Less Annotation, More Interpretation: Prior-Guided Concept Bottleneck Models for Interpretable Cancer Imaging Diagnosis

Baoqiang Ma; Kenneth Gilhuijs

REVIEW 3 major objections 5 minor 19 references

Less Annotation, More Interpretation: Prior-Guided Concept Bottleneck Models for Interpretable Cancer Imaging Diagnosis

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A prior-guided hybrid concept bottleneck model improves mean concept AUC in the 0–20% annotation regime across three cancer imaging tasks while keeping diagnostic AUC close to black-box baselines.

desk verdict Solid empirical study of a prior-guided hybrid CBM, but the 'less annotation' headline is conditional on aggregate priors computed from full training labels, which the standard baseline does not receive. read the letter →

arxiv 2608.13148 v1 pith:TBC6S7EX submitted 2026-08-13 cs.CV

Baoqiang Ma , Kenneth Gilhuijs This is my paper · ORCID

classification cs.CV

keywords conceptbottleneckmodelsinterpretablemedicalimaginglimitedsupervisionclass-conditionalpriorsannotationefficiencycancerdiagnosismammographypulmonarynodules

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 claims that dense, expensive per-patient concept annotations are unnecessary for interpretable cancer diagnosis. By adding class-conditional concept prevalence priors and a clinically initialized concept-to-diagnosis head to a concept bottleneck model, it achieves substantially higher concept detection AUC than a standard CBM when only 0–20% of patients have concept labels, while keeping diagnostic AUC close to black-box performance. The reported gains at 10% annotation are large for the three studied tasks—mass concept AUC rises from 0.619 to 0.741, calcifications from 0.650 to 0.787, and lung nodules from 0.597 to 0.642—so the practical argument is that a small annotation set plus aggregate clinical statistics suffices for transparent tumor diagnosis.

What carries the argument

The central object is the prior-guided hybrid CBM, a concept bottleneck model whose concept head $g_\theta(x)$ maps images to concept probabilities $\hat{c}=\sigma(z)$, followed by a linear diagnosis head. Three mechanisms carry the argument: (1) a distribution-matching loss $L_{\text{dist}}$ that compares the mean predicted concept vector of unannotated patients in a minibatch to a class-conditional prevalence prior $\pi_y$, giving weak supervision from aggregate label proportions; (2) initialization of the diagnosis head weight for concept $k$ as $\logit\{P(y=1\mid c_k=1)\}-\logit\{P(y=1\mid c_k=0)\}$, with bias from the malignancy base rate; and (3) an optional L2 anchor $L_{\text{reg}}$ keeping the head near that initialization. The load-bearing mechanism is the stabilized concept-to-diagnosis mapping: with a plausible start, gradient updates teach image-to-concept mappings while the diagnosis head stays close to the prior, which the drift analysis shows does not happen with random initialization.

What would settle it

Compute the data-derived priors from only the annotated 10% of training patients (instead of the full training cohort) and rerun the low-annotation experiments; if the mean concept AUC advantage over the standard CBM shrinks to near zero or reverses, the paper's central benefit is attributable to full-cohort aggregate labels rather than to the prior-guided mechanism itself.

Watch

Extended reading notes

Core claim

The central discovery is that a hybrid supervision scheme—mixing instance-level concept labels on a few patients with class-conditional concept distribution matching on the rest, plus a prior-initialized diagnosis head—lets concept bottleneck models learn fine-grained radiological concepts annotation-efficiently. The paper reports consistent mean concept ROC-AUC gains in the 0–20% annotation regime relative to a matched standard CBM, with the largest effect coming from prior initialization of the concept-to-diagnosis head rather than from the distribution loss alone. Ablations show that the prior-initialized head drifts almost not at all (relative L2 drift under 0.5%) when no concept labels are available, while randomly initialized heads drift by more than 85%. The paper also shows that zero-shot vision-language models are insufficient for fine-grained tumor-level concept recognition, scoring around 0.50 AUC.

Load-bearing premise

The method's benefit depends on having reliable class-conditional concept prevalence estimates (the expected fraction of each radiological concept in benign versus malignant cases), and in the experiments these are computed from the full training set's concept labels even when instance-level supervision is only 0–20%; if those aggregate statistics are not available without near-full annotation, the claimed reduction in annotation burden weakens.

