Pith. sign in

REVIEW 3 major objections 4 minor 44 references

Text-Guided Refinement of Multi-sequence Glioma Subregion Segmentation with a Vision-Language Foundation Model

T0 review · 3 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read A 3D vision-language foundation model can be adapted into a text-guided contour editor for glioma subregions, with correct instructions improving segmentation Dice scores beyond blank and contradictory prompts in both internal and…

desk verdict A useful but oracle-limited adaptation: the instruction branch works when the prompt is generated from the ground-truth error, but the clinical editing claim is not yet established. read the letter →

arxiv 2608.05389 v1 pith:WVMETB47 submitted 2026-08-05 cs.CV

Zach Eidex , Yu-nong Lin , Mojtaba Safari , Sean Pitroda , Ralph Weichselbaum , Zhen Tian , Xiaofeng Yang This is my paper · ORCID
classification cs.CV
keywords braintumorsegmentationtext-guidedvision-languagefoundationmodelgliomasubregionscontourrefinementinstruction-specificediting3DmedicalimageDicesimilaritycoefficient
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 tries to establish that an existing 3D vision-language segmentation foundation model can be turned into a contour editor for glioma subregions by adding a small trainable instruction branch while keeping the image encoder and decoder frozen. The central claim is that the corrections are driven by the semantic content of the instruction: on an internal glioma test set, correct text prompts raised mean subregion Dice similarity coefficient (DSC) from 0.774 to 0.796, outperforming blank prompts (0.762) and contradictory prompts (0.770). In a cross-dataset transfer set, correct prompts raised DSC from 0.527 to 0.550 and widened the gap over contradictory prompts to 0.047 DSC. The authors conclude that text-guided refinement is instruction-specific rather than generic post-processing, supporting further evaluation as a clinician-in-the-loop tool for targeted brain tumor contour correction.

What carries the argument

The central object is the instruction branch: frozen Qwen text embeddings of the correction command are projected, processed by a copied VoxTell prompt transformer against the image bottleneck, and mapped into the same multi-scale decoder-conditioning spaces as the frozen base structure prompt. At each decoder scale the correction embedding is combined with the base prompt embedding by residual weighted addition, $\tilde h^k_\ell = b^k_\ell + \lambda i^k_\ell$, with $\lambda = \tanh(\gamma)$ initialized so the instruction contributes about 0.1. This design preserves the base prompt as the dominant anchor while letting the instruction steer the frozen decoder toward a local edit, and all other weights stay frozen.

What would settle it

Give the same frozen refinement model free-form correction instructions written by clinicians on the same internal and cross-dataset cases, and compare DSC against blank and contradictory prompts; the claim of instruction-specific refinement would be falsified if correct clinician prompts fail to beat both control conditions by a comparable margin.

Watch

Extended reading notes

Core claim

The paper claims that a lightweight adaptation of the pretrained VoxTell 3D vision-language foundation model can perform one-step, instruction-specific refinement of glioma subregion masks. A trainable projection injects frozen text embeddings of a correction instruction into the multi-scale decoder-conditioning space, so the frozen decoder can make a local edit while the base target prompt remains the semantic anchor. Correct instructions improved mean subregion DSC over pretrained VoxTell, blank prompts, and contradictory prompts, with statistically significant paired differences and qualitative examples showing the intended local expansion, shrinkage, addition, or removal.

Load-bearing premise

The evaluation assumes that oracle correction prompts generated from the discrepancy between the pretrained prediction and the ground-truth reference mask behave like real clinician instructions; the paper itself notes in the Discussion that it does not measure performance with clinician-authored prompts, and if clinician phrasing differs substantially, the demonstrated gains may not transfer.

Editorial extensions

If this is right

  • On the internal glioma test set, correct text instructions raised mean subregion DSC from 0.774±0.158 to 0.796±0.137, beating blank prompts by 0.035 DSC and contradictory prompts by 0.026 DSC with Holm-adjusted p<0.001.
  • Under cross-dataset transfer, correct prompts improved mean subregion DSC from 0.527±0.287 to 0.550±0.278 and produced a larger correct-versus-contradictory separation of 0.047 DSC.
  • Because blank and contradictory prompts pass through the same frozen model and instruction branch, the correct-minus-contradictory difference is evidence that refinement depends on the semantic content of the text rather than on nonspecific post-training.
  • As a refinement tool, text-guided VoxTell is competitive but not dominant: it is similar to T1c-only nnU-Net internally, exceeds it cross-dataset, and nearly matches multimodal T1c+T2-FLAIR nnU-Net, suggesting its value is as a controllable editor rather than a replacement for a strong automatic segmenter.
  • Cohort-stratified transfer results show correct text beat contradictory text in every cross-dataset subgroup, but the best method varied by disease type, so pooled transfer metrics should be supplemented by per-cohort evaluation.

