Pith. sign in

REVIEW 4 major objections 5 minor 48 references

S3TU-Net: Structured Convolution and Superpixel Transformer for Lung Nodule Segmentation

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that S3TU-Net, a U-Net hybrid with structured convolutions, a superpixel-based transformer, and spatial-shift skip links, achieves 89.04% Dice on LIDC-IDRI and 86.40% on a private adenocarcinoma dataset, outperforming…

desk verdict Plausible incremental architecture; reported 4.5-point Dice gain not yet trustworthy because of split leakage risk and metric inconsistencies. read the letter →

arxiv 2411.12547 v1 pith:Y5FXCUOD submitted 2024-11-19 eess.IV cs.CVcs.LG

Yuke Wu , Xiang Liu , Yunyu Shi , Xinyi Chen , Zhenglei Wang , YuQing Xu , Shuo Hong Wang This is my paper · ORCID
classification eess.IVcs.CVcs.LG
keywords lungnodulesegmentationCNN-TransformerhybridsuperpixeltransformerspatialshiftMLPstructuredconvolutionDropBlockLIDC-IDRICTimage
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 argues that combining structured convolutional blocks, a superpixel-based transformer, and spatial-shift skip links in a single U-shaped network yields state-of-the-art lung nodule segmentation. On the LIDC-IDRI dataset, the proposed S3TU-Net reports a DSC of 89.04%, precision of 90.73%, IoU of 90.70%, and sensitivity of 93.70%, outperforming eight recent baselines by margins of roughly 4.5 points in DSC. On an independent private dataset of lung adenocarcinoma CT slices, it retains an 86.40% DSC, which the authors take as evidence of generalization. The paper also shows through ablation that each of the three architectural additions contributes positively, with the structured convolutions alone raising DSC by 9.41 points over a plain U-Net baseline.

What carries the argument

The load-bearing machinery is a stack of three modular innovations inside a U-Net skeleton. (1) DWF-Conv and D2BR-Conv replace standard convolution blocks: DWF-Conv inserts a large-kernel attention (LKA) module with depthwise and dilated convolutions plus scaled ReLU activations, and D2BR-Conv inserts DropBlock and batch normalization after each 3x3 convolution. (2) RM-SViT divides the encoded feature tensor into non-overlapping patches, initializes 'super tokens' by averaging features in an 8x8 grid, iteratively refines the token-superpixel association via a softmax similarity (Eq. 2), then applies multi-branch self-attention with a residual connection. (3) The S2-MLP Link expands channels threefold, spatially shifts two of the three branches in orthogonal directions, and recombines them with split attention. The paper's argument is that these modules respectively widen local receptive fields, model global context cheaply, and improve multi-scale fusion, and that the ensemble is what lifts segmentation metrics.

What would settle it

A controlled re-evaluation would split LIDC-IDRI by patient (or by nodule) rather than by slice, train S3TU-Net and at least the two strongest baselines (USG-Net and SA-UNet) under identical preprocessing, loss, and training budget, and report mean Dice over multiple seeds. If the per-nodule split shrinks S3TU-Net's DSC margin to near the reported 4.52% gap or turns it negative, the central claim as stated would be refuted. Additionally, checking whether the reported mIoU exceeds the reported DSC (90.70% vs 89.04%) is consistent with the stated IoU definition would settle concerns about metric integrity.

Watch

Extended reading notes

Core claim

The central discovery is that a U-shaped encoder-decoder can be made substantially more accurate for lung nodule segmentation by replacing plain convolutions with two specialized blocks, inserting a superpixel-based vision transformer between encoder and decoder, and adding spatial-shift MLP links at the skip connections. The DWF-Conv block uses large-kernel attention and scaled activations to broaden receptive fields, while the D2BR-Conv block applies DropBlock to regularize training. The RM-SViT module reduces the token sequence into learnable 'super tokens' via iterative sparse association, applies multi-branch self-attention, and fuses the result back with a residual connection. The S2-MLP Link performs channel-wise spatial shifts and split attention to fuse encoder and decoder features. Taken together, these components produce the reported 89.04% DSC on LIDC-IDRI and 86.40% on the EPDB adenocarcinoma dataset, with ablation experiments attributing the largest single gain to the structured convolution blocks.