Editorial extensions

If this is right

A clinically meaningful concept bottleneck for cancer imaging can be trained with as little as 0–20% patient-level concept annotations, with concept AUC close to what fully annotated models give in the tested tasks.
The diagnosis head initialized from clinical priors is robust: at 0% annotation its weights drift by less than 0.5% relative, while randomly initialized heads drift by more than 85%, explaining why prior initialization matters most when labels are absent.
Zero-shot vision-language models do not yet provide reliable fine-grained tumor-level concept predictions, so annotation-free concept bottleneck models are not a viable alternative in this setting.
Correcting concept predictions at test time improves diagnosis AUC by up to 0.139 on lung nodules, indicating that selective clinician review of concept outputs could become a practical workflow.
Data-derived class-conditional priors outperform manual BI-RADS-style priors for concept supervision, showing that cohort-calibrated aggregate statistics are the stronger prior source.

Reading between the lines

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

If the prior-initialization effect generalizes, prevalence priors drawn from historical radiology reports or clinical databases could cold-start interpretable models for new cancer types where only outcome labels exist, without any new concept annotation.
The distribution-matching loss is learning from label proportions; a natural extension is estimating $\pi_y$ iteratively from the unannotated pool (e.g., via expectation-maximization), which would remove the need for full-cohort concept statistics and directly test how much aggregate information the method truly requires.
The oracle concept-correction gains suggest a prospective clinician-in-the-loop study: have radiologists correct only high-impact concept predictions, then measure real diagnostic AUC improvement over a no-correction baseline; the paper's oracle numbers bound the possible benefit.

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

Quick take: the core experiments are sound, but the headline 'less annotation' is not actually supported by the comparison as run. The priors that drive the gains are estimated from the full training cohort's concept labels, so the hybrid model receives aggregate label information at every annotation fraction, including 0%. The matched standard CBM baseline is denied that. That is a real information mismatch, and the stress-test note is right. The authors do acknowledge this in Section 4.6, but the abstract and conclusions present it as established rather than conditional.

What's new: the specific combination of instance-level concept loss, class-conditional distribution matching, and prior-based head initialization for low-annotation cancer imaging isn't in the prior work. The experiments are carefully done: bootstrap confidence intervals, paired significance tests, ablations, and released code. The finding that prior initialization of the diagnosis head is the main driver is plausible and backed by the drift analysis. The zero-shot VLM comparison is a useful negative result; those models are not ready for fine-grained tumor-level concepts.

Soft spots: the main one is the annotation-burden claim. Since data priors require aggregate concept statistics from the full training set, the method does not yet show it can reduce annotation effort from scratch. It shows that if you already have cohort-level concept prevalence, you can train with fewer instance-level labels. That is still useful, but it's a different claim. The manual prior sweep in Fig. S1 shows weaker gains, which underscores the point. Also, the closest prior work (Nahiduzzaman et al.) is cited but not used as a baseline, so the marginal contribution over class-level prior supervision alone is not quantified. Minor points: no per-concept error analysis, and the concept-correction experiment is oracle, as the authors note.

Overall: the paper is honest and the engineering is solid. The central mechanism — anchoring the concept-to-diagnosis head to a clinical prior — is reasonable, and the empirical evidence is consistent with it. The 'less annotation' claim needs to be reframed as 'less instance-level annotation given aggregate priors.' I'd send this to peer review; a good reviewer can push them to match the information budget and add the missing baseline. For a reading group it's a useful data point in the CBM literature, though not a must-read.

Referee Report

3 major / 5 minor

Summary. The paper proposes a prior-guided hybrid concept bottleneck model (PriorHybrid-CBM) that combines limited instance-level concept annotations with class-conditional distribution matching on unannotated patients and prior initialization of the concept-to-diagnosis head. The method is evaluated on CBIS-DDSM mammographic masses and calcifications and LIDC-IDRI pulmonary nodules across concept annotation fractions from 0% to 100%. The central empirical claim is that in the clinically relevant 0-20% annotation regime, the hybrid CBM consistently improves mean concept ROC-AUC over a matched standard CBM while maintaining diagnostic AUC close to black-box models, with reported 10%-annotation improvements from 0.619 to 0.741 (masses), 0.650 to 0.787 (calcifications), and 0.597 to 0.642 (nodules). The paper also reports that zero-shot vision-language models are insufficient for fine-grained tumor-level concept prediction.