Reading between the lines

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

  • The refinement model never receives the initial mask, only the image, target prompt, and correction text; this implies the instruction must describe the needed edit well enough for the model to reconstruct the current contour, which may explain why the benefit is largest when the base segmentation is poor.
  • The larger correct-versus-contradictory gap under dataset shift hints that instruction-guided correction is most valuable when the base segmenter is out of distribution; a testable extension would corrupt base predictions deliberately and check whether text correction recovers more DSC at lower initial quality.
  • Because only a small projection branch is trained, the same recipe could likely be ported to other structures supported by the underlying foundation model, provided their common errors can be expressed in the same structured vocabulary of target, action, location, imaging evidence, edit size, and preservation constraints.
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 / 4 minor

Summary. The paper adapts the pretrained 3D vision-language segmentation model VoxTell into a lightweight text-guided refinement framework for glioma subregion segmentation. Initial masks are produced by frozen VoxTell; a trainable instruction branch maps correction text into the model's decoder-conditioning space, and the decoder then produces corrected logits directly from the image and text (the initial mask is not an input). Correction prompts are oracle prompts built from the discrepancy between the initial prediction and the reference mask, encoding target subregion, action, location, imaging evidence, edit size, and preservation constraints. The authors report that correct prompts improve mean subregion DSC over pretrained VoxTell, blank prompts, and contradictory prompts, both on an internal BraTS-GLI test set and on a cross-dataset transfer set, and they interpret this as evidence of instruction-specific, text-dependent contour editing. They also compare against nnU-Net and SAT baselines and include no-text, modality-specific, and training-size ablations.

Significance. If the central claim were established, the contribution would be practically relevant: a frozen 3D foundation model can be augmented with a small trainable text-conditioning branch, avoiding expensive full finetuning, and may offer a clinician-controllable correction interface. The paper has genuine methodological strengths: case-level paired t-tests with Holm-Bonferroni correction and paired Cohen's dz, clear reporting of effect sizes, public data and splits, cross-cohort stratification, and a template-based prompting scheme that is reproducible in principle. However, the load-bearing evidence for 'instruction-specific refinement' currently rests on oracle prompts that leak the target error into the test input, and on comparisons against blank/contradictory prompts that may be out-of-distribution for the instruction branch. The Discussion explicitly concedes that clinician-authored prompts were not evaluated; because that concession concerns the central claim, the significance of the present results is that of a controlled proof-of-concept under oracle instructions rather than a demonstrated clinician-in-the-loop tool.

