Fig. 5A] The '16x' figure is misleading relative to the experiment actually run. The abstract and Section 2.2 compare the full PanTS dataset (36,390 scans) with PANORAMA (2,238 scans), but the trained model uses only the 9,901-scan training set, which is 4.4x PANORAMA, not 16x. The scaling experiment therefore does not test the quantity claimed in the abstract. This mismatch should be corrected in the text, and the causal wording should be tied to the actual training-set sizes used in the comparison.
  3. [§3.3, §3.4, and §4, Fig. 5A] The comparison in Fig. 5A is also confounded by a label-protocol mismatch between training datasets and the PanTS test set. The PanTS test ground truth is annotated with the PanTS standard, which includes all tumor types, whereas models trained on MSD-Pancreas and PANORAMA were trained with their own label definitions (e.g., PDAC-only for PANORAMA). A model trained only on PDAC may be penalized on the PanTS test set for not detecting non-PDAC tumors, even if its detections are correct under its own training label semantics. This protocol mismatch is a separate variable from annotation scale and can bias the comparison in favor of the PanTS-trained model. The paper should acknowledge this and, if possible, provide an evaluation that controls for label-protocol differences (for example, by evaluating on a PDAC-only subset of the test set).
minor comments (5)
  1. [§3] There is a typo in the sentence describing the test set: 'thirty-party evaluation' should read 'third-party evaluation'.
  2. [Appendix B.3.2] The heading of B.3.2 repeats 'Justification of Annotating Large-Scale Tumor Datasets'; it should refer to the annotation of the 24 surrounding anatomical structures, matching the experiment actually described.
  3. [§3.1 and Table 1] The text states that the test set 'contains a higher frequency of tumor occurrences than the training set', but Table 1 shows nearly identical tumor prevalence (10.8% vs 10.7%). The sentence should be corrected or the statistic it refers to should be clarified (e.g., tumor burden per positive case).
  4. [Appendix A, Table 3] The statement that PanTS is 'over 8.5x larger than the most extensive existing dataset dedicated to pancreatic tumor detection' is inconsistent with the '16x' figure used elsewhere, because the denominators differ (compare with Trauma Det. at 4,714 CTs and PANORAMA at 2,238 CTs). The comparisons should be made against a consistent baseline.
  5. [Fig. 5A] The ROC curves are shown without confidence intervals or statistical tests for the AUC differences. Given that the test set is large, adding CIs or a DeLong test would strengthen the quantitative support for the reported differences.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the benchmark results are produced by independent training/test splits and third-party MSD evaluation, with no fitted parameter or self-citation chain driving the central claim.

full rationale

I found no circular step that reduces a claimed derivation to its own inputs. Section 4 trains a standard nnU-Net on three public training sets (MSD-Pancreas, PANORAMA, PanTS) and evaluates on the held-out PanTS test set drawn from centers unseen at training; Section 5 compares 2-class vs 28-class supervision on the same cohort. The strongest external support is the official MSD-Pancreas leaderboard, where the PanTS-trained nnU-Net is evaluated by third-party challenge organizers, so the headline segmentation gains do not rest on a self-referential fit. The abstract's causal phrase 'directly attributable to the 16x larger-scale tumor annotations' is a confounded inference, not a circular one: footnote 4 itself acknowledges that PANORAMA's annotation protocol differs (PDAC-only labeling), and the trained model uses 9,901 scans (4.4x PANORAMA) rather than the full 36,390. These are validity and framing concerns about isolating the scale variable, not cases where an output equals an input by construction. Self-citations to prior annotation workflows (e.g., Refs. 49, 36, 66 for human-in-the-loop labeling) are used as methodology support, not as the evidence for the benchmark claims, so they do not make the derivation circular. I therefore assign score 0.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

No new physical/theoretical entities are introduced. The ledger records hand-chosen thresholds and domain assumptions about annotation quality and dataset separation that the conclusions depend on.

free parameters (2)
  • DSC quality threshold = 20%
    Hand-chosen minimum Dice for flagging low-agreement cases in inter-annotator review (Section 3.5). It is a validation filter, not an input to the main result, but it is an ad hoc choice.
  • Train/test split ratio = 27% / 73%
    Chosen to create a large held-out test set; this ratio determines the size of the test benchmark and therefore the statistical power of the comparisons.
assumptions (5)
  • domain assumption Expert annotations in PanTS are treated as ground truth for training and evaluation.
    The benchmark and all conclusions assume radiologist labels are correct; annotation quality is only spot-checked on 300 training scans.
  • domain assumption Training and test sets are from entirely different centers, so performance differences reflect out-of-distribution generalization.
    Section 3.1/3.2 state the test set is from 3 centers not seen in training; patient-level overlap is not explicitly ruled out.
  • domain assumption The 11 public datasets aggregated into the training set have compatible and sufficiently accurate existing labels.
    Section 3.2/Appendix A; the original labels are not re-audited on a per-case basis except for tumors.
  • domain assumption AI-assisted annotations for the 24 non-tumor structures are equivalent in quality to fully manual annotations.
    Section 3.3: initial organ annotations were AI-generated and then manually verified/corrected; no separate quality study for these structures is reported.
  • standard math nnU-Net with identical settings is a fair base model for comparing datasets.
    nnU-Net is a standard self-configuring framework; the paper asserts configurations were held constant, but nnU-Net auto-configures some settings based on the label set.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PanTS: The Pancreatic Tumor Segmentation Dataset." pith.science (2026). https://pith.science/paper/YKSPZLU6