Significance. If the reported gains hold under a properly matched information comparison, the method would be a meaningful step toward reducing concept-annotation costs for interpretable medical imaging. The empirical work has notable strengths: percentile bootstrap confidence intervals, paired significance tests, patient-level splitting, a component ablation, and publicly released code. The paper also provides a clear negative result for zero-shot VLM transfer. However, the central 'less annotation' claim currently rests on a comparison in which the hybrid model receives aggregate concept-label statistics from the full training cohort, while the standard CBM baseline does not. This limits the validity of the headline conclusion as stated.

major comments (3)

[Section 4.6, Table 1] The principal concern raised in the review process is confirmed by the manuscript text. Section 3.2 states that 'data-derived aggregate priors were computed separately from the training patients of each fold' and that annotation fractions 'applied only to labels used in the instance-level concept loss.' As a result, at every annotation fraction, including 0%, the hybrid model receives class-conditional concept prevalence vectors pi_y (Eq. 3) and P(y|c_k) (Eq. 4) estimated from the concept labels of the entire training cohort, whereas the matched standard CBM receives none of this aggregate label information. The reported concept-AUC gains in the 0-20% regime, including the 10% figures in the abstract, therefore conflate the proposed mechanism with an information advantage. The authors should either re-run the experiments with priors estimated only from the annotated subset (or from a separate external source) to provide a truly matched comparison, or re-frame the central claim as conditional on the availability of full-training-set aggregate priors. As it stands, the 'less annotation' claim is not supported by the current experimental design.
[Section 4.4, Section 4.6] The ablation titled 'Random init.' removes prior initialization but retains the distribution-matching loss (lambda_d) and the anchoring loss (lambda_r). Since the distribution-matching targets are data-derived priors computed from the full training cohort, this ablation does not isolate the effect of prior initialization from the use of full-cohort aggregate label information. The observation in Section 4.4 that random-init diagnosis heads drift far from the data-prior direction is made at 0% concept annotation, where the standard CBM has no concept supervision at all; the comparison is therefore not information-matched. The conclusion that 'prior initialization is the main component' needs to be re-stated with the caveat that this component is evaluated in the presence of full-training-set prior statistics.
[Supplementary Fig. S1, Section 4.6] The paper's own results show that data-derived priors generally outperform manual priors, which are label-free. For example, in the 0-20% regime, mean concept AUC improves from 0.681 to 0.716 for masses and from 0.726 to 0.757 for calcifications when moving from manual to data-derived priors. This implies that the practical annotation-reduction benefit depends on having access to cohort-level label statistics, which may themselves require substantial annotation effort. The limitation section (4.6) acknowledges this, but the abstract and conclusion state the annotation reduction as established rather than conditional on such statistics. The framing should be corrected to match the evidence.

minor comments (5)

[Introduction] Typo: 'Several approached have been proposed' should be 'Several approaches have been proposed.'
[Section 2.2] Typo: 'specific dataset' is written as 'specifc dataset.'
[Section 3.1] The nodule crop size is written as '483' without a superscript; it should be 4^3 (64) or the intended notation should be clarified.
[Table captions] The table captions contain 'T able' rather than 'Table'; please fix the formatting.
[Section 3.2] The selection of lambda_d is performed using validation concept AUC over the 0-20% regime. This is a second, smaller use of concept labels beyond the training objective and should be mentioned when summarizing the label budget of the method.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the empirical comparison is self-contained and evaluated on held-out test data.

full rationale

No circularity found. The paper's central claim is an empirical comparison of concept AUC on held-out test sets, and no derivation is claimed that would reduce to its inputs. The prior vectors pi_y in Eq. (3) and P(y|c_k) in Eq. (4) are inputs estimated from the training folds; they are not fitted to test labels, and the held-out test evaluation is external. The only caveat is that data-derived priors are computed from all training patients' concept labels, which means the hybrid model receives aggregate label information that the standard CBM does not; this weakens the 'reduced annotation burden' framing but is not circularity. The paper explicitly discloses this in Section 4.6, stating that 'any reduction in clinical workload is conditional on a suitable prior being available.' Self-citations [10,11] are supporting context, not load-bearing. Therefore the derivation chain is self-contained with no circular step.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the availability and accuracy of class-conditional concept priors, which are domain assumptions, and on standard optimization assumptions. The model introduces no new entities. The two tunable weights (λ_d, λ_r) are selected by validation, which is standard practice but should be noted.

free parameters (3)

