CT Data Curation for Liver Patients: Phase Recognition in Dynamic Contrast-Enhanced CT

Adam P. Harrison; Bo Zhou; Chien-Hung Liao; Chi-Tung Cheng; Jiawen Yao; Jing Xiao; Le Lu

REVIEW 2 major objections 4 minor 17 references

CT Data Curation for Liver Patients: Phase Recognition in Dynamic Contrast-Enhanced CT

T0 review · 2 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A compact 3D vision model labels liver CT phases at 97.7% F1

desk verdict A practical vision-based PACS curation pipeline with a clean weak-label loss; the headline F1 is plausible but the evaluation rests on unstated annotation and split details that need a revision. read the letter →

arxiv 1909.02511 v2 pith:HNUHODCE submitted 2019-09-05 eess.IV cs.CV

Bo Zhou , Adam P. Harrison , Jiawen Yao , Chi-Tung Cheng , Jing Xiao , Chien-Hung Liao , Le Lu This is my paper · ORCID

classification eess.IVcs.CV

keywords datacurationPACSdynamicCTphaserecognitionliver3Dconvolutionalnetworksqueeze-and-excitationweaksupervision

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 asks whether visual information alone can replace unreliable text mining when harvesting medical imaging datasets from hospital PACS systems. It claims that a deliberately small 3D convolutional network with squeeze-and-excitation blocks, trained on noisy DICOM-tag labels, can recognize the scan phase of dynamic liver CT studies—non-contrast, arterial, venous, or delay—with a mean F1 of 0.977 on a held-out test set. Because many scans are only labeled as generic "contrast", the authors introduce an aggregated cross-entropy loss that treats contrast as the union of the three contrast phases, letting the model learn from weakly labeled data instead of discarding it. They report that this approach correctly harvests 92.7% of studies, improving on both the text-mined baseline and larger 3D architectures. If correct, the work makes large-scale, low-labor curation of clinical CT data feasible for downstream modeling.

What carries the argument

The load-bearing object is the 3DSE network paired with the aggregated cross-entropy (ACE) loss. 3DSE is a streamlined 3D convolutional classifier that downsamples volumes to 128x128x32, extracts features with two 3D convolutions, recalibrates feature channels using squeeze-and-excitation blocks to inject global context, and outputs five classes (non-contrast, arterial, venous, delay, other). The ACE loss converts each weakly labeled "contrast" scan into a training signal by defining $p_{\text{contrast}} = p_{\text{arterial}} + p_{\text{venous}} + p_{\text{delay}}$ through the logits and using a logsumexp formulation of cross entropy that is numerically stable. This lets scans without a precise phase label still contribute gradients that sharpen the model's phase predictions, which is the mechanism behind the reported F1 and harvest-rate gains.

What would settle it

Take a random sample of, say, 300 scans from the same hospital PACS distribution, have several board-certified radiologists independently label the phase, and measure the model's F1 against their consensus; if mean F1 falls substantially below 0.977, or if inter-radiologist agreement is itself below the assumed level, the claim of near-perfect curation fails on independent ground truth.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a compact 3DSE classifier trained with weak text-mined labels and an aggregated cross-entropy (ACE) loss can identify dynamic liver CT phases with mean F1 of 0.977 at scan level and correctly extract 92.7% of studies, meaning all scans in the study are classified correctly. The improvement over text mining is driven by visual cues that cannot always be recovered from imperfect DICOM tags: venous scans, for example, rise from F1 0.868 to 0.963. The ACE loss contributes by encoding the hierarchical relation that a scan labeled "contrast" is one of arterial, venous, or delay, expressed as the sum of their softmax probabilities, so weakly supervised scans help train the fine-grained outputs. The authors also report that a 19.22 MB model outperforms a 33.56 MB C3D and a 325.22 MB 3D-ResNet, attributing the latter's failure to overfitting.

Load-bearing premise

The entire evaluation depends on the manual phase labels of the validation and test sets being correct, but the paper gives no information about how many annotators labeled them, their training, or how disagreements were resolved.

Editorial extensions

If this is right

Hospitals can harvest dynamic liver CT studies from messy PACS archives without hand-curating every scan, using text-mined rules only to seed a visual model.
Weakly labeled scans marked only "contrast" no longer need to be discarded; the ACE loss turns them into usable training data for fine-grained phase outputs.
Small streamlined 3D models with SE layers are a viable alternative to much larger 3D networks for this classification task, with less overfitting and faster training.
Study-level harvest rates near 93% mean downstream analyses on multi-phase liver CT can draw on thousands of studies that would otherwise be lost or require manual review.