@misc{pith2026250701291,
  author       = {Pith},
  title        = {Pith review of: PanTS: The Pancreatic Tumor Segmentation Dataset},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YKSPZLU6}},
  note         = {Machine review of arXiv:2507.01291}
}
read the original abstract

PanTS is a large-scale, multi-institutional dataset curated to advance research in pancreatic CT analysis. It contains 36,390 CT scans from 145 medical centers, with expert-validated, voxel-wise annotations of over 993,000 anatomical structures, covering pancreatic tumors, pancreas head, body, and tail, and 24 surrounding anatomical structures such as vascular/skeletal structures and abdominal/thoracic organs. Each scan includes metadata such as patient age, sex, diagnosis, contrast phase, in-plane spacing, slice thickness, etc. AI models trained on PanTS achieve significantly better performance in pancreatic tumor detection, localization, and segmentation compared to those trained on existing public datasets. Our analysis indicates that these gains are directly attributable to the 16x larger-scale tumor annotations and indirectly supported by the 24 additional surrounding anatomical structures. As the largest and most comprehensive resource of its kind, PanTS offers a new benchmark for developing and evaluating AI models in pancreatic CT analysis.

Figures

Figures reproduced from arXiv: 2507.01291 by the authors.

Figure 1
Figure 1. Dataset characteristics and visualization. A. PanTS comprises 36,390 CT scans collected from 145 medical centers, paired with expert-validated voxel-wise annotations, 16× larger than the biggest public dataset (i.e., PANORAMA [4]) to date. B–C. The dataset includes detailed annotations for pancreatic tumors, pancreas, and its head, body, and tail, enabling spatially aware tumor localization. D. Twenty-four surroundi… view at source ↗
Figure 2
Figure 2. Geographic diversity of the PanTS dataset. Global distribution of contributing centers in the PanTS training set (purple circles) and test set (red outlines). Circle size is proportional to the base-10 logarithm (log10) of the number of CT scans contributed per country. The training set is aggregated from diverse public datasets spanning multiple countries, while the much larger test set is exclusively drawn from th… view at source ↗
Figure 3
Figure 3. Annotation standard and quality control. A–C. Voxel-wise annotations of pancreatic tu￾mors and surrounding anatomical structures shown on axial, sagittal, and coronal planes. Radiologists provide these annotations following the standard described in §3.4. D. 3D rendering on the coronal plane highlights detailed annotations of the tumor, pancreas, and key vessels, including the celiac artery (Celiac AA), superior mes… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Inter-annotator agreement on the PanTS subset. A. Distribution of DSC (%) values between two independent radiologists across 300 CT scans from the PanTS training set. Most annotations demonstrate high agreement, confirming their reliability. A minimum threshold of DSC …
Figure 5
Figure 5. Figure 5: Justification of annotating large-scale tumor datasets. A. The Receiver Operating Characteristic (ROC) curve of standard nnU-Net trained on different scale of pancreatic CT datasets, i.e., MSD-Pancreas (n = 281), PANORAMA (n = 2,238), and our PanTS dataset (n = 9,901).…
Figure 6
Figure 6. Figure 6: Justification of annotating 24 surrounding anatomical structures. We compare nnU-Net models trained with 2 classes (tumor and pancreas) versus 28 classes (tumor, pancreas head/body/tail, and 24 surrounding anatomical structures). The 28-class model significantly improv…

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Learning Segmentation from Radiology Reports

    eess.IV 2025-07 conditional novelty 7.0 of 10

    R-Super converts tumor count, size, and location information from radiology reports into voxel-wise losses that improve CT tumor segmentation beyond training with masks alone.

  2. Learning Patient-Specific Disease Dynamics with Latent Flow Matching for Longitudinal Imaging Generation

    cs.CV 2025-12 conditional novelty 5.0 of 10

    Δ-LFM generates future brain MRIs by learning a patient-specific velocity field in a latent space where trajectory magnitude grows monotonically with disease time, outperforming prior models on three Alzheimer's datasets.

