pith. sign in

arxiv: 2606.04705 · v1 · pith:VJVTFHIEnew · submitted 2026-06-03 · 💻 cs.CV · cs.AI

Enhancing MedSAM with a Lightweight Box Predictor for Medical Image Segmentation

Amirhossein Movahedisefat , Amirreza Fateh , Mohammad Reza Mohammadi This is my paper

Pith reviewed 2026-06-28 07:14 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords medical image segmentationMedSAMbounding box predictionsingle-click promptfoundation model adaptationlightweight moduleDice coefficient
0
0 comments X

The pith

A lightweight box predictor integrated into MedSAM estimates bounding boxes from single clicks to improve medical image segmentation accuracy.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper aims to enhance the MedSAM foundation model for medical segmentation by adding a module that converts a single user click into a bounding box prompt. Point prompts alone often lack enough context for irregular or low-contrast structures in medical scans. The approach uses localized embedding features to predict the box, trained separately before integration, adding just 1.6 million parameters. Evaluations on four datasets covering ultrasound, CT, and MRI show higher Dice scores than the base model. This matters because it makes interactive segmentation more reliable with minimal added complexity for clinical use.

Core claim

The Box Predictor module, trained independently on embedding features from single clicks, generates approximate bounding boxes that provide spatial guidance to MedSAM, reducing ambiguity of point prompts and yielding improved segmentation performance across diverse medical imaging modalities and anatomical structures.

What carries the argument

The lightweight Box Predictor that takes localized image embedding features from a single click and outputs an approximate bounding box for use as prompt in MedSAM.

If this is right

  • Segmentation Dice scores reach 0.89 on BUSI, 0.93 on FLARE22, 0.88 on BRISC, and 0.98 on LungSegDB.
  • Only 1.6M additional parameters are introduced with negligible inference overhead.
  • The method generalizes across CT, MRI, and ultrasound modalities on four datasets.
  • Two-stage training allows the Box Predictor to be trained independently before integration with frozen or fine-tuned MedSAM.

Where Pith is reading between the lines

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

  • Similar lightweight predictors could be developed for other prompt types like scribbles in foundation models for segmentation.
  • The approach might reduce the need for multiple clicks in interactive medical annotation workflows.
  • Testing on more varied clinical datasets could reveal if performance holds for rare pathologies.

Load-bearing premise

Training the Box Predictor independently on localized embedding features from single clicks produces bounding boxes that reliably reduce prompt ambiguity in MedSAM without introducing new failure modes on irregular structures.

What would settle it

Observing that on datasets with highly irregular structures the integrated model produces lower Dice scores than MedSAM using only point prompts would falsify the central claim.

Figures

Figures reproduced from arXiv: 2606.04705 by Amirhossein Movahedisefat, Amirreza Fateh, Mohammad Reza Mohammadi.

Figure 1
Figure 1. Figure 1: Schematic illustration of the two-stage training framework. In Stage 1, a 5 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Representative examples from the four medical segmentation datasets. Each row shows four samples from one dataset [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Performance metrics vs. box enlargement percentage across four datasets. The x-axis represents the percentage of [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Effect of input point perturbation on segmentation [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative segmentation results across four datasets. Left image: Ground truth masks (colored overlays) compared [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Representative cases where the BoxPredictor module provides substantial improvements over all baselines across three datasets: BRISC (brain MRI), FLARE22 (abdominal CT), and BUSI (breast ultrasound). Samples 1–2 are from BRISC and FLARE22 respectively, and Sample 3 is from BUSI. All methods receive the same point prompt, shown in the Point Prompt column. The ∆ value indicates the difference between our met… view at source ↗
Figure 7
Figure 7. Figure 7: Representative failure cases where the proposed MedSAM + BoxPredictor underperforms the other compared methods. The layout follows the same convention as [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
read the original abstract