major comments (3)
  1. [§2.4, Eq. (9)] The correction prompts are generated from the discrepancy between the pretrained prediction and the reference mask during both training and inference. The template in Eq. (9) encodes target, action, location, imaging evidence, edit size, and preservation constraints, with location and size derived from the error voxels. At test time, therefore, the 'correct' instruction is an oracle description of the ground-truth error, and the input leaks information about the target segmentation. A clinician without the reference mask cannot produce such a prompt, so the central claim in the Abstract and §3.1—that correct text instructions drive instruction-specific refinement—is not established for realistic inputs. The Discussion's sentence 'does not measure performance with clinician-authored prompts' concedes this, but the limitation is load-bearing rather than incidental. A non-oracle evaluation, such as prompts generated from the initial prediction alone or prompts written by human readers, is needed, or the claims must be explicitly restricted to oracle-conditioned refinement.
  2. [§2.5, Table 1] As described, the instruction branch is trained only on correct oracle prompts; blank and contradictory prompts appear only at evaluation, and the no-instruction controls replace the text with a blank during training. The correct-vs-blank and correct-vs-contradictory comparisons, which are the primary evidence for 'text-dependent contour editing' in the Discussion, therefore compare in-distribution instructions against out-of-distribution text forms. The lower DSC for blank and contradictory prompts could reflect sensitivity to unfamiliar templates or mismatched text statistics rather than semantic understanding of the instruction. A cleaner control would train with all three prompt types, or with randomized action words, and then test on held-out cases; alternatively, the evaluation should include prompts matched in template but differing only in the action word under conditions the model has seen.
  3. [§2.1, Eq. (2)] The refinement model does not receive the initial mask; Eq. (2) maps the image, target prompt, and correction instruction directly to corrected logits, and the text states that 'the initial mask itself is not passed to the model.' As a result, the model is a text-conditioned re-segmenter rather than an editor of an existing contour, and the 'preserve other regions' language in the prompts cannot be enforced locally because the current contour is never observed. Because the instruction encodes location and edit size derived from the reference-informed error, the model is given most of the information needed to identify where to change without seeing the original mask. To make the refinement claim testable, the initial mask should be provided as an input channel, or the evaluation should measure the change restricted to the specified error region and verify that the rest of the contour is preserved.
minor comments (4)
  1. [§2.4, Figures 2–3] Figure 3 labels the input as 'T1c MRI', but the correct edema prompt shown reads 'where there is no convincing FLAIR hyperintensity.' This contradicts §2.4, which states that T1c-only ED prompts use general image-support language and that FLAIR-specific phrases are reserved for T2-FLAIR experiments. Please reconcile the prompt generation rule with the displayed examples or relabel the input modality.
  2. [Table 1 caption] The caption reads 'Holm-adjusted paired t-test p-value and paired Cohen's dz are reported for each metrics'; this should be 'for each metric.'
  3. [§2.5 and throughout] The model name is spelled inconsistently as 'VoxTell' and 'Voxtell' (for example, 'The Voxtell model was initialized' in §2.5). Please use a single consistent spelling.
  4. [Table 3 caption] The 'Best' rows are not defined in the caption, and the main text refers to 'modality-selected VoxTell text' while the table uses 'VoxTell text.' Please define the selection rule and clarify the notation.

Circularity Check

1 steps flagged · score 4.0 of 10

Oracle correction prompts are derived from the test reference mask, so the headline refinement gain is partially an artifact of label leakage rather than a self-contained prediction.

  1. fitted input called prediction [Section 2.4 (Prompt Design), Eq. 9; Section 2.1 Eq. 1-3; Discussion limitations]
    "Correction prompts were generated from the disagreement between the pretrained VoxTell prediction and the reference segmentation during both model training and inference. ... A correction instruction u k is then generated from the discrepancy between the initial prediction p k 0 and the reference mask y k. ... does not measure performance with clinician-authored prompts."

    At test time, the correction instruction u is a deterministic function of the reference label y (action from false-negative/false-positive counts, location from error centroid, edit size from error volume). The headline endpoint DSC(G(x,t,u), y) therefore measures the model against the same label used to construct its input. Training on the same oracle-prompt distribution makes the 'correct improves over pretrained' result an expected consequence of the protocol, not an independent prediction of refinement. The blank/contradictory controls are weaker than they appear because both are generated from y or fall outside the training distribution of correct prompts.

full rationale

The only substantial circularity concern is the oracle-prompt protocol. The paper's own equations show that the correction instruction is constructed from the discrepancy between the initial prediction and the reference mask, and the same reference mask is then used to compute DSC. This is a genuine label-leakage issue: the 'correct' test input contains information about the target segmentation by construction, so the reported improvement over pretrained VoxTell is partly a measure of how well the model follows an oracle description rather than an independent test of general text-guided refinement. The correct-versus-contradictory comparison provides some independent evidence of semantic sensitivity, because contradictory prompts reverse the action while preserving the template and local region; however, those prompts are generated from the same reference-derived error information and are outside the distribution used for training. The paper explicitly acknowledges the clinician-prompt limitation, which is commendable but does not remove the input-label coupling. No load-bearing self-citations, imported uniqueness theorems, or ansatz-by-citation patterns were found. The central architectural claim remains an empirical benchmark with a real but partial circularity, so a score of 4 is appropriate rather than a higher score.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The framework rests on the pretrained VoxTell model and on oracle prompts derived from ground-truth segmentation errors. The only hand-chosen numeric parameters are the instruction-fusion initialization and the edit-size bins; no new physical entities are introduced.