Reference graph

Works this paper leans on

68 extracted references · 49 canonical work pages · cited by 2 Pith papers

  1. [1]

    Sauter, Ivan Nesic, Marko Obradovic, Shan Yang, Sebastian Manneck, Carl Glessgen, Johanna M

    Lorraine Abel, Jakob Wasserthal, Thomas Weikert, Alexander W. Sauter, Ivan Nesic, Marko Obradovic, Shan Yang, Sebastian Manneck, Carl Glessgen, Johanna M. Ospel, Bram Stieltjes, Daniel T. Boll, and Björn Friebe. Automated detection of pancreatic cystic lesions on ct using deep learning. Diagnostics, 11(5), 2021. ISSN 2075-4418. doi: 10.3390/diagnostics110...

    work page doi:10.3390/diagnostics11050901 2021

  2. [2]

    Hugo J. W. L. Aerts, Emmanuel R. Velazquez, Ralph T. H. Leijenaar, Chintan Parmar, Patrick Grossmann, Sara Carvalho, Johan Bussink, Ronald Monshouwer, Benjamin Haibe-Kains, David Rietveld, Frank Hoebers, Marieke M. Rietbergen, René Leemans, Andre Dekker, John Quacken- bush, Robert Gillies, and Philippe Lambin. A whole-body FDG-PET/CT dataset with manually...

    doi:10.1038/s41597-022-01718-3 2022

  3. [3]

    Erickson, Sheri Kirk, Yu Lee, Marston W

    Oguz Akin, Paula Elnajjar, Mark Heller, Rosemary Jarosz, Bradley J. Erickson, Sheri Kirk, Yu Lee, Marston W. Linehan, Rajan Gautam, Ramesh Vikram, Kelly M. Garcia, Charles Roche, Emanuele Bonaccio, and Jeffrey Filippini. The Cancer Genome Atlas Kidney Renal Clear Cell Carcinoma Collection (TCGA-KIRC) (Version 3), 2016. URL https://doi.org/10.7937/ K9/TCIA...

    work page 2016

  4. [4]

    The panorama study protocol: Pancreatic cancer diagnosis-radiologists meet ai

    N Alves, M Schuurmans, D Rutkowski, et al. The panorama study protocol: Pancreatic cancer diagnosis-radiologists meet ai. zenodo, 2024

    work page 2024

  5. [5]

    The medical segmentation decathlon

    Michela Antonelli, Annika Reinke, Spyridon Bakas, Keyvan Farahani, Bennett A Landman, Geert Litjens, Bjoern Menze, Olaf Ronneberger, Ronald M Summers, Bram van Ginneken, et al. The medical segmentation decathlon. arXiv preprint arXiv:2106.05735, 2021

    arXiv 2021

  6. [6]

    The medical segmentation decathlon

    Michela Antonelli, Annika Reinke, Spyridon Bakas, Keyvan Farahani, Annette Kopp-Schneider, Bennett A Landman, Geert Litjens, Bjoern Menze, Olaf Ronneberger, Ronald M Summers, et al. The medical segmentation decathlon. Nature communications, 13(1):1–13, 2022

    work page 2022

  7. [7]

    Pedro RAS Bassi, Wenxuan Li, Yucheng Tang, Fabian Isensee, Zifu Wang, Jieneng Chen, Yu-Cheng Chou, Yannick Kirchhoff, Maximilian Rokuss, Ziyan Huang, Jin Ye, Junjun He, Tassilo Wald, Constantin Ulrich, Michael Baumgartner, Saikat Roy, Klaus H. Maier-Hein, Paul Jaeger, Yiwen Ye, Yutong Xie, Jianpeng Zhang, Ziyang Chen, Yong Xia, Zhaohu Xing, Lei Zhu, Youse...

    work page 2024

  8. [8]

    Label Critic: Design Data Before Models

    Pedro RAS Bassi, Qilong Wu, Wenxuan Li, Sergio Decherchi, Andrea Cavalli, Alan Yuille, and Zongwei Zhou. Label critic: Design data before models. arXiv preprint arXiv:2411.02753,

    work page Pith review arXiv

Show all 68 references
  1. [9]

    Radgpt: Con- structing 3d image-text tumor datasets

    Pedro RAS Bassi, Mehmet Can Yavuz, Kang Wang, Xiaoxi Chen, Wenxuan Li, Sergio Decherchi, Andrea Cavalli, Yang Yang, Alan Yuille, and Zongwei Zhou. Radgpt: Con- structing 3d image-text tumor datasets. arXiv preprint arXiv:2501.04678 , 2025. URL https://github.com/MrGiovanni/RadGPT. 10

    2025 arXiv

  2. [10]

    The liver tumor segmentation benchmark (lits)