λ_d (distribution matching weight) = 0.7 (mass), 0.3 (calcification), 1.0 (nodule)
Selected per task by mean five-fold validation concept AUC over the 0-20% annotation regime (Section 3.2).
λ_r (anchor regularization weight) = 0.05
Weak anchoring chosen for the main configuration (Section 2.3); ablation removes it.
λ_c (concept loss weight) = not reported (assumed 1)
Paper does not specify the value of λ_c in the hybrid objective (Eq. 2); implicitly set to 1.

assumptions (4)

domain assumption Concept labels are reliable and the predefined concepts capture clinically relevant attributes.
Section 3.1 defines concepts from BI-RADS and LIDC thresholds; the method assumes these labels are meaningful and correctly annotated.
domain assumption Class-conditional priors π_y can be estimated accurately from training data and generalize to the test distribution.
Section 2.2 uses these priors as supervision; if priors are biased or unrepresentative, the distribution matching loss may misguide training.
domain assumption A linear diagnosis head on concept probabilities is sufficient to represent the concept-to-diagnosis mapping.
Section 2.1 uses a linear head on concept probabilities; this is the concept bottleneck assumption underlying the interpretability claim.
standard math Standard optimization and loss functions (BCE with logits, L2 regularization) behave as expected.
Used throughout the method; no unusual mathematical assumptions.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Less Annotation, More Interpretation: Prior-Guided Concept Bottleneck Models for Interpretable Cancer Imaging Diagnosis." pith.science (2026). https://pith.science/paper/TBC6S7EX

@misc{pith2026260813148,
  author       = {Pith},
  title        = {Pith review of: Less Annotation, More Interpretation: Prior-Guided Concept Bottleneck Models for Interpretable Cancer Imaging Diagnosis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TBC6S7EX}},
  note         = {Machine review of arXiv:2608.13148}
}

read the original abstract

Concept bottleneck models (CBMs) can improve the transparency of cancer image diagnostic prediction by expressing predictions through radiological concepts. However, their dependence on instance-level concept annotations limits practical applicability. We propose a prior-guided hybrid CBM that integrates limited concept annotations, class-conditional concept distribution matching on unannotated patients, and prior initialization of the concept-to-diagnosis head. We evaluate the method on CBIS-DDSM mammographic masses and calcifications and LIDC-IDRI pulmonary nodules across 0-100% concept annotation. In the clinically relevant 0-20% annotation regime, the hybrid CBM consistently improves mean concept AUC over a matched standard CBM, while maintaining diagnostic performance close to black-box models. At 10% annotation specifically, concept AUC increases from 0.619 to 0.741 for masses, from 0.650 to 0.787 for calcifications, and from 0.597 to 0.642 for pulmonary nodules. Ablation experiments identify prior initialization as the main component contributing to improved concept detection, likely by stabilizing the concept-to-diagnosis head. Zero-shot VLMs remain insufficient for reliable fine-grained tumor-level concept prediction. These findings suggest that structured priors can substantially reduce the annotation burden of interpretable cancer imaging models.

Figures

Figures reproduced from arXiv: 2608.13148 by the authors.

**Figure 1.** Overview of PriorHybrid-CBM. A task-specific image backbone predicts radiological concepts, which are used by a linear diagnosis head. Patients with concept annotations receive instance-level concept supervision; patients without concept annotations receive class-conditional prior supervision. Prior concept knowledge also initializes (with optional anchors) the concept-to-diagnosis head. 4. Show that current modal… view at source ↗

**Figure 2.** Concept annotation efficiency. Mean concept ROC-AUC is reported for the prior-guided hybrid CBM, matched standard CBM, and zero-shot VLM reference. Error bars denote percentile bootstrap 95% confidence intervals. Stars denote paired bootstrap significance for hybrid CBM versus standard CBM. The zero-shot VLM references were consistently weaker for concept prediction, achieving AUCs around 0.50. Therefore, prompt-bas… view at source ↗

**Figure 3.** Diagnostic annotation efficiency. Diagnosis ROC-AUC is shown for the priorguided hybrid CBM, matched standard CBM, black-box reference, and zero-shot VLM reference. Error bars denote percentile bootstrap 95% confidence intervals for trained CBM models; the black-box reference is shown as a horizontal line with bootstrap confidence band [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 13 canonical work pages

[1]

Nature Medicine25, 954–961 (2019)

Ardila, D., Kiraly, A.P., Bharadwaj, S., Choi, B., Reicher, J.J., Peng, L., Tse, D., Etemadi, M., Ye, W., Corrado, G., Naidich, D.P., Shetty, S.: End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nature Medicine25, 954–961 (2019). https://doi.org/10.1038/ s41591-019-0447-x

work page 2019
[2]