free parameters (2)
  • Initial instruction scale lambda_0 = ~0.1 (via gamma initialization)
    The fusion scalar lambda = tanh(gamma) is initialized so the instruction branch contributes about 0.1 relative to the base prompt; this is a hand-chosen hyperparameter.
  • Edit size thresholds = small <3 cc, medium 3-12 cc, large >=12 cc
    The prompt template discretizes requested edit size into three hand-chosen volume ranges; this affects prompt wording and the difficulty of the generated instructions.
assumptions (4)
  • domain assumption VoxTell v1.1 provides a valid frozen 3D vision-language representation for brain tumor segmentation.
    The entire refinement method builds on VoxTell's frozen encoder, decoder, and text-conditioning pathway; poor VoxTell representations would undermine the framework. Section 2.3.
  • domain assumption Oracle prompts derived from the pretrained-vs-reference discrepancy are a valid proxy for real clinician correction instructions.
    The training and evaluation prompts encode the exact location, action, size, and evidence of the error from ground truth; the authors acknowledge this is not clinician-authored language. Section 2.4 and Discussion.
  • domain assumption BraTS-GLI reference segmentations are accurate ground truth.
    All metrics and prompt generation depend on these labels; label noise would bias the oracle prompts. Section 2.2.
  • standard math Paired t-tests on case-level DSC differences are valid.
    The statistical inference assumes approximately normal paired differences; n=250 internal and n=100 cross-dataset support the normality assumption under the CLT. Section 2.6.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Text-Guided Refinement of Multi-sequence Glioma Subregion Segmentation with a Vision-Language Foundation Model." pith.science (2026). https://pith.science/paper/WVMETB47

@misc{pith2026260805389,
  author       = {Pith},
  title        = {Pith review of: Text-Guided Refinement of Multi-sequence Glioma Subregion Segmentation with a Vision-Language Foundation Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WVMETB47}},
  note         = {Machine review of arXiv:2608.05389}
}
abstract

Background: Accurate glioma subregion delineation is important for radiotherapy planning and longitudinal monitoring, but manual contour correction is time-consuming. Models such as nnU-Net may generalize imperfectly and lack clinician-directed text correction. Purpose: We investigated adapting a three-dimensional (3D) vision-language foundation model for text-guided brain tumor segmentation refinement. Methods: We developed a lightweight VoxTell-based framework. Pretrained VoxTell generated initial masks. Oracle prompts derived from segmentation errors encoded target, action, location, imaging evidence, edit size, and preservation constraints. Frozen Qwen/VoxTell prompt embeddings were injected through trainable projections into its multiscale decoder conditioning; other weights remained frozen. Training, validation, and testing used 901, 100, and 250 BraTS-GLI cases. Cross-dataset transfer was evaluated on 100 meningioma, metastasis, pediatric tumor, and UPENN-GBM cases. Results: On the internal test set using post-contrast T1-weighted input, correct instructions improved subregion Dice similarity coefficient (DSC; enhancing tumor, edema, and necrotic/non-enhancing core) from $0.774\pm0.158$ to $0.796\pm0.137$. They outperformed blank prompts ($0.762\pm0.155$; Holm-adjusted $p<0.001$, $d_z=0.71$) and contradictory prompts ($0.770\pm0.163$; $p<0.001$, $d_z=0.48$). In cross-dataset testing, correct instructions improved DSC from $0.527\pm0.287$ to $0.550\pm0.278$ and outperformed contradictory instructions ($0.504\pm0.275$; $p<0.001$, $d_z=0.43$). Conclusion: A 3D vision-language foundation model can perform instruction-guided refinement of glioma subregion segmentations. Sensitivity to correct, blank, and contradictory prompts suggests text-dependent contour editing rather than nonspecific post-processing, supporting further evaluation as a clinician-in-the-loop tool.

Figures

Figures reproduced from arXiv: 2608.05389 by the authors.

Figure 1
Figure 1. VoxTell text-guided refinement architecture. A frozen pretrained VoxTell pathway encodes the MRI patch and base structure prompt. A copied instruction branch encodes the correction command and maps it into the same multi-scale decoder-conditioning spaces. At each decoder scale, the structure prompt embedding and instruction embedding are fused by learned weighted addition, b k ℓ + λik ℓ , before being passed to the … view at source ↗
Figure 2
Figure 2. Internal GLI qualitative examples. Each row shows one patient and one target-specific correction instruction. All tumor compartments are shown together: NCR yellow, ED green, and ET red. Red boxes identify the common tumor zoom region used across methods. Target DSC values below the panels quantify the highlighted target response. 9 [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Cross-dataset qualitative examples. Correct text is compared with blank and contradictory text on the same image and target; SAT-Pro and T1c nnU-Net are included as foundation-model and automatic comparators. Red boxes identify the common tumor zoom region used across methods. The examples demonstrate instruction-sensitive behavior under dataset shift. 3.2 Comparison with Baseline Methods Text-guided VoxTell was eva… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 37 canonical work pages