    Patrick Bilic, Patrick Ferdinand Christ, Eugene V orontsov, Grzegorz Chlebus, Hao Chen, Qi Dou, Chi-Wing Fu, Xiao Han, Pheng-Ann Heng, Jürgen Hesser, et al. The liver tumor segmentation benchmark (lits). arXiv preprint arXiv:1901.04056, 2019

    1901 arXiv

  3. [11]

    Large-scale pancreatic cancer detection via non-contrast ct and deep learning

    Kai Cao, Yingda Xia, Jiawen Yao, Xu Han, Lukas Lambert, Tingting Zhang, Wei Tang, Gang Jin, Hui Jiang, Xu Fang, et al. Large-scale pancreatic cancer detection via non-contrast ct and deep learning. Nature medicine, 29(12):3033–3043, 2023

    2023

  4. [12]

    Monai: An open-source framework for deep learning in healthcare

    M Jorge Cardoso, Wenqi Li, Richard Brown, Nic Ma, Eric Kerfoot, Yiheng Wang, Benjamin Murrey, Andriy Myronenko, Can Zhao, Dong Yang, et al. Monai: An open-source framework for deep learning in healthcare. arXiv preprint arXiv:2211.02701, 2022

    2022 arXiv

  5. [13]

    the advance of solid tumor research in china

    Luohai Chen, Wei Wang, Kaizhou Jin, Bing Yuan, Huangying Tan, Jian Sun, Yu Guo, Yanji Luo, Shi-Ting Feng, Xianjun Yu, et al. Special issue “the advance of solid tumor research in china”: Prediction of sunitinib efficacy using computed tomography in patients with pancreatic neu...

    2023

  6. [14]

    Embracing massive medical data

    Yu-Cheng Chou, Zongwei Zhou, and Alan Yuille. Embracing massive medical data. In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 24–35. Springer, 2024. URL https://github.com/MrGiovanni/OnlineLearning

    2024

  7. [15]

    Diagnosis and detection of pancreatic cancer

    Linda C Chu, Michael G Goggins, and Elliot K Fishman. Diagnosis and detection of pancreatic cancer. The Cancer Journal, 23(6):333–342, 2017

    2017

  8. [16]

    Clinical clues of pre-symptomatic pancreatic ductal adenocarcinoma prior to its diagnosis: a retrospective review of ct scans and laboratory tests

    Hwe Hoon Chung, Kyung Sook Lim, and Joo Kyung Park. Clinical clues of pre-symptomatic pancreatic ductal adenocarcinoma prior to its diagnosis: a retrospective review of ct scans and laboratory tests. Clinics and Practice, 12(1):70–77, 2022

    2022

  9. [17]

    Rsna 2023 abdominal trauma detection

    Errol Colak, Hui-Ming Lin, Robyn Ball, Melissa Davis, Adam Flanders, Sabeena Jalal, Kirti Magudia, Brett Marinelli, Savvas Nicolaou, Luciano Prevedello, Jeff Rudie, George Shih, Maryam Vazirabad, and John Mongan. Rsna 2023 abdominal trauma detection. https: //kaggle.com/compet...

    2023

  10. [18]

    Dints: Differentiable neural network topology search for 3d medical image segmentation

    Yufan He, Dong Yang, Holger Roth, Can Zhao, and Daguang Xu. Dints: Differentiable neural network topology search for 3d medical image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5841–5850, 2021

    2021

  11. [19]

    The kits19 challenge data: 300 kidney tumor cases with clinical context, ct semantic segmentations, and surgical outcomes

    Nicholas Heller, Niranjan Sathianathen, Arveen Kalapara, Edward Walczak, Keenan Moore, Heather Kaluzniak, Joel Rosenberg, Paul Blake, Zachary Rengel, Makinna Oestreich, et al. The kits19 challenge data: 300 kidney tumor cases with clinical context, ct semantic segmentations, a...

    1904 arXiv

  12. [20]

    An international challenge to use artificial intelligence to define the state-of-the-art in kidney and kidney tumor segmentation in ct imaging., 2020

    Nicholas Heller, Sean McSweeney, Matthew Thomas Peterson, Sarah Peterson, Jack Rickman, Bethany Stai, Resha Tejpaul, Makinna Oestreich, Paul Blake, Joel Rosenberg, et al. An international challenge to use artificial intelligence to define the state-of-the-art in kidney and kid...

    2020

  13. [22]

    Pancreatic steatosis on computed tomography is an early imaging feature of pre-diagnostic pancreatic cancer: A preliminary study in overweight patients

    Sanne A Hoogenboom, Candice W Bolan, Anthony Chuprin, Maria T Raimondo, Jeanin E van Hooft, Michael B Wallace, and Massimo Raimondo. Pancreatic steatosis on computed tomography is an early imaging feature of pre-diagnostic pancreatic cancer: A preliminary study in overweight p...