Load-bearing premise

The reported advantage assumes the LIDC-IDRI 9:1 split and the comparison protocols are fair, meaning slices from the same nodule are not shared between training and testing and all baselines were measured under identical preprocessing and metric definitions; if either condition fails, the 4.5-point DSC lead could evaporate.

Editorial extensions

If this is right

  • On the LIDC-IDRI benchmark, the full model improves DSC by 4.52% over USG-Net and 4.69% over SA-UNet, the two best prior baselines, and raises sensitivity by 3.16% over SA-UNet.
  • Ablation shows the structured convolution blocks (Block1) contribute the largest gain, adding 9.41% DSC over the U-Net baseline, which supports the design choice of specialized early-stage convolution.
  • The RM-SViT module achieves its best result with a single iteration and an 8x8 grid on 128x128 inputs, indicating that cheap superpixel attention suffices for this task size.
  • The EPDB result (86.40% DSC across AAH/AIS/MIA/IAC subtypes) suggests the model transfers to adenocarcinoma staging-relevant nodule shapes without retraining.

Reading between the lines

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

  • Because superpixel tokenization compresses the sequence length, the RM-SViT design may scale to 3D volumes or high-resolution images better than global ViTs, though the paper only demonstrates 2D 128x128 slices.
  • The slice-level data split, if not per-nodule, is a common source of inflated Dice; a per-patient split on LIDC-IDRI would give a more clinically meaningful estimate and would also let readers judge whether the claimed gain over baselines persists.
  • The reported mIoU higher than DSC is atypical for binary segmentation unless the IoU is computed on a probabilistic or multi-class basis; clarifying this definition would strengthen the comparison.
  • If the gains replicate under a fair protocol, the same three-module recipe (structured convolution, superpixel transformer, spatial-shift MLP skip) could be transferred to other small-structure segmentation tasks like skin lesions or prostate MRI.
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.

Referee Report

4 major / 5 minor

Summary. The paper proposes S3TU-Net, a U-shaped CNN-Transformer hybrid for lung nodule segmentation in CT images. The architecture combines two structured convolutional blocks (DWF-Conv, D2BR-Conv), a superpixel-based residual multi-branch vision transformer (RM-SViT), and a spatial-shift MLP skip-connection module (S2-MLP Link). The authors report a DSC of 89.04%, precision of 90.73%, mIoU of 90.70%, and sensitivity of 93.70% on LIDC-IDRI, claim a 4.52% DSC improvement over the best compared method, and report a DSC of 86.40% on a private EPDB dataset. The paper also includes ablation studies and comparisons against seven published segmentation networks.

Significance. If the reported results hold under a clean evaluation protocol, the architecture is a credible contribution to lung nodule segmentation: it combines several recent design ideas (large-kernel attention, DropBlock, spatial-shift MLPs, and superpixel transformers) in a U-Net framework and demonstrates generalization on an independent private dataset. The ablation study on EPDB is a useful strength. However, the paper does not release code or trained models, and the central quantitative claims currently rest on an evaluation protocol with several unresolved issues: inconsistent headline numbers, an undefined IoU metric that is reported above DSC, a slice-level data split that may leak nodule information, and test-set-based hyperparameter selection. For these reasons, the significance depends on whether the authors can repair the evaluation and confirm the reported margins.