  1. [1]

    Nature Methods , volume=

    nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation , author=. Nature Methods , volume=. 2021 , doi=

    work page 2021

  2. [2]

    Scientific Data , volume=

    The University of Pennsylvania glioblastoma (UPenn-GBM) cohort: advanced MRI, clinical, genomics, and radiomics , author=. Scientific Data , volume=. 2022 , publisher=

    work page 2022

  3. [3]

    Medical Physics , volume=

    Deep learning based apparent diffusion coefficient map generation from multi-parametric MR images for patients with diffuse gliomas , author=. Medical Physics , volume=. 2025 , doi=

    work page 2025

  4. [4]

    2023 , eprint=

    Deep Learning in MRI-guided Radiation Therapy: A Systematic Review , author=. 2023 , eprint=

    work page 2023

  5. [5]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Segment Anything , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  6. [6]

    Nature Communications , volume=

    Segment anything in medical images , author=. Nature Communications , volume=. 2024 , doi=

    work page 2024

  7. [7]

    npj Digital Medicine , volume=

    Large-vocabulary segmentation for medical images with text prompts , author=. npj Digital Medicine , volume=. 2025 , doi=

    work page 2025

  8. [8]

    2025 , eprint=

    VoxTell: Free-Text Promptable Universal 3D Medical Image Segmentation , author=. 2025 , eprint=

    work page 2025

Show all 44 references
  1. [9]

    IEEE Transactions on Medical Imaging , volume=

    The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) , author=. IEEE Transactions on Medical Imaging , volume=. 2015 , doi=

    2015

  2. [10]

    arXiv preprint arXiv:1811.02629 , year=

    Identifying the best machine learning algorithms for brain tumor segmentation, progression assessment, and overall survival prediction in the BRATS challenge , author=. arXiv preprint arXiv:1811.02629 , year=

    arXiv

  3. [11]

    International Journal of Radiation Oncology Biology Physics , volume=

    NRG Oncology Assessment of Artificial Intelligence Deep Learning-Based Auto-segmentation for Radiation Therapy: Current Developments, Clinical Considerations, and Future Directions , author=. International Journal of Radiation Oncology Biology Physics , volume=. 2024 , doi=

    2024

  4. [12]

    SPIE Medical Imaging , year=

    Deep Learning-based Multi-Organ CT Segmentation with Adversarial Data Augmentation , author=. SPIE Medical Imaging , year=

  5. [13]

    arXiv preprint arXiv:2412.15115 , year=

    Qwen2.5 Technical Report , author=. arXiv preprint arXiv:2412.15115 , year=

    arXiv

  6. [14]

    Medical Image Computing and Computer-Assisted Intervention , pages=

    U-Net: Convolutional Networks for Biomedical Image Segmentation , author=. Medical Image Computing and Computer-Assisted Intervention , pages=

  7. [15]

    IEEE Transactions on Medical Imaging , volume=

    Brain Tumor Segmentation Using Convolutional Neural Networks in MRI Images , author=. IEEE Transactions on Medical Imaging , volume=. 2016 , doi=

    2016

  8. [16]

    Medical Image Analysis , volume=

    Brain tumor segmentation with Deep Neural Networks , author=. Medical Image Analysis , volume=. 2017 , doi=

    2017

  9. [17]

    International Conference on 3D Vision , pages=

    V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation , author=. International Conference on 3D Vision , pages=. 2016 , doi=

    2016

  10. [18]

    Medical Image Computing and Computer-Assisted Intervention , pages=

    3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation , author=. Medical Image Computing and Computer-Assisted Intervention , pages=

  11. [19]

    Medical Image Analysis , volume=

    Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation , author=. Medical Image Analysis , volume=. 2017 , doi=

    2017

  12. [20]

    Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries , pages=

    3D MRI brain tumor segmentation using autoencoder regularization , author=. Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries , pages=

  13. [21]

    Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

    UNETR: Transformers for 3D Medical Image Segmentation , author=. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

  14. [22]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Self-supervised pre-training of Swin Transformers for 3D medical image analysis , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  15. [23]