Reading between the lines

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

The same aggregated-loss trick applies to any curation problem where a coarse label is a union of fine classes, such as "tumor" versus its subtypes, so ACE could be reused beyond CT phases.
Cases where text labels conflict with visual content suggest a hybrid system that lets the visual model flag or correct suspicious DICOM tags at harvest time, an extension the authors only gesture toward when listing future work on harmonizing text and vision.
A natural testable extension is to run the same 3DSE+ACE pipeline on a different organ's dynamic protocol, such as renal or pancreatic CT, and check whether the 0.977-level F1 transfers or whether new phase-specific visual cues are required.
The reported superiority of 3DSE over 3D-ResNet may reflect dataset scale as much as architecture; on larger training sets the larger model might close the gap, since the paper's training set is about 30K scans.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

Dear colleague,

Short version: this is a useful, well-scoped paper on using visual information to curate dynamic liver CT studies from hospital PACS. The core idea—train a small 3D network on noisy DICOM text labels, add an aggregated cross-entropy loss for generic 'contrast' labels, and evaluate on manually labeled scans—is sensible, and the reported gains are large. The ACE loss is a simple logsumexp trick, clearly derived and numerically stable. The study-level harvest rate (92.7% of studies fully correct) is the metric that matters for data curation. If the results hold, this saves real radiology-lab time.

What's genuinely new is applying visual phase recognition to PACS curation rather than text mining alone, plus the weak-label aggregation for contrast-only annotations. The 3DSE architecture is not groundbreaking, but it's appropriately small and shows that a compact model with global channel attention can beat much larger 3D nets. Citations look normal, and the comparison against text mining and larger models is fair.

Soft spots, in proportion.

First, the manual labels. Section 2.1 says 'we then manually annotated a validation set and a test set' and nothing else: no annotator expertise, no instructions, no inter-observer agreement. Since the entire evaluation is measured against these labels, this is a real gap. It's not fatal—the visual cues are standard radiology—but a revision should include at least a small consensus reading or a detailed annotation protocol.

Second, the train/validation/test split is not explicitly at the patient level. Patient counts are 3,267 + 101 + 196 = 3,564, versus 4,666 total, leaving ~1,100 patients unaccounted for. With 1–3 studies per patient, a test study could come from a patient whose other study is in training, which could inflate F1 through anatomy or scanner correlation. I suspect it's an omitted detail rather than a deliberate shortcut, but it must be stated and verified.

Third, no code or data is released, and training details are sparse beyond a learning-rate sweep. For a practical pipeline paper, that limits reproducibility but is common at this venue.

The stress-test note says these issues could undermine the headline numbers. I think they are legitimate reviewer questions, but they are addressable in revision. The text-mining baseline is weak, so the absolute F1 might drop a little under a fully clean split, but the relative gains from the ACE loss and 3DSE are credible.

Who should read this: people working on medical image curation, PACS mining, or weak supervision in clinical imaging. It's a solid systems paper, not a conceptual breakthrough. It deserves serious peer review; my recommendation is conditional accept pending clear confirmation of patient-level separation, annotation details, and ideally code or data.

Best, [Your name]

Referee Report

2 major / 4 minor

Summary. This paper addresses automated curation of dynamic contrast-enhanced CT liver studies from a hospital PACS. The authors use simple text-mined DICOM labels as noisy supervision, train a compact 3D squeeze-and-excitation network (3DSE), and introduce an aggregated cross-entropy (ACE) loss that treats a 'contrast' label as the union of arterial, venous, and delay phases. On a manually annotated test set of 231 studies, the model achieves a mean scan-level F1 of 0.977 and a study-level harvest accuracy of 92.7%, outperforming text mining, C3D, and 3D-ResNet. The paper includes statistical significance testing for scan-level F1.