major comments (4)
  1. [Section IV-B, Table V, Abstract] The numerical results are internally inconsistent. The abstract and Table V report DSC 89.04%, mIoU 90.70%, precision 90.73%, and sensitivity 93.70%, while the text of Section IV-B states that S3TU-Net achieves a maximum DSC of 88.87%, MIoU of 91.14%, sensitivity of 93.48%, and precision of 91.97%. These are different numbers, and the subsequent claims of improvements by 4.52% in DSC and 3.16% in sensitivity are computed from the Table V values. The authors must state which numbers are the official results, correct the inconsistency, and recompute all comparative claims consistently.
  2. [Section III-C, Section III-A] The LIDC-IDRI split is described only at the slice level: 5,717 training images and 757 test images from 6,474 total slices, with the same section reporting 1,303 nodules. The paper never states that the split is patient-disjoint or nodule-disjoint. Because adjacent CT slices of the same nodule are near-duplicates, a slice-level split can place nearly identical images in both training and test sets, inflating DSC and sensitivity through memorization. The authors must clarify the split granularity, and if it is slice-level, they must repeat the main comparison under a patient-level or nodule-level split and report whether the claimed 4.52% DSC advantage persists.
  3. [Section III-B, Eqs. (15)-(17), Table V] The metric definitions are incomplete and appear to produce impossible values. Equations (15)-(17) define DSC, sensitivity, and precision using raw probabilities, but no definition is given for IoU/mIoU, and no thresholding or binarization procedure is stated. In Table V, S3TU-Net reports mIoU 90.70% alongside DSC 89.04%; for any fixed pair of binary masks, IoU = DSC/(2-DSC), which would give at most about 80.3% foreground IoU for DSC 89.04%. If the reported mIoU is instead a mean of per-class IoUs (including background), that definition must be stated explicitly and must be used for all compared methods. The caption of Table III also labels results as '(Max)', indicating best-of-run selection; the authors should report mean and standard deviation over runs, or at least justify why maximum statistics are a valid basis for comparison.
  4. [Section IV-A.2, Fig. 9] The choice of the superpixel grid size (8x8) and the iteration count (1) is made by comparing performance in Fig. 9, which appears to be evaluated on the LIDC-IDRI test set. Selecting hyperparameters using the test set, and then reporting test-set performance with those hyperparameters, introduces selection bias and makes the reported margins over baselines (which are not given the same tuned hyperparameters) difficult to interpret. The authors should either tune on a held-out validation split and report test-set results, or disclose the selection procedure and quantify its effect.
minor comments (5)
  1. [Eqs. (7)-(8)] The spatial shift formulas in Equations (7) and (8) are formatted incorrectly: the piecewise cases are written as assignments with conditions placed on the right-hand side, making the intended operations unclear. Please rewrite them as standard piecewise definitions.
  2. [References] Several citations are mismatched: reference [41] is listed as a ResNeSt paper but is cited for LIDC-IDRI in Section III-A, and reference [45] is listed as a roof-extraction paper but is cited for Attention-UNet in Table V. The LIDC-IDRI dataset and the Attention-UNet method need correct citations. References [15] and [19] also appear to duplicate the same roof-extraction paper.
  3. [Fig. 7 caption] The caption of Fig. 7 refers to 'stage IV lung adenocarcinoma (AAH/MIA/IAC/AIS)', but AAH, AIS, MIA, and IAC are pathological subtypes, not stages; please correct the terminology.
  4. [Section IV-A.1] The text states that 'the DSC and sensitivity are increased by 9.41%' for Baseline+Block1 compared with Baseline; from Table III, DSC increases by 9.41 percentage points (77.44 to 86.85) but sensitivity increases by 8.41 percentage points (86.94 to 95.35), so the sentence should be corrected.
  5. [Section II-B] There is a typo in the description of DropBlock parameters: 'block size and y.block size' should presumably be 'block size and drop block size' or similar.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical architecture-and-evaluation study with no derivation chain reducing to its inputs; the external EPDB validation is independent.

full rationale

S3TU-Net is presented as a U-Net hybrid with structured convolution blocks, an RM-SViT transformer, and an S2-MLP Link. The module equations (Eqs. 1-14) are self-contained definitions of the proposed operators, not fitted quantities from which the reported DSC/mIoU numbers are derived. The headline LIDC-IDRI results (DSC 89.04%, etc.) are experimental measurements on a held-out test portion, and the EPDB private dataset provides an external generalization check, so the central claims are not equivalent to their inputs by construction. The ablation study selects a superpixel grid size of 8×8 and one iteration based on observed performance; this is model selection rather than a fitted parameter renamed as a prediction, and if the selection and the slice-level 9:1 split were performed on the same test data, that is a leakage/correctness risk, not a circularity step. The mIoU-above-DSC values and the '(Max)' statistics indicate metric-definition and best-of-run reporting concerns, but they do not make the derivation circular. No load-bearing self-citation chain or imported uniqueness theorem appears; references to prior work by potential co-authors (e.g., [10]) are contextual and are not used to justify the central result. The future-work section notes annotation challenges but does not assert a circular limitation. Accordingly, the circularity score is 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 4 invented entities

