CLISC: Bridging clip and sam by enhanced cam for unsupervised brain tumor segmentation

Guotai Wang; Jia Fu; Shichuan Zhang; Wenjun Liao; Xiaochuan Ma

REVIEW 4 major objections 6 minor 1 cited by

CLISC: Bridging clip and sam by enhanced cam for unsupervised brain tumor segmentation

T0 review · 4 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read An unsupervised cascade that chains CLIP, CAM, and SAM segments brain tumors at 85.60% DSC on BraTS2020 without any manual annotations.

desk verdict CLISC reports a strikingly strong unsupervised whole-tumor segmentation result, but the unvalidated CLIP labeling stage and missing code are real caveats. read the letter →

arxiv 2501.16246 v1 pith:UBTOZN52 submitted 2025-01-27 cs.CV

Xiaochuan Ma , Jia Fu , Wenjun Liao , Shichuan Zhang , Guotai Wang This is my paper · ORCID

classification cs.CV

keywords unsupervisedsegmentationbraintumorfoundationmodelsCLIPSAMclassactivationmappingself-trainingBraTS2020

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

This paper tries to establish that an unsupervised pipeline can get close to supervised performance on brain tumor segmentation by chaining two foundation models. CLIP supplies image-level tumor/non-tumor labels for every FLAIR slice, a classifier trained on those labels produces enhanced class activation maps, and those maps generate box and point prompts for SAM to produce segmentation pseudo-labels. A 3D U-Net is then trained on the pseudo-labels, with a self-training step that keeps only labels on which SAM and the network agree. On BraTS2020, the pipeline reports 85.60% average whole-tumor DSC and 6.72 mm HD95, about 10 points above five unsupervised baselines and within 2.4 points of fully supervised training. The evaluation is limited to whole-tumor FLAIR segmentation; tumor substructures and other organs are left to future work.

What carries the argument

The load-bearing mechanism is a four-stage pseudo-label cascade. First, CLIP's cosine-similarity scores classify each FLAIR slice as tumor or non-tumor, and those labels train a ResNet classifier. Second, Layer-CAM produces a raw activation map, and an adaptive masking step hides the most discriminative region and retrains the classifier so the next CAM covers more of the tumor. Third, the enhanced CAM defines a bounding box and a center point plus four corner points that prompt SAM, yielding per-slice masks that are stacked into 3D pseudo-labels. Fourth, a 3D U-Net trains on those pseudo-labels, then a self-training round re-prompts SAM with the network's own prediction and keeps only volumes whose pseudo-label agrees with SAM above a percentile threshold.

What would settle it

Run the CLIP label step on the BraTS2020 training slices and compare the argmax labels against ground-truth tumor presence. If the slice-level accuracy is close to chance, or if replacing CLIP labels with ground-truth labels does not materially change the final segmentation DSC, the central claim would be weakened; the paper's own ablations do not include this measurement.

Watch

Extended reading notes

Core claim

CLISC claims that slice-level labels extracted from CLIP with two English text prompts are sufficient to bootstrap a cascade that produces high-quality whole-tumor segmentations. After filtering CLIP labels through a ResNet classifier, Layer-CAM localization with adaptive masking, SAM prompting with a bounding box plus five points, and a similarity-based self-training round, the 3D segmentation network reaches 85.60% DSC on BraTS2020, surpassing zero-shot SAM prompted by ground-truth boxes (80.59% DSC) and approaching fully supervised training (87.97% DSC). The paper treats this as evidence that adapting foundation models through an intermediate CAM stage, rather than using them directly for inference, closes much of the gap between unsupervised and supervised medical image segmentation.

Load-bearing premise

The pipeline assumes CLIP's zero-shot slice-level tumor/non-tumor labels on FLAIR MRI are accurate enough to train a reliable classifier; the paper never reports this label accuracy against ground truth, and every later stage inherits its errors.