Semantic segmentation in medical imaging is a critical yet challenging task due to data scarcity and high variability across modalities. While foundation models like the Segment Anything Model (SAM) show promise, they often struggle with medical images without specific adaptation. Moreover, point prompts, despite being the most natural form of user interaction, provide insufficient spatial context for reliable segmentation, particularly when target structures are irregular or poorly contrasted. In this paper, we propose an enhanced segmentation framework that integrates a lightweight Box Predictor module into the MedSAM architecture. The Box Predictor estimates an approximate bounding box from a single user click using localized image embedding features, providing spatial guidance that reduces the ambiguity of point prompts, while introducing only 1.6M additional parameters and negligible inference overhead. We introduce a two-stage training pipeline where the Box Predictor is trained independently before being integrated into MedSAM. To validate the generalization capability of our method, we conduct extensive evaluations on four diverse datasets (FLARE22, BRISC, BUSI, LungSegDB) spanning distinct imaging modalities, including CT, MRI, and Ultrasound. Our method improves segmentation accuracy and robustness across varied anatomical structures and imaging domains, achieving Dice scores of 0.89 (BUSI), 0.93 (FLARE22), 0.88 (BRISC), and 0.98 (LungSegDB). Code is available at https://github.com/Amirhosseinmovahedi/MedSAM-BoxPredictor

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The manuscript proposes enhancing MedSAM by integrating a lightweight (1.6M-parameter) Box Predictor module that estimates an approximate bounding box from a single user click on localized image embeddings. This is intended to reduce point-prompt ambiguity for irregular or low-contrast structures. The method uses a two-stage training pipeline (independent Box Predictor training followed by integration into MedSAM) and reports Dice scores of 0.89 (BUSI), 0.93 (FLARE22), 0.88 (BRISC), and 0.98 (LungSegDB) on four datasets spanning CT, MRI, and ultrasound, claiming improved accuracy and robustness with negligible inference overhead.

Significance. If the performance gains hold under proper controls, the work would demonstrate a practical, low-parameter way to convert point prompts into box prompts for medical foundation models, addressing a known limitation of SAM-family models on variable medical data. The two-stage design and public code release would also support reproducibility.

major comments (3)
  1. [Abstract] Abstract: The central performance claim (Dice scores of 0.89–0.98) is presented without any baseline comparisons (e.g., original MedSAM with point prompts only, or SAM with box prompts), error bars, statistical tests, or dataset split details. This prevents assessment of whether the reported gains are attributable to the Box Predictor or to other factors.
  2. [Abstract] Abstract (two-stage pipeline description): The claim that the independently trained Box Predictor reliably reduces prompt ambiguity when integrated into MedSAM lacks supporting evidence such as box-IoU metrics, ablation on joint vs. separate training, or failure-case analysis on irregular structures (e.g., BUSI/BRISC). Without this, the integration step remains an unvalidated assumption.
  3. [Abstract] Abstract (evaluation): No ablation is described that isolates the contribution of the Box Predictor (e.g., MedSAM with oracle boxes vs. predicted boxes), leaving open whether the Dice improvements reflect genuine ambiguity reduction or dataset-specific effects.
minor comments (1)
  1. [Abstract] Abstract: The phrase 'extensive evaluations' is used but the provided results consist only of four aggregate Dice numbers; more granular metrics (e.g., per-structure or per-modality breakdowns) would strengthen the generalization claim.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and will revise the abstract to better contextualize the reported results.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central performance claim (Dice scores of 0.89–0.98) is presented without any baseline comparisons (e.g., original MedSAM with point prompts only, or SAM with box prompts), error bars, statistical tests, or dataset split details. This prevents assessment of whether the reported gains are attributable to the Box Predictor or to other factors.

    Authors: The full manuscript contains baseline comparisons against MedSAM with point prompts and specifies the dataset splits used. We will revise the abstract to include these comparisons and error bars. revision: yes

  2. Referee: [Abstract] Abstract (two-stage pipeline description): The claim that the independently trained Box Predictor reliably reduces prompt ambiguity when integrated into MedSAM lacks supporting evidence such as box-IoU metrics, ablation on joint vs. separate training, or failure-case analysis on irregular structures (e.g., BUSI/BRISC). Without this, the integration step remains an unvalidated assumption.

    Authors: The methods and results sections provide box-IoU metrics for the Box Predictor, ablations on joint versus separate training, and discussion of performance on irregular structures. We will update the abstract to reference these supporting results. revision: yes

  3. Referee: [Abstract] Abstract (evaluation): No ablation is described that isolates the contribution of the Box Predictor (e.g., MedSAM with oracle boxes vs. predicted boxes), leaving open whether the Dice improvements reflect genuine ambiguity reduction or dataset-specific effects.