Significance. If the reported results are reproducible, the paper offers a practical solution to a recognized bottleneck in medical imaging ML: extracting large labeled datasets from real-world PACS. The 3DSE model is lightweight (19.22 MB), the ACE loss is a simple and generalizable way to incorporate hierarchical weak labels, and the comparison is against a noisy baseline that mimics realistic deployment. The paper's main value is its evaluation on a large heterogeneous dataset, but the current manuscript leaves two evaluation-critical details underspecified: the exact construction and patient-level disjointness of the train/validation/test splits, and the protocol for the manual annotation serving as ground truth. These must be clarified before the headline accuracy and harvest rates can be fully trusted.

major comments (2)

[Section 2.1] The reported counts for the training, validation, and test sets do not add up to the stated totals: the paper gives 4,666 patients, 7,680 studies, and 43,010 scans in total, while the validation and test sets account for 101 and 196 patients, 150 and 231 studies, and 801 and 1,262 scans, respectively. This leaves 4,369 patients, 7,299 studies, and 40,947 scans, but the training set is reported as 3,267 patients, 5,164 studies, and 29,891 scans, leaving roughly 1,100 patients and 2,100 studies unaccounted for. The authors must specify the exact filtering or subsampling steps and state explicitly whether the split is at the patient level. Without this, it is unclear whether any patient's studies appear in both training and test, which could inflate the reported F1 and harvest rates.
[Section 2.1] The validation and test sets are the sole ground truth for the evaluation, but the paper provides no information about the annotation process: who annotated, what instructions or reference criteria were used, whether disagreements were resolved by consensus, or any inter-observer agreement statistic. The reported F1 of 0.977 and harvest rate of 92.7% are measures of agreement with these manual labels; without details on their reliability, the clinical significance of these numbers is unclear. Please describe the annotator qualifications, the annotation interface or instructions, and any quality assurance measures.

minor comments (4)

[Table 3] The text mining row reports 80.9% accuracy, but 186/231 equals 80.5%; please reconcile the denominator or the percentage.
[Section 3] The paper states that the D-phase degradation 'does not meet statistical significance' but does not report p-values; providing the actual p-values or confidence intervals would allow the reader to assess the strength of the significance claims.
[Table 2] The row label '3DSE-SE' is ambiguous; please rename it to something like '3DSE (without SE)' for clarity.
[Section 2.1] The sentence 'The text-matching rules are more than sufficient to reliably extract labels based on text alone' is contradicted by the later results showing a text-mining F1 of 0.938; please rephrase to reflect the actual performance.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the ACE loss is a modeling choice and the headline F1 is evaluated against independent manual labels.

full rationale

The central claim is an empirical evaluation claim. The ACE loss (Sec. 2.3, Eqs. 1-5) defines pC = pA + pV + pD and derives a logsumexp loss for weakly labeled "contrast" scans; this is a modeling assumption about how to supervise partial labels, not a fitted parameter or a prediction defined in terms of the reported F1. The model is trained on noisy text-mined labels (Sec. 2.1, with rules in supplementary Table 1) and evaluated on a separately manually annotated validation/test set (Sec. 2.1), so the reported scan-level and study-level F1 numbers are not forced by the text-mining rules or by the ACE loss. No load-bearing derivation in the paper reduces to its own inputs. Self-citations ([7], [17]) concern prior segmentation and visualization works and are not used to justify the phase-recognition performance or to exclude alternative methods. The unvalidated manual labels and patient-level split ambiguity are annotation/experimental-design risks, not circularity. Therefore no circular step is present.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The central claim does not depend on any fitted free parameters in the derivation; the ACE loss is parameter-free. The load-bearing assumptions are domain assumptions about label quality and the visual learnability of phase. No new entities are introduced.

assumptions (4)

domain assumption Manually annotated test set labels are accurate ground truth.
Section 2.1 describes manual annotation of validation and test sets; the entire evaluation relies on these labels being correct.
domain assumption Text-mined DICOM labels are noisy but weakly informative.
Section 2.1 states simple rules can extract scan descriptions but errors and omissions make labels unreliable, motivating the visual model.
domain assumption The contrast label is the union of arterial, venous, and delay phases.
Section 2.3, Eq. (1) defines pC = pA + pV + pD, which assumes these three phases partition the 'contrast' category.
domain assumption The 3D SE network can learn discriminative visual cues from noisy labels.
The entire method depends on the assumption that phase information is visually encoded in the CT volumes and that a 3D CNN can extract it, as argued in the introduction and shown in Fig. 1.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CT Data Curation for Liver Patients: Phase Recognition in Dynamic Contrast-Enhanced CT." pith.science (2026). https://pith.science/paper/HNUHODCE