    Complex & Intelligent Systems , volume=

    Deep learning based brain tumor segmentation: a survey , author=. Complex & Intelligent Systems , volume=. 2023 , doi=

    2023

  16. [24]

    Nature Communications , volume=

    The Medical Segmentation Decathlon , author=. Nature Communications , volume=. 2022 , doi=

    2022

  17. [25]

    Radiology: Artificial Intelligence , volume=

    TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images , author=. Radiology: Artificial Intelligence , volume=. 2023 , doi=

    2023

  18. [26]

    Medical Imaging with Deep Learning , year=

    SAM.MD: Zero-shot medical image segmentation capabilities of the Segment Anything Model , author=. Medical Imaging with Deep Learning , year=

  19. [27]

    Medical Image Analysis , volume=

    Segment anything model for medical images? , author=. Medical Image Analysis , volume=. 2024 , doi=

    2024

  20. [28]

    Computers in Biology and Medicine , year=

    Segment Anything Model for Medical Image Segmentation: Current Applications and Future Directions , author=. Computers in Biology and Medicine , year=

  21. [29]

    Medical Imaging with Deep Learning , year=

    MedSegDiff: Medical Image Segmentation with Diffusion Probabilistic Model , author=. Medical Imaging with Deep Learning , year=

  22. [30]

    arXiv preprint arXiv:2102.04306 , year=

    TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation , author=. arXiv preprint arXiv:2102.04306 , year=

    arXiv

  23. [31]

    IEEE Transactions on Medical Imaging , volume=

    UNet++: Redesigning Skip Connections to Exploit Multiscale Features in Image Segmentation , author=. IEEE Transactions on Medical Imaging , volume=. 2020 , doi=

    2020

  24. [32]

    IEEE Transactions on Medical Imaging , volume=

    Reducing the Hausdorff Distance in Medical Image Segmentation with Convolutional Neural Networks , author=. IEEE Transactions on Medical Imaging , volume=. 2020 , doi=

    2020

  25. [33]

    Journal of Medical Internet Research , volume=

    Clinically applicable segmentation of head and neck anatomy for radiotherapy: deep learning algorithm development and validation study , author=. Journal of Medical Internet Research , volume=. 2021 , doi=

    2021

  26. [34]

    International Journal of Radiation Oncology Biology Physics , volume=

    Deep learning algorithm for auto-delineation of high-risk oropharyngeal clinical target volumes with built-in dice similarity coefficient parameter optimization function , author=. International Journal of Radiation Oncology Biology Physics , volume=

  27. [35]

    Frontiers in Oncology , volume=

    Deep Deconvolutional Neural Network for Target Segmentation of Nasopharyngeal Cancer in Planning Computed Tomography Images , author=. Frontiers in Oncology , volume=

  28. [36]

    Medical Physics , volume=

    Automatic segmentation of head and neck organs at risk using deep learning , author=. Medical Physics , volume=. 2019 , doi=

    2019

  29. [37]

    Radiotherapy and Oncology , volume=

    Deep learning-based auto-segmentation of pelvic organs at risk and clinical target volumes for radiotherapy treatment of cervical cancer , author=. Radiotherapy and Oncology , volume=. 2019 , doi=

    2019

  30. [38]

    Radiation Oncology , volume=

    Evaluation of deep learning-based auto-segmentation for radiotherapy treatment planning: a systematic review , author=. Radiation Oncology , volume=

  31. [39]

    Advances in Neural Information Processing Systems , volume=

    Generative Adversarial Nets , author=. Advances in Neural Information Processing Systems , volume=

  32. [40]

    Advances in Neural Information Processing Systems , volume=

    Attention is All You Need , author=. Advances in Neural Information Processing Systems , volume=

  33. [41]

    International Conference on Learning Representations , year=

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale , author=. International Conference on Learning Representations , year=

  34. [42]

    2026 , issn =

    Vision foundation model for 3D magnetic resonance imaging segmentation, classification, and registration , journal =. 2026 , issn =. doi:https://doi.org/10.1016/j.media.2026.103992 , author =

    2026

  35. [43]

    2024 , eprint=

    VISTA3D: A Unified Segmentation Foundation Model For 3D Medical Imaging , author=. 2024 , eprint=

    2024

  36. [44]

    2025 , eprint=

    MedDINOv3: How to Adapt Vision Foundation Models for Medical Image Segmentation? , author=. 2025 , eprint=

    2025

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.