Editorial extensions

If this is right

On BraTS2020 FLAIR whole-tumor segmentation, the unsupervised CLISC pipeline reaches 85.60% average DSC and 6.72 mm HD95, more than 10 DSC points above five state-of-the-art unsupervised baselines.
CLISC beats SAM in zero-shot inference even when SAM is given prompts derived from ground truth: 85.60% versus 80.59% DSC.
The gap to fully supervised training is small: 85.60% versus 87.97% DSC, and CLISC's HD95 is better (6.72 mm versus 7.49 mm).
Ablations show each stage contributes: adaptive masking raises CAM DSC from 51.10% to 60.22%, combined box-and-point SAM prompts reach 74.21% DSC, and self-training with similarity filtering raises the final DSC from 82.29% to 85.60%.
The method also handles small and tiny tumors better than the unsupervised baselines, and for tiny tumors it exceeds fully supervised training.

Reading between the lines

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

The same cascade could be transferred to other organs or modalities where CLIP gives weak slice-level signal, provided the CAM stage can localize the salient region; the paper's future-work sentence points this way.
A testable extension is to replace CLIP with a medical-domain vision-language model and measure how much of the gain comes from better slice labels rather than from CAM or SAM.
The similarity-based filtering between SAM output and student prediction is a generic denoising rule that could be applied to any pseudo-label pipeline, independent of CLIP.
If CLIP slice-label accuracy is low, the entire chain inherits the error; reporting that accuracy against ground-truth labels would make the method's reliability transparent.

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.

Desk Editor's Note

read the letter

Here's my read on CLISC (arXiv:2501.16246). If the numbers are reproducible, this is an impressive result: 85.6 Dice for whole-tumor FLAIR segmentation on BraTS2020 with zero manual labels, within 2.4 points of full supervision and above SAM prompted with ground-truth boxes. The pipeline is a coherent three-stage cascade — CLIP slice-level labels to train a classifier, LayerCAM with adaptive masking (AMDA) to refine ROIs, SAM prompts from those ROIs to get pseudo-masks, then a 3D U-Net self-trained with a similarity filter (S3F) against SAM. What's genuinely new is the combination plus the two refinements; the individual pieces are known but the integration is sensible and the ablations are clean. Each stage adds a clear, monotonic improvement: AMDA lifts CAM DSC from 51.1 to 60.2, adding box+point prompts gets 74.2, training the U-Net on SAM pseudo-labels jumps to 82.3, and S3F filtering brings it to 85.6. That's good evidence the gains come from the whole system, not one lucky component.

The main soft spot is the first stage: CLIP's slice-level tumor/non-tumor labels are never validated against ground truth. The paper uses two English text prompts applied to FLAIR MRI — a modality far from CLIP's training data — and the entire downstream chain inherits any error there. No agreement statistics are reported, so a reader can't tell whether the method is robust to label noise or just happens to work on this dataset. That's a real gap, though the final held-out score is an indirect sanity check; if CLIP labels were badly wrong, you'd expect the pipeline to collapse. Still, for transferability, this needs reporting.

Second, no code is released, and the hyperparameters (alpha=20, beta=20) aren't explicitly tied to the validation split. The ablation plots in Fig. 3 suggest they did tune on validation, but it should be stated. One note: the stress-test raised that alpha=20 masks 80% of pixels, which seems inconsistent with "larger alpha masks all tumor areas." That concern doesn't hold up on reading — larger alpha gives a higher threshold and thus a smaller, more concentrated mask, which can indeed erase the whole tumor. So the paper's explanation is fine.

Bottom line: this deserves a serious referee. It's a strong empirical claim with well-designed experiments and a sensible method. I'd recommend sending it to peer review, with the expectation that the authors release code and report CLIP label accuracy on a sample. If those hold, it's a solid contribution for anyone working on annotation-efficient medical segmentation.

Referee Report

4 major / 6 minor

Summary. The paper proposes CLISC, a three-stage unsupervised pipeline for whole-tumor FLAIR segmentation on BraTS2020. CLIP provides slice-level pseudo-labels that train a ResNet-50 classifier; Layer-CAM with adaptive masking (AMDA) produces ROIs; ROIs generate box/point prompts for SAM; SAM masks are stacked into 3D pseudo-labels; a 3D U-Net is trained with these pseudo-labels and refined with a similarity-based filter (S3F). On a 20% held-out test set, the method reports average DSC 85.60% and HD95 6.72 mm, exceeding five unsupervised baselines by more than 10 DSC points and approaching fully supervised training (87.97% DSC, 7.49 mm HD95).