    2021

  14. [23]

    Artificial intelligence in pancreatic cancer

    Bowen Huang, Haoran Huang, Shuting Zhang, Dingyue Zhang, Qingya Shi, Jianzhou Liu, and Junchao Guo. Artificial intelligence in pancreatic cancer. Theranostics, 12(16):6931, 2022. 11

    2022

  15. [24]

    nnu-net: a self-configuring method for deep learning-based biomedical image segmentation

    Fabian Isensee, Paul F Jaeger, Simon AA Kohl, Jens Petersen, and Klaus H Maier-Hein. nnu-net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods, 18(2):203–211, 2021

    2021

  16. [25]

    nnu-net revisited: A call for rigorous validation in 3d medical image segmentation

    Fabian Isensee, Tassilo Wald, Constantin Ulrich, Michael Baumgartner, Saikat Roy, Klaus Maier-Hein, and Paul F Jaeger. nnu-net revisited: A call for rigorous validation in 3d medical image segmentation. arXiv preprint arXiv:2404.09556, 2024

    2024 arXiv

  17. [26]

    Amos: A large-scale abdominal multi-organ benchmark for versatile medical image segmentation

    Yuanfeng Ji, Haotian Bai, Chongjian Ge, Jie Yang, Ye Zhu, Ruimao Zhang, Zhen Li, Lingyan Zhanng, Wanling Ma, Xiang Wan, et al. Amos: A large-scale abdominal multi-organ benchmark for versatile medical image segmentation. Advances in neural information processing systems, 35:36...

    2022

  18. [27]

    Label-assemble: Leveraging multiple datasets with partial labels

    Mintong Kang, Bowen Li, Zengle Zhu, Yongyi Lu, Elliot K Fishman, Alan Yuille, and Zongwei Zhou. Label-assemble: Leveraging multiple datasets with partial labels. In IEEE International Symposium on Biomedical Imaging, pages 1–5. IEEE, 2023. URL https://github.com/ MrGiovanni/La...

    2023

  19. [28]

    Scaling laws for neural language models

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling laws for neural language models. arXiv preprint arXiv:2001.08361, 2020

    2001 arXiv

  20. [29]

    Chaos challenge- combined (ct-mr) healthy abdominal organ segmentation

    A Emre Kavur, N Sinem Gezer, Mustafa Barı¸ s, Sinem Aslan, Pierre-Henri Conze, Vladimir Groza, Duc Duy Pham, Soumick Chatterjee, Philipp Ernst, Sava¸ s Özkan, et al. Chaos challenge- combined (ct-mr) healthy abdominal organ segmentation. Medical Image Analysis, 69:101950, 2021

    2021

  21. [30]

    Scalable neural architecture search for 3d medical image segmentation

    Sungwoong Kim, Ildoo Kim, Sungbin Lim, Woonhyuk Baek, Chiheon Kim, Hyungjoo Cho, Boogeon Yoon, and Taesup Kim. Scalable neural architecture search for 3d medical image segmentation. In International Conference on Medical Image Computing and Computer-Assisted Intervention, page...

    2019

  22. [31]

    A retrospective preliminary study of intrapancreatic late en- hancement as a noteworthy imaging finding in the early stages of pancreatic adenocarcinoma

    Yoshihiro Konno, Yasuhiro Sugai, Masafumi Kanoto, Keisuke Suzuki, Toshitada Hiraka, Yuki Toyoguchi, and Kazuho Niino. A retrospective preliminary study of intrapancreatic late en- hancement as a noteworthy imaging finding in the early stages of pancreatic adenocarcinoma. Europ...

    2023

  23. [32]

    Miccai multi-atlas labeling beyond the cranial vault–workshop and challenge

    Bennett Landman, Zhoubing Xu, J Igelsias, Martin Styner, T Langerak, and Arno Klein. Miccai multi-atlas labeling beyond the cranial vault–workshop and challenge. In Proc. MICCAI Multi-Atlas Labeling Beyond Cranial Vault—Workshop Challenge, volume 5, page 12, 2015

    2015

  24. [33]

    Early detection and localization of pancreatic cancer by label-free tumor synthesis

    Bowen Li, Yu-Cheng Chou, Shuwen Sun, Hualin Qiao, Alan Yuille, and Zongwei Zhou. Early detection and localization of pancreatic cancer by label-free tumor synthesis. MICCAI Workshop on Big Task Small Data, 1001-AI, 2023. URL https://github.com/MrGiovanni/ SyntheticTumors

    2023

  25. [34]

    Abdomenatlas: A large-scale, detailed-annotated, & multi-center dataset for efficient transfer learning and open algorithmic benchmarking

    Wenxuan Li, Chongyu Qu, Xiaoxi Chen, Pedro RAS Bassi, Yijia Shi, Yuxiang Lai, Qian Yu, Huimin Xue, Yixiong Chen, Xiaorui Lin, et al. Abdomenatlas: A large-scale, detailed-annotated, & multi-center dataset for efficient transfer learning and open algorithmic benchmarking. Medic...