This ledger accounts for the paper's hand-chosen hyperparameters, its unstated evaluation assumptions, and its four named modules, which are compositions of published building blocks. There are no fitted physical constants or invented natural entities; the core uncertainty lives in the evaluation protocol: the test-set-driven choice of grid size and iteration count, the Max-versus-Average statistics, the slice-level split, and the undefined mIoU.

free parameters (6)
  • Superpixel grid size = 8x8
    Chosen by comparing 8x8 vs 16x16 performance on the test set; Section IV-A.2 concludes 8x8 is best for 128x128 inputs.
  • RM-SViT iteration count = 1
    Selected from iter = 0/1/2/3 experiments on the test set; Section IV-A.2.
  • DropBlock block size = 7
    Hand-set for 128x128 input images; Section III-C.
  • Initial learning rate = 0.001
    Adam optimizer with a scheduler and one warm-up epoch; Section III-C.
  • Batch size and epoch count = 16, 300
    Section III-C; no early stopping rule is specified, which matters because reported metrics are labeled Max.
  • Base channel width C
    The channel count of the first encoder stage and the growth factor across stages are never specified, so a re-implementation must guess them.
assumptions (5)
  • domain assumption Radiologist annotations in LIDC-IDRI (nodules marked by at least two of four radiologists) are treated as ground truth.
    Section III-A uses these annotations as targets without quantifying inter-observer variability.
  • domain assumption Slices are independent samples; no nodule- or patient-level split is enforced.
    Section III-C splits 6,474 slices at 9:1; adjacent slices of one nodule may co-occur in training and test sets.
  • ad hoc to paper DSC, sensitivity, and precision computed with raw probabilities (Eqs. 15-17) are comparable to standard thresholded metrics.
    The paper never discloses the binarization threshold, and the unstated mIoU computation produces values above DSC, which standard definitions cannot produce for the same masks.
  • ad hoc to paper Reporting maximum metrics over the training run is a valid summary.
    Table III header '(Max)' and Section IV report best-epoch numbers without specifying how the best epoch is chosen; EPDB numbers in Table IV are labeled '(Average)', mixing statistics.
  • domain assumption Grid-averaged superpixel initialization and iterative softmax association preserve nodule-relevant information.
    Section II-C Eqs. (1)-(3); at an 8x8 grid on 128x128 input, small early-stage nodules may be diluted by averaging with background.
invented entities (4)
  • DWF-Conv block
    purpose: Depth-weighted feature convolution using LKA and scaled ReLU at the first encoder and decoder stages.
    Composition of LKA (Visual Attention Network) and SE-style re-weighting; validated only on the paper's own datasets, no code released.
  • D2BR-Conv block
    purpose: Double convolution with DropBlock regularization to reduce overfitting in the middle stages.
    DropBlock plus standard double convolution; no external validation.
  • S2-MLP Link
    purpose: Spatial-shift MLP with split attention at the skip connections for multi-scale fusion.
    S2-MLP (Yu et al., 2022) plus ResNeSt split attention; new combination, validated only on the paper's datasets.
  • RM-SViT module
    purpose: Residual multi-branch attention over superpixel tokens to capture long-range dependencies.
    Reproduces the super-token mechanism of the superpixel transformer literature without citing it, with an added residual connection; no independent evidence outside this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of S3TU-Net: Structured Convolution and Superpixel Transformer for Lung Nodule Segmentation." pith.science (2026). https://pith.science/paper/Y5FXCUOD