@misc{pith2026190902511,
  author       = {Pith},
  title        = {Pith review of: CT Data Curation for Liver Patients: Phase Recognition in Dynamic Contrast-Enhanced CT},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HNUHODCE}},
  note         = {Machine review of arXiv:1909.02511}
}

read the original abstract

As the demand for more descriptive machine learning models grows within medical imaging, bottlenecks due to data paucity will exacerbate. Thus, collecting enough large-scale data will require automated tools to harvest data/label pairs from messy and real-world datasets, such as hospital PACS. This is the focus of our work, where we present a principled data curation tool to extract multi-phase CT liver studies and identify each scan's phase from a real-world and heterogenous hospital PACS dataset. Emulating a typical deployment scenario, we first obtain a set of noisy labels from our institutional partners that are text mined using simple rules from DICOM tags. We train a deep learning system, using a customized and streamlined 3D SE architecture, to identify non-contrast, arterial, venous, and delay phase dynamic CT liver scans, filtering out anything else, including other types of liver contrast studies. To exploit as much training data as possible, we also introduce an aggregated cross entropy loss that can learn from scans only identified as "contrast". Extensive experiments on a dataset of 43K scans of 7680 patient imaging studies demonstrate that our 3DSE architecture, armed with our aggregated loss, can achieve a mean F1 of 0.977 and can correctly harvest up to 92.7% of studies, which significantly outperforms the text-mined and standard-loss approach, and also outperforms other, and more complex, model architectures.

Figures

Figures reproduced from arXiv: 1909.02511 by the authors.

**Figure 2.** Our 3DSE network is designed to have a relatively small amount of parameters [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Respond-CAM [17] visualizations of 3DSE from three different dynamic CT scans. (A) the 3DSE focuses on contrast accumulation in the cardiac region; (V): 3DSE focuses on contrast remnants in the cardiac blood pool, liver portal veins, and kidney veins; (D): 3DSE focuses on contrast accumulation in the ureters of the kidney [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 1.** Figure 1: Additional respond-CAM visualizations of 3DSE. [PITH_FULL_IMAGE:figures/full_fig_p010_1.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

[1]

Medical Image Analysis 42 (2017) 60–88

Litjens, G.J.S., Kooi, T., Bejnordi, B.E., Setio, A.A.A., Ciompi, F., Ghafoorian, M., van der Laak, J.A.W.M., van Ginneken, B., Snchez, C.I.: A survey on deep learning in medical image analysis. Medical Image Analysis 42 (2017) 60–88

work page 2017
[2]

In: Asian Conference on Computer Vision, Springer (2018) 298–313

Zhou, B., Lin, X., Eck, B., Hou, J., Wilson, D.: Generation of virtual dual energy images from standard single-shot radiographs using multi-scale and conditional adversarial network. In: Asian Conference on Computer Vision, Springer (2018) 298–313

work page 2018
[3]

In: IEEE CVPR

Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L.: Imagenet: A large-scale hierarchical image database. In: IEEE CVPR. (2009) 248–255

work page 2009
[4]

In: Journal of Digital Imaging

Kohli, M.D., Summers, R.M., Geis, J.R.: Medical image data and datasets in the era of machine learning: Whitepaper from the 2016 c-mimi meeting dataset session. In: Journal of Digital Imaging. (2017)

work page 2017
[5]

In: Artiﬁcial Intelligence in Medical Imaging

Harvey, H., Glocker, B.: A standardised approach for preparing imaging data for machine learning tasks in radiology. In: Artiﬁcial Intelligence in Medical Imaging. Springer (2019) 61–72

work page 2019
[6]

Journal of Medical Imaging 5(3) (2018) 036501

Yan, K., Wang, X., Lu, L., Summers, R.M.: Deeplesion: automated mining of large-scale lesion annotations and universal lesion detection with deep learning. Journal of Medical Imaging 5(3) (2018) 036501

work page 2018
[7]

In: 2019 IEEE Winter Conference on Applications of Computer Vision (WACV), IEEE (2019) 1–10

Zhou, B., Chen, A., Crawford, R., Dogdas, B., Goldmarcher, G.: A progressively- trained scale-invariant and boundary-aware deep neural network for the automatic 3d segmentation of lung lesions. In: 2019 IEEE Winter Conference on Applications of Computer Vision (WACV), IEEE (2019) 1–10

work page 2019
[8]

AAAI (2019)

Irvin, J., Rajpurkar, P., et al.: Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison. AAAI (2019)

work page 2019

Show all 17 references

[9]

AMIA Jt Summits Transl Sci Proc (2018) 188–196

Peng, Y., Wang, X., Lu, L., Bagheri, M., Summers, R., Lu, Z.: Negbio: a high- performance tool for negation and uncertainty detection in radiology reports. AMIA Jt Summits Transl Sci Proc (2018) 188–196

2018
[10]