Significance. If valid, the result is significant: it demonstrates that a cascade of frozen foundation models can close much of the gap between unsupervised and fully supervised whole-tumor segmentation, and it outperforms SAM prompted by ground truth. The evaluation is well structured: a held-out test set, five strong baselines, paired t-tests, a size-stratified breakdown, and a three-part ablation. The method is also simple in that CLIP and SAM are used off-the-shelf. However, the load-bearing first stage—CLIP slice-level pseudo-labels—is never validated against any ground truth, and the hyperparameter selection protocol is not explicit; these issues must be addressed before the central claim is fully supported.

major comments (4)

[Sec. 2.1.1, Eq. (1)] The paper never reports the accuracy of the CLIP-derived slice-level pseudo-labels yi = argmax(p_i) against ground-truth tumor presence, even though D1 is the foundation of every subsequent stage (classifier, CAM, AMDA, SAM prompts, pseudo-labels D3/D4). If CLIP's false-negative rate on tumor slices is substantial, the entire pipeline inherits missing or spurious labels, and the final 85.60% DSC would not be evidence that the pipeline is robust. Please report slice-level classification accuracy, a confusion matrix, and a sensitivity analysis (e.g., replacing CLIP labels with oracle labels or perturbing a fraction of labels) on a held-out split.
[Sec. 2.1.2] The definition of Mi = Q0_i > Tα, where Tα is 'the α-th percentile,' is ambiguous and appears inconsistent with the stated goal of masking the most discriminative region. For α=20, Mi selects the top 80% of pixels (if Q0 values are sorted ascending), so masking Mi removes most of the image, not just the discriminative region. Please clarify whether Tα should be the (100−α)-th percentile or whether Mi is defined differently, and report the mean fraction of pixels masked for each α in Fig. 3(a).
[Sec. 3.1 and Fig. 3] The choice α=β=20 is not tied to a stated validation procedure. If the test set was used to select these hyperparameters, the reported DSC is optimistically biased. Please state explicitly how Fig. 3 was generated (validation vs test), and report performance on the 10% validation split for all configurations, including the final model.
[Sec. 2.3] The S3F filter uses the similarity between the network's own prediction Si and SAM's output S_hat_i, where the SAM prompts are generated from Si. This internal-consistency criterion can be high even when both are confidently wrong, so F(Si,S_hat_i) does not directly measure pseudo-label quality. The ablation shows a gain, but the mechanism remains unclear; please analyze how filtering changes the agreement between retained pseudo-labels and ground truth (if measurable) and discuss the risk of self-confirmation.

minor comments (6)

[Sec. 2.2] The phrase 'four corner points (background) of the bounding box' is ambiguous; do these points lie inside or outside the box? Please specify their coordinates.
[Table 2] Define exactly what '5 points' and 'box' mean for SAM prompts in the ablation; a reader cannot reproduce the prompt settings from the current text.
[Fig. 3] Label the vertical axes; currently the reader cannot tell whether DSC or HD95 is plotted.
[Sec. 3.1] The manuscript does not mention whether the validation split is used for early stopping or model selection for the 3D U-Net; please clarify.
[General] No code or pre-trained models are released; given the many non-standard steps (AMDA, S3F), releasing code would substantially aid reproducibility.
[Fig. 1] There are a few typographical issues in Figure 1 (e.g., duplicated '𝓓𝓓' symbols), which should be cleaned up.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: CLISC's final score is measured against held-out ground truth and pseudo-labels come from external CLIP and SAM; the missing CLIP-label validation is a correctness gap, not a circular step.