@misc{pith2026241112547,
  author       = {Pith},
  title        = {Pith review of: S3TU-Net: Structured Convolution and Superpixel Transformer for Lung Nodule Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y5FXCUOD}},
  note         = {Machine review of arXiv:2411.12547}
}
read the original abstract

The irregular and challenging characteristics of lung adenocarcinoma nodules in computed tomography (CT) images complicate staging diagnosis, making accurate segmentation critical for clinicians to extract detailed lesion information. In this study, we propose a segmentation model, S3TU-Net, which integrates multi-dimensional spatial connectors and a superpixel-based visual transformer. S3TU-Net is built on a multi-view CNN-Transformer hybrid architecture, incorporating superpixel algorithms, structured weighting, and spatial shifting techniques to achieve superior segmentation performance. The model leverages structured convolution blocks (DWF-Conv/D2BR-Conv) to extract multi-scale local features while mitigating overfitting. To enhance multi-scale feature fusion, we introduce the S2-MLP Link, integrating spatial shifting and attention mechanisms at the skip connections. Additionally, the residual-based superpixel visual transformer (RM-SViT) effectively merges global and local features by employing sparse correlation learning and multi-branch attention to capture long-range dependencies, with residual connections enhancing stability and computational efficiency. Experimental results on the LIDC-IDRI dataset demonstrate that S3TU-Net achieves a DSC, precision, and IoU of 89.04%, 90.73%, and 90.70%, respectively. Compared to recent methods, S3TU-Net improves DSC by 4.52% and sensitivity by 3.16%, with other metrics showing an approximate 2% increase. In addition to comparison and ablation studies, we validated the generalization ability of our model on the EPDB private dataset, achieving a DSC of 86.40%.

Figures

Figures reproduced from arXiv: 2411.12547 by the authors.

Figure 1
Figure 1. The overall framework of S3TU-Net. The framework is divided into three broad categories of modules, two novel convolutional modules (DWF-Conv/ D2BR-Conv), multi-spatial dimensional connectors (S2 -MLP Link), and residual connection-based superpixel vision transformer (RM-SViT). vides a solution for long-distance dependency modeling and has shown promise in visual tasks. Vision Transformer (ViT) [30, 31] exemplifies … view at source ↗
Figure 2
Figure 2. The architecture of traditional convolutional block. (a) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The architecture of Newly proposed convolutional [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The architecture of RM-SViT Module. The encoder [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: The architecture of S2 -MLP Link Module. Firstly, MLP is used to expand the channel c of the feature map into 3 × c and divide it into three parts (F1,F2,F3) along the channel dimension. F1 and F2 are spatially shifted according to different directions, and F3 remains …
Figure 6
Figure 6. Figure 6: Comparison results of performance. Various performance comparison results on LIDC-IDRI dataset (Baseline/ +Block [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: The Example Images From the EPDB Dataset. Randomly shown are the original images, annotated masks, segmentation [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison (data were randomly extracted from LIDC-IDRI).(1) Original image. (2/3) Segmentation results [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Performance comparison on Block2. iter=0/1/2/3 and size=8 × 8/ 16 × 16.(a) System CPU utilization. (b) GPU power consumption. (c) Sensitivity. (d) MIoU. (e) Dice. (left to right, top to bottom) their generalization capability diminishes when handling com￾plex adenocarc…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 37 canonical work pages

  1. [45]

    A bag of tricks for fine-grained roof extraction,

    J. Hu, Z. Huang, F. Shen, D. He, and Q. Xian, “A bag of tricks for fine-grained roof extraction,” 2023

    work page 2023

  2. [1]

    Lung cancer prediction using electronic claims records: A transformer-based approach,

    H.-Y . Chen, H.-M. Wang, C.-H. Lin, R. Yang, and C.- C. Lee, “Lung cancer prediction using electronic claims records: A transformer-based approach,” IEEE Journal of Biomedical and Health Informatics , 2023

    work page 2023

  3. [2]

    Modality- specific segmentation network for lung tumor segmenta- tion in pet-ct images,