RadioGraphics 37(5) (2017) 1388–1400

Burrowes, D.P., Medellin, A., Harris, A.C., Milot, L., Wilson, S.R.: Contrast- enhanced us approach to the diagnosis of focal liver masses. RadioGraphics 37(5) (2017) 1388–1400

2017
[11]

In: IEEE international conference on computer vision

Tran, D., Bourdev, L., Fergus, R., Torresani, L., Paluri, M.: Learning spatiotem- poral features with 3d convolutional networks. In: IEEE international conference on computer vision. (2015) 4489–4497

2015
[12]

Wein, H.S., Bredno, J., Lehmann, T.M.: Quality of dicom header information for image categorization

Gueld, M.O., Kohnen, M., Keysers, D., andBerthold B. Wein, H.S., Bredno, J., Lehmann, T.M.: Quality of dicom header information for image categorization. In: Proceedings of SPIE Medical Imaging. (2002) Data Curation for Patients: Phase Recognition 9

2002
[13]

In: IEEE CVPR

Hara, K., Kataoka, H., Satoh, Y.: Learning spatio-temporal features with 3d resid- ual networks for action recognition. In: IEEE CVPR. (2017) 3154–3160

2017
[14]

In: IEEE conference on computer vision and pattern recognition

Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: IEEE conference on computer vision and pattern recognition. (2018) 7132–7141

2018
[15]

In: Proceedings of the 18th Conference on Computational Linguistics - Volume 2

Yeh, A.: More accurate tests for the statistical signiﬁcance of result diﬀerences. In: Proceedings of the 18th Conference on Computational Linguistics - Volume 2. COLING ’00, Stroudsburg, PA, USA (2000) 947–953

2000
[16]

Scandinavian Journal of Statistics 6 (1979) 65–70

Holm, S.: A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics 6 (1979) 65–70

1979
[17]

arterial

Zhao, G., Zhou, B., Wang, K., Jiang, R., Xu, M.: Respond-cam: Analyzing deep models for 3d imaging data by visualizations. In: Medical Image Computing and Computer-Assisted Intervention, Springer (2018) 485–492 CT Data Curation for Liver Patients: Phase Recognition in Dynamic ...

2018 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Medical Image Analysis 42 (2017) 60–88

Litjens, G.J.S., Kooi, T., Bejnordi, B.E., Setio, A.A.A., Ciompi, F., Ghafoorian, M., van der Laak, J.A.W.M., van Ginneken, B., Snchez, C.I.: A survey on deep learning in medical image analysis. Medical Image Analysis 42 (2017) 60–88

work page 2017

[2] [2]

In: Asian Conference on Computer Vision, Springer (2018) 298–313

Zhou, B., Lin, X., Eck, B., Hou, J., Wilson, D.: Generation of virtual dual energy images from standard single-shot radiographs using multi-scale and conditional adversarial network. In: Asian Conference on Computer Vision, Springer (2018) 298–313

work page 2018

[3] [3]

In: IEEE CVPR

Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L.: Imagenet: A large-scale hierarchical image database. In: IEEE CVPR. (2009) 248–255

work page 2009

[4] [4]

In: Journal of Digital Imaging

Kohli, M.D., Summers, R.M., Geis, J.R.: Medical image data and datasets in the era of machine learning: Whitepaper from the 2016 c-mimi meeting dataset session. In: Journal of Digital Imaging. (2017)

work page 2017

[5] [5]

In: Artiﬁcial Intelligence in Medical Imaging

Harvey, H., Glocker, B.: A standardised approach for preparing imaging data for machine learning tasks in radiology. In: Artiﬁcial Intelligence in Medical Imaging. Springer (2019) 61–72

work page 2019

[6] [6]

Journal of Medical Imaging 5(3) (2018) 036501

Yan, K., Wang, X., Lu, L., Summers, R.M.: Deeplesion: automated mining of large-scale lesion annotations and universal lesion detection with deep learning. Journal of Medical Imaging 5(3) (2018) 036501

work page 2018

[7] [7]

In: 2019 IEEE Winter Conference on Applications of Computer Vision (WACV), IEEE (2019) 1–10

Zhou, B., Chen, A., Crawford, R., Dogdas, B., Goldmarcher, G.: A progressively- trained scale-invariant and boundary-aware deep neural network for the automatic 3d segmentation of lung lesions. In: 2019 IEEE Winter Conference on Applications of Computer Vision (WACV), IEEE (2019) 1–10

work page 2019

[8] [8]

AAAI (2019)

Irvin, J., Rajpurkar, P., et al.: Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison. AAAI (2019)

work page 2019