full rationale

The derivation chain is not circular. CLIP pseudo-labels in Sec. 2.1.1 are obtained by zero-shot text-image similarity (Eq. 1) using two fixed prompts, so D1 is produced by an external frozen model and not derived from the segmentation target. The classifier, Layer-CAM, AMDA, and SAM prompts are all downstream of this external signal, and the final 3D U-Net is evaluated against held-out BraTS2020 ground truth after a 7:1:2 split, so no reported DSC or HD95 is fitted to the test labels. The self-training stage (Sec. 2.3) does reuse the network's own first-round predictions Si as candidate labels, but only after agreement filtering F(Si, S_hat_i) > Tbeta with frozen SAM, and the final metric remains external; this is a mild confirmation effect rather than an identity by construction. The one minor self-citation ([1], corresponding author) appears only as a general performance citation in the introduction and is not load-bearing. The absence of a measured CLIP pseudo-label accuracy is a genuine validation gap and the paper's weakest assumption, but a missing validation is not circularity under the definitions used here. Overall, the central claim is self-contained against external benchmarks, so the circularity score is low.

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

The central claim rests on four domain assumptions about the transfer of CLIP and SAM to medical images and about the validity of the similarity filter. There are two fitted hyperparameters (alpha and beta). No new physical or model entities are introduced; AMDA and S3F are procedures, not entities.

free parameters (2)

AMDA threshold percentile alpha = 20
In Sec 2.1.2, M_i = Q0_i > T_alpha with T_alpha the alpha-th percentile of the raw CAM. Alpha=20 was selected by scanning 0, 10, 20, and 30 and reporting the best DSC in Fig 3(a).
S3F filtering percentile beta = 20
In Sec 2.3, D4 keeps volumes with F(S_i, S_hat_i) above the beta-th percentile. Beta=20 was selected by scanning 0, 10, 20, 30, and 40 and reporting the best DSC in Fig 3(b).

assumptions (4)

domain assumption CLIP's zero-shot text-image comparison with two hand-written English prompts yields reliable slice-level tumor and non-tumor labels on FLAIR brain MRI.
Sec 2.1.1 builds D1 = {X_i, y_i} from y_i = argmax(p_i) with prompts T0 and T1; the paper does not evaluate the accuracy of these labels against ground truth.
domain assumption LayerCAM on the ResNet50 classifier trained on CLIP pseudo-labels highlights the full tumor region, especially after AMDA retraining.
Sec 2.1.2 and Sec 2.2 use Q1 to define ROIs and prompts; if CAM misses parts of the tumor, SAM prompts will miss those parts.
domain assumption SAM, prompted with the bounding box plus center and four corner points, produces high-quality tumor masks on FLAIR images.
Sec 2.2 uses Eq (3) to generate all segmentation pseudo-labels from SAM; no per-slice quality check is applied before stacking.
domain assumption DSC agreement between the first-round 3D U-Net prediction and SAM's re-prediction is a valid indicator of pseudo-label quality.
Sec 2.3 defines F(S_i, S_hat_i) and filters by T_beta; the paper does not verify this proxy against ground-truth quality.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CLISC: Bridging clip and sam by enhanced cam for unsupervised brain tumor segmentation." pith.science (2026). https://pith.science/paper/UBTOZN52

@misc{pith2026250116246,
  author       = {Pith},
  title        = {Pith review of: CLISC: Bridging clip and sam by enhanced cam for unsupervised brain tumor segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UBTOZN52}},
  note         = {Machine review of arXiv:2501.16246}
}

read the original abstract

Brain tumor segmentation is important for diagnosis of the tumor, and current deep-learning methods rely on a large set of annotated images for training, with high annotation costs. Unsupervised segmentation is promising to avoid human annotations while the performance is often limited. In this study, we present a novel unsupervised segmentation approach that leverages the capabilities of foundation models, and it consists of three main steps: (1) A vision-language model (i.e., CLIP) is employed to obtain image-level pseudo-labels for training a classification network. Class Activation Mapping (CAM) is then employed to extract Regions of Interest (ROIs), where an adaptive masking-based data augmentation is used to enhance ROI identification.(2) The ROIs are used to generate bounding box and point prompts for the Segment Anything Model (SAM) to obtain segmentation pseudo-labels. (3) A 3D segmentation network is trained with the SAM-derived pseudo-labels, where low-quality pseudo-labels are filtered out in a self-learning process based on the similarity between the SAM's output and the network's prediction. Evaluation on the BraTS2020 dataset demonstrates that our approach obtained an average Dice Similarity Score (DSC) of 85.60%, outperforming five state-of-the-art unsupervised segmentation methods by more than 10 percentage points. Besides, our approach outperforms directly using SAM for zero-shot inference, and its performance is close to fully supervised learning.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