    D. Xiang, B. Zhang, Y . Lu, and S. Deng, “Modality- specific segmentation network for lung tumor segmenta- tion in pet-ct images,” IEEE Journal of Biomedical and Health Informatics, vol. 27, no. 3, pp. 1237–1248, 2022

    work page 2022

  4. [3]

    Deep learning methods for lung cancer segmentation in whole-slide histopathology images—the acdc@ lunghp challenge 2019,

    Z. Li, J. Zhang, T. Tan, X. Teng, X. Sun, H. Zhao, L. Liu, Y . Xiao, B. Lee, Y . Liet al., “Deep learning methods for lung cancer segmentation in whole-slide histopathology images—the acdc@ lunghp challenge 2019,” IEEE Jour- nal of Biomedical and Health Informatics , vol. 25, no. 2, pp. 429–440, 2020

    work page 2019

  5. [4]

    Spectrum of lung adenocarcinoma,

    B. D. Hutchinson, G. S. Shroff, M. T. Truong, and J. P. Ko, “Spectrum of lung adenocarcinoma,” in Seminars in Ultrasound, CT and MRI, vol. 40, no. 3. Elsevier, 2019, pp. 255–264

    work page 2019

  6. [5]

    Lung adenocarcinoma,

    D. J. Myers and J. M. Wallen, “Lung adenocarcinoma,” in StatPearls [Internet]. StatPearls Publishing, 2023

    work page 2023

  7. [6]

    Prognostic considerations of the new world health organization classification of lung adeno- carcinoma,

    A. C. Borczuk, “Prognostic considerations of the new world health organization classification of lung adeno- carcinoma,” European Respiratory Review , vol. 25, no. 142, pp. 364–371, 2016

    work page 2016

  8. [7]

    Who classification of tumours of the lung, pleura, thymus and heart,

    W. H. Organization et al., “Who classification of tumours of the lung, pleura, thymus and heart,” WHO/IARC Classification of Tumours, , vol. 7, 2015

    work page 2015

Show all 48 references
  1. [8]

    Role of pet/ct in management of early lung adenocarcinoma,

    X. Shao, R. Niu, Z. Jiang, X. Shao, and Y . Wang, “Role of pet/ct in management of early lung adenocarcinoma,” American Journal of Roentgenology , vol. 214, no. 2, pp. 437–445, 2020

    2020

  2. [9]

    Lung ade- nocarcinomas: correlation of computed tomography and pathology findings,

    J. Cohen, E. Reymond, A. Jankowski, E. Brambilla, F. Arbib, S. Lantuejoul, and G. Ferretti, “Lung ade- nocarcinomas: correlation of computed tomography and pathology findings,” Diagnostic and interventional imag- ing, vol. 97, no. 10, pp. 955–963, 2016

    2016

  3. [10]

    Integrating lung parenchyma segmentation and nodule detection with deep multi-task learning,

    W. Liu, X. Liu, H. Li, M. Li, X. Zhao, and Z. Zhu, “Integrating lung parenchyma segmentation and nodule detection with deep multi-task learning,” IEEE Journal of Biomedical and Health Informatics , vol. 25, no. 8, pp. 3073–3081, 2021

    2021

  4. [11]

    Advancing pose-guided image synthesis with progressive conditional diffusion models,

    F. Shen, H. Ye, J. Zhang, C. Wang, X. Han, and Y . Wei, “Advancing pose-guided image synthesis with progressive conditional diffusion models,” in The Twelfth International Conference on Learning Representations , 2023

    2023

  5. [12]

    Boosting consistency in story visualization with rich-contextual conditional diffusion models,

    F. Shen, H. Ye, S. Liu, J. Zhang, C. Wang, X. Han, and W. Yang, “Boosting consistency in story visualization with rich-contextual conditional diffusion models,” arXiv preprint arXiv:2407.02482, 2024

    2024 arXiv

  6. [13]

    Pulmonary nodule detection in medical images: a survey,

    J. Zhang, Y . Xia, H. Cui, and Y . Zhang, “Pulmonary nodule detection in medical images: a survey,” Biomedi- cal Signal Processing and Control, vol. 43, pp. 138–147, 2018