Medical Physics38(2), 915–931 (2011)

Armato, S.G., et al.: The lung image database consortium (LIDC) and image database resource initiative (IDRI): A completed reference database of lung nodules on CT scans. Medical Physics38(2), 915–931 (2011)

work page 2011
[3]

arXiv preprint arXiv:1904.00625 (2019)

Chen, S., Ma, K., Zheng, Y.: Med3D: Transfer learning for 3d medical image analysis. arXiv preprint arXiv:1904.00625 (2019)

arXiv 2019
[4]

In: Medical Image Computing and Computer Assisted Intervention

Ghosh, S., Poynton, C.B., Visweswaran, S., Batmanghelich, K.: Mammo-CLIP: A vision language foundation model to enhance data efficiency and robustness in mammography. In: Medical Image Computing and Computer Assisted Intervention. pp. 632–642 (2024)

work page 2024
[5]

Nature Biomedical Engineering (2025).https://doi.org/10.1038/s41551-025-01599-y

Hamamci, I.E., Er, S., Wang, C., et al.: Developing generalist foundation mod- els from a multimodal dataset for 3d computed tomography. Nature Biomedical Engineering (2025).https://doi.org/10.1038/s41551-025-01599-y

doi:10.1038/s41551-025-01599-y 2025
[6]

arXiv preprint arXiv:2406.18992 (2024)

Hu, L., Huang, T., Xie, H., Gong, X., Ren, C., Hu, Z., Yu, L., Ma, P., Wang, D.: Semi-supervised concept bottleneck models. arXiv preprint arXiv:2406.18992 (2024)

arXiv 2024
[7]

In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

Huang, G., Liu, Z., van der Maaten, L., Weinberger, K.Q.: Densely connected convolutional networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 4700–4708 (2017).https://doi.org/10.1109/ CVPR.2017.243

work page 2017
[8]

In: Proceedings of the 37th International Conference on Machine Learning

Koh, P.W., Nguyen, T., Tang, Y.S., Mussmann, S., Pierson, E., Kim, B., Liang, P.: Concept bottleneck models. In: Proceedings of the 37th International Conference on Machine Learning. pp. 5338–5348 (2020)

work page 2020

Show all 19 references

[9]

Scientific Data4, 170177 (2017) Prior-Guided CBMs for Cancer Imaging 11

Lee, R.S., Gimenez, F., Hoogi, A., Miyake, K.K., Gorovoy, M., Rubin, D.L.: A curated mammography data set for use in computer-aided detection and diagnosis research. Scientific Data4, 170177 (2017) Prior-Guided CBMs for Cancer Imaging 11

2017
[10]

In: 23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026, London, United Kingdom, April 8–11, 2026

Ma, B., Madzia-Madzou, D.K., Kraaijveld, R.C.J., Ouyang, J.: Ranking XAI methods for head and neck cancer outcome prediction. In: 23rd IEEE International Symposium on Biomedical Imaging, ISBI 2026, London, United Kingdom, April 8–11, 2026. pp. 1–4. IEEE (2026). https://doi.org...

2026
[11]

Ma, B., Madzia-Madzou, D.K., Ouyang, J., Gilhuijs, K.G.A.: Can vision-language models enable more efficient concept-based learning with less supervision for in- terpretable lung nodule diagnosis? In: MIDL 2026 Short Papers (2026),https: //openreview.net/forum?id=OyOHg986R5, poster

2026
[12]

Nature577, 89–94 (2020).https://doi.org/10.1038/s41586-019-1799-6

McKinney, S.M., Sieniek, M., Godbole, V., Godwin, J., Antropova, N., Ashrafian, H., Back, T., Chesus, M., Corrado, G.S., Darzi, A., Etemadi, M., Garcia-Vicente, F., et al.: International evaluation of an AI system for breast cancer screening. Nature577, 89–94 (2020).https://do...