Prompt Engineering in Segment Anything Model: Methodologies, Applications, and Emerging Challenges
cs.CV 2025-07 conditional novelty 2.0 of 10

A structured survey of prompt engineering methods for the Segment Anything Model, covering geometric, textual, and multimodal prompts and their applications.

Reference graph

Works this paper leans on

24 extracted references · 22 canonical work pages · cited by 1 Pith paper

[1]

and structure

INTRODUCTION Magnetic Resonance Imaging (MRI) plays an important role in brain tumor diagnosis and treatment planning, due to its good soft-tissue contrast for visualizing tumor morphology Corresponding author: Guotai Wang (guotai.wang@uestc.edu.cn). and structure. Deep learning has achieved remarkable perfor- mance on brain tumor segmentation when traine...

arXiv 2025
[2]

an image of brain tissue showing typical signal intensity without any re- gions of abnormal intensity or suspicious mass

METHODOLOGY The overall structure of our framework is illustrated in Fig. 1. First, we use CLIP-derived image-level labels to supervise a classification network and generate enhanced CAM [14] with Adaptive Masking-based Data Augmentation (AMDA). Next, segmentation pseudo-labels are obtained by leveraging SAM [10] with CAM-derived prompts. Finally, a 3D se...

work page
[3]

Dataset and Implementation We evaluated our method on the BraTS2020 that comprises 369 cases each containing 3D volumes in four modalities

EXPERIMENTS AND RESULTS 3.1. Dataset and Implementation We evaluated our method on the BraTS2020 that comprises 369 cases each containing 3D volumes in four modalities. In this work, we focus on the segmentation of the whole tumor from FLAIR images and ignore the other modalities. The dataset was split into 7:1:2 for training, validation and test- ing, re...

work page
[4]

It uses CLIP-derived image labels to supervise a classification network with Adaptive Masking- based Data Augmentation for enhanced CAM

CONCLUSION In conclusion, we proposed a novel unsupervised brain tu- mor segmentation framework CLISC that bridges CLIP and SAM by enhanced CAM. It uses CLIP-derived image labels to supervise a classification network with Adaptive Masking- based Data Augmentation for enhanced CAM. The high- quality CAMs are then used to generate prompts for SAM to obtain ...

work page
[5]

Ethical ap- proval was not required as confirmed by the license attached with the open-access data

COMPLIANCE WITH ETHICAL STANDARDS This research study was conducted retrospectively using hu- man subject data made available in open access. Ethical ap- proval was not required as confirmed by the license attached with the open-access data

work page
[6]

ACKNOWLEDGMENT This work was supported by the National Natural Science Foundation of China under Grant 62271115, and by the Fun- damental Research Funds for the Central Universities under Grant ZYGX2022YGRH019

work page
[7]

Automatic brain tumor segmentation based on cascaded convolutional neural networks with uncertainty estimation,

Guotai Wang, Wenqi Li, S ´ebastien Ourselin, and Tom Vercauteren, “Automatic brain tumor segmentation based on cascaded convolutional neural networks with uncertainty estimation,”Frontiers in computational neu- roscience, vol. 13, pp. 56, 2019

work page 2019
[8]

Deep learning models and traditional automated techniques for brain tumor segmentation in MRI: a review,

Parvathy Jyothi and A Robert Singh, “Deep learning models and traditional automated techniques for brain tumor segmentation in MRI: a review,”Artificial intelli- gence review, vol. 56, no. 4, pp. 2923–2969, 2023

work page 2023

Show all 24 references

[9]