    Authors: The experimental evaluation in the full manuscript includes ablations that isolate the Box Predictor by comparing against oracle boxes. We will revise the abstract to summarize this ablation. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical results on external datasets with no derivations or self-referential reductions

full rationale

The paper describes an empirical two-stage training pipeline for a lightweight Box Predictor integrated into MedSAM, evaluated via Dice scores on four external datasets (BUSI, FLARE22, BRISC, LungSegDB). No equations, mathematical derivations, fitted parameters renamed as predictions, or load-bearing self-citations appear in the abstract or described content. Performance claims are presented as direct experimental outcomes rather than reductions to inputs by construction. The method is self-contained against external benchmarks with no visible circular steps.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim rests on the effectiveness of the proposed module when trained separately and the assumption that the four listed datasets capture sufficient modality and anatomy variation for generalization claims.

free parameters (1)
  • Box Predictor parameters = 1.6M
    The module adds 1.6M trainable parameters whose values are determined by training on the medical datasets.
axioms (1)
  • domain assumption Point prompts alone provide insufficient spatial context for reliable segmentation of irregular or poorly contrasted medical structures.
    Stated directly in the abstract as the motivation for introducing box prediction.
invented entities (1)
  • Lightweight Box Predictor module no independent evidence
    purpose: Estimates an approximate bounding box from a single user click using localized image embedding features
    New component introduced by the paper; no independent evidence outside the reported experiments is provided.

pith-pipeline@v0.9.1-grok · 5807 in / 1334 out tokens · 32722 ms · 2026-06-28T07:14:18.572836+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 11 canonical work pages · 1 internal anchor

  1. [1]

    Icad- segnet: Efficient and accurate semi-automatic segmentation of intracra- nial atherosclerotic plaques based on low-rank adaptation,

    J. Yuan, L. Mei, Q. Wang, Y . Cao, Z. Li, and B. Liu, “Icad- segnet: Efficient and accurate semi-automatic segmentation of intracra- nial atherosclerotic plaques based on low-rank adaptation,”Biomedical Signal Processing and Control, vol. 112, p. 108936, 2026

    2026

  2. [2]

    Adapting sam with a triple-prompt strategy for one-shot semantic segmentation,

    A. Fateh, M. R. Mohammadi, and M. R. Jahed-Motlagh, “Adapting sam with a triple-prompt strategy for one-shot semantic segmentation,” Neurocomputing, p. 133301, 2026

    2026

  3. [3]

    Msdnet: Multi- scale decoder for few-shot semantic segmentation via transformer-guided prototyping,

    A. Fateh, M. R. Mohammadi, and M. R. J. Motlagh, “Msdnet: Multi- scale decoder for few-shot semantic segmentation via transformer-guided prototyping,”Image and Vision Computing, p. 105672, 2025

    2025

  4. [4]

    Cdsg-sam: A cross-domain self-generating prompt few-shot brain tumor segmentation pipeline based on sam,

    Y . Yang, X. Fang, X. Li, Y . Han, and Z. Yu, “Cdsg-sam: A cross-domain self-generating prompt few-shot brain tumor segmentation pipeline based on sam,”Biomedical Signal Processing and Control, vol. 100, p. 106936, 2025

    2025

  5. [5]

    Sam-myonet: A fine-grained perception myocardial ultrasound segmentation network based on segment anything model with prior knowledge driven,

    Y . Ying, X. Fang, Y . Zhao, X. Zhao, Y . Zhou, G. Du, Y . Zhan, T. Gao, A. Li, D. Sunet al., “Sam-myonet: A fine-grained perception myocardial ultrasound segmentation network based on segment anything model with prior knowledge driven,”Biomedical Signal Processing and Control, vol. 110, p. 108117, 2025

    2025

  6. [6]

    Advances in deep learning for semantic segmentation of low-contrast images: A systematic review of methods, challenges, and future directions,

    C. Urrea and M. V ´elez, “Advances in deep learning for semantic segmentation of low-contrast images: A systematic review of methods, challenges, and future directions,”Sensors, vol. 25, no. 7, p. 2043, 2025

    2043

  7. [7]

    Sit-sam: A semantic-integration transformer that adapts the segment anything model to zero-shot medical image semantic segmentation,