    2024

  26. [35]

    How well do supervised models transfer to 3d image segmentation? In International Conference on Learning Representations, 2024

    Wenxuan Li, Alan Yuille, and Zongwei Zhou. How well do supervised models transfer to 3d image segmentation? In International Conference on Learning Representations, 2024. URL https://github.com/MrGiovanni/SuPreM

    2024

  27. [36]

    Scalemai: Accelerating the development of trusted datasets and ai models

    Wenxuan Li, Pedro RAS Bassi, Tianyu Lin, Yu-Cheng Chou, Xinze Zhou, Yucheng Tang, Fabian Isensee, Kang Wang, Qi Chen, Xiaowei Xu, et al. Scalemai: Accelerating the development of trusted datasets and ai models. arXiv preprint arXiv:2501.03410 , 2025. URL https: //github.com/Mr...

    2025

  28. [37]

    Text-driven tumor synthesis

    Xinran Li, Yi Shuai, Chen Liu, Qi Chen, Qilong Wu, Pengfei Guo, Dong Yang, Can Zhao, Pedro RAS Bassi, Daguang Xu, et al. Text-driven tumor synthesis. arXiv preprint arXiv:2412.18589, 2024. URL https://github.com/MrGiovanni/TextoMorph. 12

    2024 arXiv

  29. [38]

    Clip-driven universal model for organ segmentation and tumor detection

    Jie Liu, Yixiao Zhang, Jie-Neng Chen, Junfei Xiao, Yongyi Lu, Bennett A Landman, Yixuan Yuan, Alan Yuille, Yucheng Tang, and Zongwei Zhou. Clip-driven universal model for organ segmentation and tumor detection. In Proceedings of the IEEE/CVF International Confer- ence on Compu...

    2023

  30. [39]

    Universal and extensible language- vision models for organ segmentation and tumor detection from abdominal computed tomog- raphy

    Jie Liu, Yixiao Zhang, Kang Wang, Mehmet Can Yavuz, Xiaoxi Chen, Yixuan Yuan, Hao- liang Li, Yang Yang, Alan Yuille, Yucheng Tang, et al. Universal and extensible language- vision models for organ segmentation and tumor detection from abdominal computed tomog- raphy. Medical I...

    2024

  31. [40]

    Word: Revisiting organs segmentation in the whole abdominal region

    Xiangde Luo, Wenjun Liao, Jianghong Xiao, Tao Song, Xiaofan Zhang, Kang Li, Guotai Wang, and Shaoting Zhang. Word: Revisiting organs segmentation in the whole abdominal region. arXiv preprint arXiv:2111.02403, 2021

    2021 arXiv

  32. [41]

    Word: A large scale dataset, benchmark and clinical applicable study for abdominal organ segmentation from ct image

    Xiangde Luo, Wenjun Liao, Jianghong Xiao, Jieneng Chen, Tao Song, Xiaofan Zhang, Kang Li, Dimitris N Metaxas, Guotai Wang, and Shaoting Zhang. Word: A large scale dataset, benchmark and clinical applicable study for abdominal organ segmentation from ct image. Medical Image Ana...

    2022

  33. [42]

    Abdomenct-1k: Is abdominal organ segmentation a solved problem? arXiv preprint arXiv:2010.14808, 2020

    Jun Ma, Yao Zhang, Song Gu, Yichi Zhang, Cheng Zhu, Qiyuan Wang, Xin Liu, Xingle An, Cheng Ge, Shucheng Cao, et al. Abdomenct-1k: Is abdominal organ segmentation a solved problem? arXiv preprint arXiv:2010.14808, 2020

    2010 arXiv

  34. [43]

    Abdomenct-1k: Is abdominal organ segmentation a solved problem

    Jun Ma, Yao Zhang, Song Gu, Cheng Zhu, Cheng Ge, Yichi Zhang, Xingle An, Congcong Wang, Qiyuan Wang, Xin Liu, et al. Abdomenct-1k: Is abdominal organ segmentation a solved problem. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021

    2021

  35. [45]

    Automatic organ and pan-cancer segmentation in abdomen ct: the flare 2023 challenge

    Jun Ma, Yao Zhang, Song Gu, Cheng Ge, Ershuai Wang, Qin Zhou, Ziyan Huang, Pengju Lyu, Jian He, and Bo Wang. Automatic organ and pan-cancer segmentation in abdomen ct: the flare 2023 challenge. arXiv preprint arXiv:2408.12534, 2024

    2023 arXiv

  36. [46]