Self-semantic contour adaptation for cross modality brain tumor segmentation,

Xiaofeng Liu, Fangxu Xing, Georges El Fakhri, and Jonghye Woo, “Self-semantic contour adaptation for cross modality brain tumor segmentation,” in ISBI, 2022, pp. 1–5

2022
[10]

Brain tumor segmentation based on a hybrid clustering technique,

Eman Abdel-Maksoud, Mohammed Elmogy, and Rashid Al-Awadi, “Brain tumor segmentation based on a hybrid clustering technique,” Egyptian Informat- ics Journal, vol. 16, no. 1, pp. 71–81, 2015

2015
[11]

Level set method with automatic selective local statistics for brain tumor segmentation in MR images,

Kiran Thapaliya, Jae-Young Pyun, Chun-Su Park, and Goo-Rak Kwon, “Level set method with automatic selective local statistics for brain tumor segmentation in MR images,” Computerized Medical Imaging and Graphics, vol. 37, no. 7-8, pp. 522–537, 2013

2013
[12]

Bayesian skip-autoencoders for un- supervised hyperintense anomaly detection in high res- olution brain MRI,

Christoph Baur, Benedikt Wiestler, Shadi Albarqouni, and Nassir Navab, “Bayesian skip-autoencoders for un- supervised hyperintense anomaly detection in high res- olution brain MRI,” in ISBI, 2020, pp. 1905–1909

2020
[13]

Unsupervised region-based anomaly detection in brain MRI with ad- versarial image inpainting,

Bao Nguyen, Adam Feldman, Sarath Bethapudi, An- drew Jennings, and Chris G Willcocks, “Unsupervised region-based anomaly detection in brain MRI with ad- versarial image inpainting,” in ISBI, 2021, pp. 1127– 1131

2021
[14]

Constrained unsupervised anomaly segmentation,

Julio Silva-Rodr ´ıguez, Valery Naranjo, and Jose Dolz, “Constrained unsupervised anomaly segmentation,” Medical Image Analysis, vol. 80, pp. 102526, 2022

2022
[15]

Self- supervised tumor segmentation with sim2real adapta- tion,

Xiaoman Zhang, Weidi Xie, Chaoqin Huang, Ya Zhang, Xin Chen, Qi Tian, and Yanfeng Wang, “Self- supervised tumor segmentation with sim2real adapta- tion,” IEEE Journal of Biomedical and Health Infor- matics, vol. 27, no. 9, pp. 4373–4384, 2023

2023
[16]

Segment anything,

Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C Berg, Wan-Yen Lo, et al., “Segment anything,” in ICCV, 2023, pp. 4015– 4026

2023
[17]

Test-time adap- tation with SaLIP: a cascade of SAM and CLIP for zero- shot medical image segmentation,

Sidra Aleem, Fangyijie Wang, Mayug Maniparam- bil, Eric Arazo, Julia Dietlmeier, Kathleen Curran, Noel EO’ Connor, and Suzanne Little, “Test-time adap- tation with SaLIP: a cascade of SAM and CLIP for zero- shot medical image segmentation,” in CVPR, 2024, pp. 5184–5193

2024
[18]

Learning transferable visual models from natural lan- guage supervision,

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sas- try, Amanda Askell, Pamela Mishkin, Jack Clark, et al., “Learning transferable visual models from natural lan- guage supervision,” in ICML, 2021, pp. 8748–8763

2021
[19]

Im- plicit field learning for unsupervised anomaly detection in medical images,

Sergio Naval Marimont and Giacomo Tarroni, “Im- plicit field learning for unsupervised anomaly detection in medical images,” in MICCAI, 2021, pp. 189–198

2021
[20]

Layercam: Explor- ing hierarchical class activation maps for localization,

Peng-Tao Jiang, Chang-Bin Zhang, Qibin Hou, Ming- Ming Cheng, and Yunchao Wei, “Layercam: Explor- ing hierarchical class activation maps for localization,” IEEE Transactions on Image Processing , vol. 30, pp. 5875–5888, 2021