    W. Shi, J. He, and Y . Shen, “Sit-sam: A semantic-integration transformer that adapts the segment anything model to zero-shot medical image semantic segmentation,”Biomedical Signal Processing and Control, vol. 110, p. 108086, 2025

    2025

  8. [8]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inInternational Conference on Medical image computing and computer-assisted intervention. Springer, 2015, pp. 234–241

    2015

  9. [9]

    Encoder- decoder with atrous separable convolution for semantic image segmen- tation,

    L.-C. Chen, Y . Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder- decoder with atrous separable convolution for semantic image segmen- tation,” inProceedings of the European conference on computer vision (ECCV), 2018, pp. 801–818

    2018

  10. [10]

    Labelling instructions matter in biomedical image analysis,

    T. R ¨adsch, A. Reinke, V . Weru, M. D. Tizabi, N. Schreck, A. E. Kavur, B. Pekdemir, T. Roß, A. Kopp-Schneider, and L. Maier-Hein, “Labelling instructions matter in biomedical image analysis,”Nature Machine Intelligence, vol. 5, no. 3, pp. 273–283, 2023

    2023

  11. [11]

    Deep learning with noisy labels: Exploring techniques and remedies in medical image analysis,

    D. Karimi, H. Dou, S. K. Warfield, and A. Gholipour, “Deep learning with noisy labels: Exploring techniques and remedies in medical image analysis,”Medical image analysis, vol. 65, p. 101759, 2020

    2020

  12. [12]

    Comparison of fine-tuning strategies for transfer learning in medical image classification,

    A. Davila, J. Colan, and Y . Hasegawa, “Comparison of fine-tuning strategies for transfer learning in medical image classification,”Image and Vision Computing, vol. 146, p. 105012, 2024

    2024

  13. [13]

    Navigating data scarcity using foundation models: A benchmark of few-shot and zero-shot learning approaches in medical imaging,

    S. Woerner and C. F. Baumgartner, “Navigating data scarcity using foundation models: A benchmark of few-shot and zero-shot learning approaches in medical imaging,” inInternational Workshop on Founda- tion Models for General Medical AI. Springer, 2024, pp. 30–39

    2024

  14. [14]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Loet al., “Segment anything,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 4015–4026

    2023

  15. [15]

    Ma-sam: Modality-agnostic sam adaptation for 3d medical image segmentation,

    C. Chen, J. Miao, D. Wu, A. Zhong, Z. Yan, S. Kim, J. Hu, Z. Liu, L. Sun, X. Liet al., “Ma-sam: Modality-agnostic sam adaptation for 3d medical image segmentation,”Medical Image Analysis, vol. 98, p. 103310, 2024

    2024

  16. [16]

    Segment anything model for medical image segmentation: Current applications and future directions,

    Y . Zhang, Z. Shen, and R. Jiao, “Segment anything model for medical image segmentation: Current applications and future directions,”Com- puters in Biology and Medicine, vol. 171, p. 108238, 2024

    2024

  17. [17]

    Segment anything model for medical images?

    Y . Huang, X. Yang, L. Liu, H. Zhou, A. Chang, X. Zhou, R. Chen, J. Yu, J. Chen, C. Chenet al., “Segment anything model for medical images?”Medical Image Analysis, vol. 92, p. 103061, 2024

    2024

  18. [18]

    Towards segment anything model (sam) for med- ical image segmentation: A survey,

    Y . Zhang and R. Jiao, “Towards segment anything model (sam) for med- ical image segmentation: A survey,”arXiv preprint arXiv:2305.03678, 2023

    work page arXiv 2023

  19. [19]

    Segment anything in medical images,

    J. Ma, Y . He, F. Li, L. Han, C. You, and B. Wang, “Segment anything in medical images,”Nature Communications, vol. 15, no. 1, p. 654, 2024

    2024

  20. [20]

    Fully automated sam for single-source domain generalization in medical image segmentation,

    H. Zhuo, L. Ma, H. Zhao, S. Zhou, D. Sun, and Y . Fu, “Fully automated sam for single-source domain generalization in medical image segmentation,”arXiv preprint arXiv:2507.17281, 2025

    work page arXiv 2025

  21. [21]

    Eviprompt: A training-free evidential prompt generation method for adapting segment anything model in medical images,