    Pancreatic cancer: A review of clinical diagnosis, epidemiology, treatment and outcomes

    Andrew McGuigan, Paul Kelly, Richard C Turkington, Claire Jones, Helen G Coleman, and R Stephen McCain. Pancreatic cancer: A review of clinical diagnosis, epidemiology, treatment and outcomes. World journal of gastroenterology, 24(43):4846, 2018

    2018

  37. [47]

    Moawad, Diana Fuentes, Ahmed Morshid, Ahmed M

    Ahmed W. Moawad, Diana Fuentes, Ahmed Morshid, Ahmed M. Khalaf, Mohamed M. Elmohr, Ahmed Abusaif, John D. Hazle, Ahmed O. Kaseb, Mohamed Hassan, Amir Mahvash, Jan Szklaruk, Ali Qayyom, and Khaled Elsayes. Multimodality annotated HCC cases with and without advanced imaging segm...

    2021 doi

  38. [48]

    Radiomics-based machine-learning models can detect pancreatic cancer on prediagnostic computed tomography scans at a substantial lead time before clinical diagnosis

    Sovanlal Mukherjee, Anurima Patra, Hala Khasawneh, Panagiotis Korfiatis, Naveen Rajamo- han, Garima Suman, Shounak Majumder, Ananya Panda, Matthew P Johnson, Nicholas B Larson, et al. Radiomics-based machine-learning models can detect pancreatic cancer on prediagnostic compute...

    2022

  39. [49]

    Abdomenatlas-8k: Annotating 8,000 abdominal ct volumes for multi-organ segmentation in three weeks

    Chongyu Qu, Tiezheng Zhang, Hualin Qiao, Jie Liu, Yucheng Tang, Alan Yuille, and Zongwei Zhou. Abdomenatlas-8k: Annotating 8,000 abdominal ct volumes for multi-organ segmentation in three weeks. In Conference on Neural Information Processing Systems, volume 21, 2023. URL https...

    2023

  40. [50]

    Ct-org, a new dataset for multiple organ segmentation in computed tomography

    Blaine Rister, Darvin Yi, Kaushik Shivakumar, Tomomi Nobashi, and Daniel L Rubin. Ct-org, a new dataset for multiple organ segmentation in computed tomography. Scientific Data, 7(1): 1–9, 2020. 13

    2020

  41. [51]

    Deeporgan: Multi-level deep convolutional networks for automated pancreas segmentation

    Holger R Roth, Le Lu, Amal Farag, Hoo-Chang Shin, Jiamin Liu, Evrim B Turkbey, and Ronald M Summers. Deeporgan: Multi-level deep convolutional networks for automated pancreas segmentation. In International conference on medical image computing and computer- assisted interventi...

    2015

  42. [52]

    Cancer statistics, 2024

    Rebecca L Siegel, Angela N Giaquinto, and Ahmedin Jemal. Cancer statistics, 2024. CA: a cancer journal for clinicians, 74(1):12–49, 2024

    2024

  43. [53]

    Cancer statistics, 2025

    Rebecca L Siegel, Tyler B Kratzer, Angela N Giaquinto, Hyuna Sung, and Ahmedin Jemal. Cancer statistics, 2025. Ca, 75(1):10, 2025

    2025

  44. [54]

    Computerized tomography scan in pre-diagnostic pancreatic ductal adenocarcinoma: Stages of progression and potential benefits of early intervention: A retrospective study

    Dhruv Pratap Singh, Shannon Sheedy, Ajit H Goenka, Michael Wells, Nam Ju Lee, John Barlow, Ayush Sharma, Harika Kandlakunta, Shruti Chandra, Sushil Kumar Garg, et al. Computerized tomography scan in pre-diagnostic pancreatic ductal adenocarcinoma: Stages of progression and pot...

    2020

  45. [55]

    Self-supervised pre-training of swin transformers for 3d medical image analysis

    Yucheng Tang, Dong Yang, Wenqi Li, Holger R Roth, Bennett Landman, Daguang Xu, Vishwesh Nath, and Ali Hatamizadeh. Self-supervised pre-training of swin transformers for 3d medical image analysis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognit...