    2018

  7. [14]

    Deep feature transfer learning in combination with traditional features predicts survival among patients with lung adenocarcinoma,

    R. Paul, S. H. Hawkins, Y . Balagurunathan, M. B. Schabath, R. J. Gillies, L. O. Hall, and D. B. Goldgof, “Deep feature transfer learning in combination with traditional features predicts survival among patients with lung adenocarcinoma,” Tomography, vol. 2, no. 4, p. 388, 2016

    2016

  8. [15]

    A bag of tricks for fine-grained roof extraction,

    J. Hu, Z. Huang, F. Shen, D. He, and Q. Xian, “A bag of tricks for fine-grained roof extraction,” in IGARSS 2023-2023 IEEE International Geoscience and Remote Sensing Symposium. IEEE, 2023

    2023

  9. [16]

    A fast weak-supervised pulmonary nodule segmentation method based on modified self-adaptive fcm algorithm,

    H. Liu, F. Geng, Q. Guo, C. Zhang, and C. Zhang, “A fast weak-supervised pulmonary nodule segmentation method based on modified self-adaptive fcm algorithm,” Soft Computing, vol. 22, pp. 3983–3995, 2018

    2018

  10. [17]

    Au- tomated pulmonary nodule detection in ct images using deep convolutional neural networks,

    H. Xie, D. Yang, N. Sun, Z. Chen, and Y . Zhang, “Au- tomated pulmonary nodule detection in ct images using deep convolutional neural networks,”Pattern recognition, vol. 85, pp. 109–119, 2019

    2019

  11. [18]

    Pulmonary lung nodule detection from computed tomography images using two-stage convolutional neural network,

    S. Jain, P. Choudhari, and M. Gour, “Pulmonary lung nodule detection from computed tomography images using two-stage convolutional neural network,”The Com- puter Journal, vol. 66, no. 4, pp. 785–795, 2023

    2023

  12. [19]

    A rubust method for roof extraction and height estimation,

    J. Hu, Z. Huang, F. Shen, D. He, and Q. Xian, “A rubust method for roof extraction and height estimation,” in IGARSS 2023-2023 IEEE International Geoscience and Remote Sensing Symposium . IEEE, 2023

    2023

  13. [20]

    Segmentation of the pulmonary nodule and the attached vessels in the ct scan of the chest using morphological features and topological skeleton of the nodule,

    M. B. Tavakoli, M. Orooji, M. Teimouri, and R. Sha- habifar, “Segmentation of the pulmonary nodule and the attached vessels in the ct scan of the chest using morphological features and topological skeleton of the nodule,” IET Image Processing, vol. 14, no. 8, pp. 1520– 1528, 2020

    2020

  14. [21]

    Clustering by transmission learning from data density to label manifold with statistical diffusion,

    Y . Zhang, F.-l. Chung, and S. Wang, “Clustering by transmission learning from data density to label manifold with statistical diffusion,” Knowledge-Based Systems , vol. 193, p. 105330, 2020

    2020

  15. [22]

    U-net: Con- volutional networks for biomedical image segmenta- tion,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Con- volutional networks for biomedical image segmenta- tion,” in Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international confer- ence, Munich, Germany, October 5-9, 2015, proceedings, part III...

    2015

  16. [23]

    Review of semantic segmentation of medical images using mod- ified architectures of unet,

    M. Krithika Alias AnbuDevi and K. Suganthi, “Review of semantic segmentation of medical images using mod- ified architectures of unet,” Diagnostics, vol. 12, no. 12, p. 3064, 2022

    2022

  17. [24]

    A cascaded dual-pathway residual network for lung nodule segmentation in ct images,

    H. Liu, H. Cao, E. Song, G. Ma, X. Xu, R. Jin, Y . Jin, and C.-C. Hung, “A cascaded dual-pathway residual network for lung nodule segmentation in ct images,” Physica Medica, vol. 63, pp. 112–121, 2019

    2019

  18. [25]