    Y . Xu, J. Tang, A. Men, and Q. Chen, “Eviprompt: A training-free evidential prompt generation method for adapting segment anything model in medical images,”IEEE Transactions on Image Processing, 2024

    2024

  22. [22]

    Sam 2: Segment anything in images and videos,

    N. Ravi, V . Gabeur, Y .-T. Hu, R. Hu, C. Ryali, T. Ma, H. Khedr, R. R ¨adle, C. Rolland, L. Gustafson, E. Mintun, J. Pan, K. V . Alwala, N. Carion, C.-Y . Wu, R. Girshick, P. Doll´ar, and C. Feichtenhofer, “Sam 2: Segment anything in images and videos,” inInternational Conference on Learning Representations (ICLR), 2025

    2025

  23. [23]

    Medical sam 2: Segment medical images as video via segment anything model 2,

    J. Zhu, A. Hamdi, Y . Qi, Y . Jin, and J. Wu, “Medical sam 2: Segment medical images as video via segment anything model 2,”arXiv preprint arXiv:2408.00874, 2024

    work page arXiv 2024

  24. [24]

    Sam2-unet: Segment anything 2 makes strong encoder for natural and medical image segmentation,

    X. Xiong, Z. Wu, S. Tan, W. Li, F. Tang, Y . Chen, S. Li, J. Ma, and G. Li, “Sam2-unet: Segment anything 2 makes strong encoder for natural and medical image segmentation,”arXiv preprint arXiv:2408.08870, 2024

    work page arXiv 2024

  25. [25]

    SAM 3: Segment Anything with Concepts

    N. Carion, L. Gustafson, Y .-T. Hu, S. Debnath, R. Hu, D. Suris, C. Ryali, K. V . Alwala, H. Khedr, A. Huang, J. Lei, T. Ma, B. Guo, A. Kalla, M. Marks, J. Greer, M. Wang, P. Sun, R. R¨adle, T. Afouras, E. Mavroudi, K. Xu, T.-H. Wu, Y . Zhou, L. Momeni, R. Hazra, S. Ding, S. Vaze, F. Porcher, F. Li, S. Li, A. Kamath, H. K. Cheng, P. Doll ´ar, N. Ravi, K. ...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  26. [26]

    Medicosam: Towards foun- dation models for medical image segmentation,

    A. Archit, L. Freckmann, and C. Pape, “Medicosam: Towards foun- dation models for medical image segmentation,”arXiv preprint arXiv:2501.11734, 2025

    work page arXiv 2025

  27. [27]

    Pp- sam: Perturbed prompts for robust adaption of segment anything model for polyp segmentation,

    M. M. Rahman, M. Munir, D. Jha, U. Bagci, and R. Marculescu, “Pp- sam: Perturbed prompts for robust adaption of segment anything model for polyp segmentation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 4989–4995

    2024

  28. [28]

    Masksam: Auto-prompt sam with mask classification for volumetric medical image segmentation,

    B. Xie, H. Tang, B. Duan, D. Cai, Y . Yan, and G. Agam, “Masksam: Auto-prompt sam with mask classification for volumetric medical image segmentation,” inProceedings of the IEEE/CVF International Confer- ence on Computer Vision, 2025, pp. 24 423–24 433

    2025

  29. [29]

    Learning to prompt segment anything models,

    J. Huang, K. Jiang, J. Zhang, H. Qiu, L. Lu, S. Lu, and E. Xing, “Learning to prompt segment anything models,”arXiv preprint arXiv:2401.04651, 2024

    work page arXiv 2024

  30. [30]

    Robust box prompt based sam for medical image segmentation,

    Y . Huang, X. Yang, H. Zhou, Y . Cao, H. Dou, F. Dong, and D. Ni, “Robust box prompt based sam for medical image segmentation,” in International Workshop on Machine Learning in Medical Imaging. Springer, 2024, pp. 1–11. 16

    2024

  31. [31]

    Pa-sam: Prompt adapter sam for high-quality image segmentation,

    Z. Xie, B. Guan, W. Jiang, M. Yi, Y . Ding, H. Lu, and L. Zhang, “Pa-sam: Prompt adapter sam for high-quality image segmentation,” in 2024 IEEE International Conference on Multimedia and Expo (ICME). IEEE, 2024, pp. 1–6