    2022

  46. [56]

    Efficient 3d representation learning for medical image analysis

    Yucheng Tang, Jie Liu, Zongwei Zhou, Xin Yu, and Yuankai Huo. Efficient 3d representation learning for medical image analysis. World Scientific Annual Review of Artificial Intelligence, 2024

    2024

  47. [57]

    Abdominal or pelvic enhanced CT images within 10 days before surgery of 230 patients with stage II colorectal cancer (StageII-Colorectal-CT), 2022

    Tao Tong and Ming Li. Abdominal or pelvic enhanced CT images within 10 days before surgery of 230 patients with stage II colorectal cancer (StageII-Colorectal-CT), 2022. URL https://doi.org/10.7937/p5k5-tg43. [Dataset]

    2022 doi

  48. [58]

    Ct abnormalities of the pancreas associated with the subsequent diagnosis of clinical stage i pancreatic ductal adenocarcinoma more than 1 year later: a case-control study

    Fumihito Toshima, Ryosuke Watanabe, Dai Inoue, Norihide Yoneda, Tatsuya Yamamoto, Naoki Sasahira, Takashi Sasaki, Masato Matsuyama, Kaori Minehiro, Ukihide Tateishi, et al. Ct abnormalities of the pancreas associated with the subsequent diagnosis of clinical stage i pancreatic...

    2021

  49. [59]

    Multi-modal learning from unpaired images: Application to multi-organ segmentation in ct and mri

    Vanya V Valindria, Nick Pawlowski, Martin Rajchl, Ioannis Lavdas, Eric O Aboagye, Andrea G Rockall, Daniel Rueckert, and Ben Glocker. Multi-modal learning from unpaired images: Application to multi-organ segmentation in ct and mri. In 2018 IEEE winter conference on application...

    2018

  50. [60]

    Visualizing data using t-sne

    Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne. Journal of machine learning research, 9(11), 2008

    2008

  51. [61]

    Totalseg- mentator: robust segmentation of 104 anatomic structures in ct images

    Jakob Wasserthal, Hanns-Christian Breit, Manfred T Meyer, Maurice Pradella, Daniel Hinck, Alexander W Sauter, Tobias Heye, Daniel T Boll, Joshy Cyriac, Shan Yang, et al. Totalseg- mentator: robust segmentation of 104 anatomic structures in ct images. Radiology: Artificial Inte...

    2023

  52. [62]

    The felix project: Deep networks to detect pancreatic neoplasms

    Yingda Xia, Qihang Yu, Linda Chu, Satomi Kawamoto, Seyoun Park, Fengze Liu, Jieneng Chen, Zhuotun Zhu, Bowen Li, Zongwei Zhou, et al. The felix project: Deep networks to detect pancreatic neoplasms. medRxiv, 2022

    2022

  53. [63]

    C2fnas: Coarse-to-fine neural architecture search for 3d medical image segmentation

    Qihang Yu, Dong Yang, Holger Roth, Yutong Bai, Yixiao Zhang, Alan L Yuille, and Daguang Xu. C2fnas: Coarse-to-fine neural architecture search for 3d medical image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4126–4135, 2020

    2020

  54. [64]

    Scaling vision transform- ers

    Xiaohua Zhai, Alexander Kolesnikov, Neil Houlsby, and Lucas Beyer. Scaling vision transform- ers. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 12104–12113, 2022

    2022

  55. [65]

    Challenges in diagnosis of pancreatic cancer

    Lulu Zhang, Santosh Sanagapalli, and Alina Stoita. Challenges in diagnosis of pancreatic cancer. World journal of gastroenterology, 24(19):2047, 2018. 14

    2018

  56. [66]

    Leveraging ai predicted and expert revised annotations in interactive segmentation: Continual tuning or full training? In IEEE International Symposium on Biomedical Imaging

    Tiezheng Zhang, Xiaoxi Chen, Chongyu Qu, Alan Yuille, and Zongwei Zhou. Leveraging ai predicted and expert revised annotations in interactive segmentation: Continual tuning or full training? In IEEE International Symposium on Biomedical Imaging. IEEE, 2024. URL https://github....

    2024

  57. [67]

    Contin- ual learning for abdominal multi-organ and tumor segmentation

    Yixiao Zhang, Xinyi Li, Huimiao Chen, Alan L Yuille, Yaoyao Liu, and Zongwei Zhou. Contin- ual learning for abdominal multi-organ and tumor segmentation. In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 35–45. Springer, 2023. URL...

    2023

  58. [68]

    Pancreatic cancer: a review of risk factors, diagnosis, and treatment

    ZhiYu Zhao and Wei Liu. Pancreatic cancer: a review of risk factors, diagnosis, and treatment. Technology in cancer research & treatment, 19:1533033820962117, 2020

    2020

  59. [69]

    Assembling existing labels from public datasets to diagnose novel diseases: Covid-19 in late 2019.NeurIPS Workshop on Medical Imaging meets NeurIPS, 2022

    Zengle Zhu, Mintong Kang, Alan Yuille, and Zongwei Zhou. Assembling existing labels from public datasets to diagnose novel diseases: Covid-19 in late 2019.NeurIPS Workshop on Medical Imaging meets NeurIPS, 2022. URL https://github.com/MrGiovanni/LabelAssemble. 15 Appendix Tabl...

    2019

  60. [2024]

    URL https://github.com/PedroRASB/LabelCritic

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.