2021
[21]

Learning deep features for discriminative localization,

Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba, “Learning deep features for discriminative localization,” in CVPR, 2016, pp. 2921–2929

2016
[22]

Grad-cam: Visual explanations from deep net- works via gradient-based localization,

Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra, “Grad-cam: Visual explanations from deep net- works via gradient-based localization,” in ICCV, 2017, pp. 618–626

2017
[23]

Deep residual learning for image recognition,

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun, “Deep residual learning for image recognition,” in CVPR, 2016, pp. 770–778

2016
[24]

3D U-Net: learning dense volumetric segmentation from sparse an- notation,

¨Ozg¨un C ¸ ic ¸ek, Ahmed Abdulkadir, Soeren S Lienkamp, Thomas Brox, and Olaf Ronneberger, “3D U-Net: learning dense volumetric segmentation from sparse an- notation,” in MICCAI, 2016, pp. 424–432

2016

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

and structure

INTRODUCTION Magnetic Resonance Imaging (MRI) plays an important role in brain tumor diagnosis and treatment planning, due to its good soft-tissue contrast for visualizing tumor morphology Corresponding author: Guotai Wang (guotai.wang@uestc.edu.cn). and structure. Deep learning has achieved remarkable perfor- mance on brain tumor segmentation when traine...

arXiv 2025

[2] [2]

an image of brain tissue showing typical signal intensity without any re- gions of abnormal intensity or suspicious mass

METHODOLOGY The overall structure of our framework is illustrated in Fig. 1. First, we use CLIP-derived image-level labels to supervise a classification network and generate enhanced CAM [14] with Adaptive Masking-based Data Augmentation (AMDA). Next, segmentation pseudo-labels are obtained by leveraging SAM [10] with CAM-derived prompts. Finally, a 3D se...

work page

[3] [3]

Dataset and Implementation We evaluated our method on the BraTS2020 that comprises 369 cases each containing 3D volumes in four modalities

EXPERIMENTS AND RESULTS 3.1. Dataset and Implementation We evaluated our method on the BraTS2020 that comprises 369 cases each containing 3D volumes in four modalities. In this work, we focus on the segmentation of the whole tumor from FLAIR images and ignore the other modalities. The dataset was split into 7:1:2 for training, validation and test- ing, re...

work page

[4] [4]

It uses CLIP-derived image labels to supervise a classification network with Adaptive Masking- based Data Augmentation for enhanced CAM

CONCLUSION In conclusion, we proposed a novel unsupervised brain tu- mor segmentation framework CLISC that bridges CLIP and SAM by enhanced CAM. It uses CLIP-derived image labels to supervise a classification network with Adaptive Masking- based Data Augmentation for enhanced CAM. The high- quality CAMs are then used to generate prompts for SAM to obtain ...

work page

[5] [5]

Ethical ap- proval was not required as confirmed by the license attached with the open-access data

COMPLIANCE WITH ETHICAL STANDARDS This research study was conducted retrospectively using hu- man subject data made available in open access. Ethical ap- proval was not required as confirmed by the license attached with the open-access data

work page

[6] [6]

ACKNOWLEDGMENT This work was supported by the National Natural Science Foundation of China under Grant 62271115, and by the Fun- damental Research Funds for the Central Universities under Grant ZYGX2022YGRH019

work page

[7] [7]

Automatic brain tumor segmentation based on cascaded convolutional neural networks with uncertainty estimation,

Guotai Wang, Wenqi Li, S ´ebastien Ourselin, and Tom Vercauteren, “Automatic brain tumor segmentation based on cascaded convolutional neural networks with uncertainty estimation,”Frontiers in computational neu- roscience, vol. 13, pp. 56, 2019

work page 2019

[8] [8]

Deep learning models and traditional automated techniques for brain tumor segmentation in MRI: a review,

Parvathy Jyothi and A Robert Singh, “Deep learning models and traditional automated techniques for brain tumor segmentation in MRI: a review,”Artificial intelli- gence review, vol. 56, no. 4, pp. 2923–2969, 2023

work page 2023