    2024

  32. [32]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clarket al., “Learning transferable visual models from natural language supervision,” inInternational conference on machine learning. PmLR, 2021, pp. 8748–8763

    2021

  33. [33]

    Medclip: Contrastive learning from unpaired medical images and text,

    Z. Wang, Z. Wu, D. Agarwal, and J. Sun, “Medclip: Contrastive learning from unpaired medical images and text,” inProceedings of the Conference on Empirical Methods in Natural Language Processing. Conference on Empirical Methods in Natural Language Processing, vol. 2022, 2022, p. 3876

    2022

  34. [34]

    Enhancing sam with efficient prompting and preference optimiza- tion for semi-supervised medical image segmentation,

    A. Konwer, Z. Yang, E. Bas, C. Xiao, P. Prasanna, P. Bhatia, and T. Kass- Hout, “Enhancing sam with efficient prompting and preference optimiza- tion for semi-supervised medical image segmentation,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 20 990–21 000

    2025

  35. [35]

    Towards universal text-driven ct image segmentation,

    Y . Li, Y . Lai, M. Thor, D. Marshall, Z. Buchwald, D. S. Yu, and X. Yang, “Towards universal text-driven ct image segmentation,”arXiv preprint arXiv:2503.06030, 2025

    work page arXiv 2025

  36. [36]

    Generalized intersection over union: A metric and a loss for bounding box regression,

    H. Rezatofighi, N. Tsoi, J. Gwak, A. Sadeghian, I. Reid, and S. Savarese, “Generalized intersection over union: A metric and a loss for bounding box regression,” inProceedings of the IEEE/CVF conference on com- puter vision and pattern recognition, 2019, pp. 658–666

    2019

  37. [37]

    Unleashing the strengths of unlabelled data in deep learning-assisted pan-cancer abdominal organ quantification: the flare22 challenge,

    J. Ma, Y . Zhang, S. Gu, C. Ge, S. Mae, A. Young, C. Zhu, X. Yang, K. Meng, Z. Huanget al., “Unleashing the strengths of unlabelled data in deep learning-assisted pan-cancer abdominal organ quantification: the flare22 challenge,”The Lancet Digital Health, vol. 6, no. 11, pp. e815– e826, 2024

    2024

  38. [38]

    Dataset of breast ultrasound images,

    W. Al-Dhabyani, M. Gomaa, H. Khaled, and A. Fahmy, “Dataset of breast ultrasound images,”Data in brief, vol. 28, p. 104863, 2020

    2020

  39. [39]

    Brisc: Annotated dataset for brain tumor segmentation and classification,

    A. Fateh, Y . Rezvani, S. Moayedi, S. Rezvani, F. Fateh, M. Fateh, and V . Abolghasemi, “Brisc: Annotated dataset for brain tumor segmentation and classification,”Scientific Data, 2026

    2026

  40. [40]

    Fusionlungnet: Multi- scale fusion convolution with refinement network for lung ct image segmentation,

    S. Rezvani, M. Fateh, Y . Jalali, and A. Fateh, “Fusionlungnet: Multi- scale fusion convolution with refinement network for lung ct image segmentation,”Biomedical Signal Processing and Control, vol. 107, p. 107858, 2025

    2025

  41. [41]

    Is sam 2 better than sam in medical image segmentation?

    S. Sengupta, S. Chakrabarty, and R. Soni, “Is sam 2 better than sam in medical image segmentation?”arXiv preprint arXiv:2408.04212, 2024

    work page arXiv 2024

  42. [42]

    Segment anything in medical images and videos: Benchmark and deployment,

    J. Ma, S. Kim, F. Li, M. Baharoon, R. Asakereh, H. Lyu, and B. Wang, “Segment anything in medical images and videos: Benchmark and deployment,”arXiv preprint arXiv:2408.03322, 2024

    work page arXiv 2024

  43. [43]

    Segment anything model 2: An application to 2d and 3d medical images,

    H. Dong, H. Gu, Y . Chen, J. Yang, Y . Chen, and M. A. Mazurowski, “Segment anything model 2: An application to 2d and 3d medical images,”arXiv preprint arXiv:2408.00756, 2024

    work page arXiv 2024