2020 doi
[13]

arXiv preprint arXiv:2511.01131 (2025)

Nahiduzzaman, M., Korevaar, S., Bab-Hadiashar, A., Tennakoon, R.: Weakly supervised concept learning with class-level priors for interpretable medical diagnosis. arXiv preprint arXiv:2511.01131 (2025)

2025
[14]

In: International Conference on Learning Representations (2023)

Oikarinen, T., Das, S., Nguyen, L.M., Weng, T.W.: Label-free concept bottleneck models. In: International Conference on Learning Representations (2023)

2023
[15]

Computers in Biology and Medicine198, 111145 (2025)

Patrício, C., Rio-Torto, I., Cardoso, J.S., Teixeira, L.F., Neves, J.C.: CBVLM: Training-free explainable concept-based large vision language models for medical image classification. Computers in Biology and Medicine198, 111145 (2025). https://doi.org/10.1016/j.compbiomed.2025.111145

2025
[16]

In: Proceedings of the 26th International Conference on Machine Learning

Quadrianto, N., Smola, A.J., Caetano, T.S., Le, Q.V.: Estimating labels from label proportions. In: Proceedings of the 26th International Conference on Machine Learning. pp. 776–783 (2009)

2009
[17]

In: Proceedings of the 38th International Conference on Machine Learning

Radford, A., et al.: Learning transferable visual models from natural language supervision. In: Proceedings of the 38th International Conference on Machine Learning. pp. 8748–8763 (2021)

2021
[18]

In: Proceedings of the IEEE International Conference on Computer Vision

Selvaraju, R.R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., Batra, D.: Grad- CAM: Visual explanations from deep networks via gradient-based localization. In: Proceedings of the IEEE International Conference on Computer Vision. pp. 618–626 (2017)

2017
[19]

In: Proceedings of the 34th International Conference on Machine Learning

Sundararajan, M., Taly, A., Yan, Q.: Axiomatic attribution for deep networks. In: Proceedings of the 34th International Conference on Machine Learning. pp. 3319–3328 (2017) 12 B. Ma and K. Gilhuijs A Complete Hybrid-CBM Prior and Distribution Sweep 0 5 10 20 50 75 100 0.5 0.6 ...

2017

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

Nature Medicine25, 954–961 (2019)

Ardila, D., Kiraly, A.P., Bharadwaj, S., Choi, B., Reicher, J.J., Peng, L., Tse, D., Etemadi, M., Ye, W., Corrado, G., Naidich, D.P., Shetty, S.: End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nature Medicine25, 954–961 (2019). https://doi.org/10.1038/ s41591-019-0447-x

work page 2019

[2] [2]

Medical Physics38(2), 915–931 (2011)

Armato, S.G., et al.: The lung image database consortium (LIDC) and image database resource initiative (IDRI): A completed reference database of lung nodules on CT scans. Medical Physics38(2), 915–931 (2011)

work page 2011

[3] [3]

arXiv preprint arXiv:1904.00625 (2019)

Chen, S., Ma, K., Zheng, Y.: Med3D: Transfer learning for 3d medical image analysis. arXiv preprint arXiv:1904.00625 (2019)

arXiv 2019

[4] [4]

In: Medical Image Computing and Computer Assisted Intervention

Ghosh, S., Poynton, C.B., Visweswaran, S., Batmanghelich, K.: Mammo-CLIP: A vision language foundation model to enhance data efficiency and robustness in mammography. In: Medical Image Computing and Computer Assisted Intervention. pp. 632–642 (2024)

work page 2024

[5] [5]

Nature Biomedical Engineering (2025).https://doi.org/10.1038/s41551-025-01599-y

Hamamci, I.E., Er, S., Wang, C., et al.: Developing generalist foundation mod- els from a multimodal dataset for 3d computed tomography. Nature Biomedical Engineering (2025).https://doi.org/10.1038/s41551-025-01599-y

doi:10.1038/s41551-025-01599-y 2025

[6] [6]

arXiv preprint arXiv:2406.18992 (2024)

Hu, L., Huang, T., Xie, H., Gong, X., Ren, C., Hu, Z., Yu, L., Ma, P., Wang, D.: Semi-supervised concept bottleneck models. arXiv preprint arXiv:2406.18992 (2024)

arXiv 2024

[7] [7]

In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

Huang, G., Liu, Z., van der Maaten, L., Weinberger, K.Q.: Densely connected convolutional networks. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 4700–4708 (2017).https://doi.org/10.1109/ CVPR.2017.243

work page 2017

[8] [8]

In: Proceedings of the 37th International Conference on Machine Learning

Koh, P.W., Nguyen, T., Tang, Y.S., Mussmann, S., Pierson, E., Kim, B., Liang, P.: Concept bottleneck models. In: Proceedings of the 37th International Conference on Machine Learning. pp. 5338–5348 (2020)

work page 2020