    A novel deep learning network and its application for pulmonary nodule segmentation,

    D. Lu, J. Chu, R. Zhao, Y . Zhang, and G. Tian, “A novel deep learning network and its application for pulmonary nodule segmentation,” Computational Intelligence and Neuroscience, vol. 2022, no. 1, p. 7124902, 2022

    2022

  19. [26]

    Pulmonary nodules segmentation based on crf 3d-unet structure,

    T. Hou, J. Zhao, Y . Qiang, S. Wang, and P. Wang, “Pulmonary nodules segmentation based on crf 3d-unet structure,” Comput Eng Des , vol. 41, no. 6, pp. 1663– 1669, 2020

    2020

  20. [27]

    Vits vs. cnns for 3d medical image segmentation: Are transformers all you need?

    X. Arregui Garc ´ıa, “Vits vs. cnns for 3d medical image segmentation: Are transformers all you need?” Master’s thesis, 2023

    2023

  21. [28]

    Attention is all you need,

    A. Vaswani, “Attention is all you need,” Advances in Neural Information Processing Systems , 2017

    2017

  22. [29]

    Enhancing landslide segmen- tation with guide attention mechanism and fast fourier transformer,

    K. Yan, F. Shen, and Z. Li, “Enhancing landslide segmen- tation with guide attention mechanism and fast fourier transformer,” in International Conference on Intelligent Computing. Springer, 2024, pp. 296–307

    2024

  23. [30]

    An image is worth 16x16 words: Trans- formers for image recognition at scale,

    A. Dosovitskiy, “An image is worth 16x16 words: Trans- formers for image recognition at scale,” arXiv preprint arXiv:2010.11929, 2020

    2010 arXiv

  24. [31]

    Swin transformer: Hierarchical vi- sion transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vi- sion transformer using shifted windows,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 012–10 022

    2021

  25. [32]

    Dropout: a simple way to pre- vent neural networks from overfitting,

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: a simple way to pre- vent neural networks from overfitting,” The journal of machine learning research, vol. 15, no. 1, pp. 1929–1958, 2014

    1929

  26. [33]

    Dropblock: A regu- larization method for convolutional networks,

    G. Ghiasi, T.-Y . Lin, and Q. V . Le, “Dropblock: A regu- larization method for convolutional networks,” Advances in neural information processing systems , vol. 31, 2018

    2018

  27. [34]

    Visual attention network,

    M.-H. Guo, C.-Z. Lu, Z.-N. Liu, M.-M. Cheng, and S.- M. Hu, “Visual attention network,” Computational Visual Media, vol. 9, no. 4, pp. 733–752, 2023

    2023

  28. [35]

    Freeu: Free lunch in diffusion u-net,

    C. Si, Z. Huang, Y . Jiang, and Z. Liu, “Freeu: Free lunch in diffusion u-net,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 4733–4743

    2024

  29. [36]

    Squeeze-and-excitation networks,

    J. Hu, L. Shen, and G. Sun, “Squeeze-and-excitation networks,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 7132– 7141

    2018

  30. [37]

    Malunet: A multi-attention and light-weight unet for skin lesion segmentation,

    J. Ruan, S. Xiang, M. Xie, T. Liu, and Y . Fu, “Malunet: A multi-attention and light-weight unet for skin lesion segmentation,” in 2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) . IEEE, 2022, pp. 1150–1156

    2022

  31. [38]

    S2-mlp: Spatial-shift mlp architecture for vision,

    T. Yu, X. Li, Y . Cai, M. Sun, and P. Li, “S2-mlp: Spatial-shift mlp architecture for vision,” in Proceedings of the IEEE/CVF winter conference on applications of computer vision, 2022, pp. 297–306

    2022

  32. [39]

    Mlp-mixer: An all-mlp architecture for vision,

    I. O. Tolstikhin, N. Houlsby, A. Kolesnikov, L. Beyer, X. Zhai, T. Unterthiner, J. Yung, A. Steiner, D. Keysers, J. Uszkoreit et al. , “Mlp-mixer: An all-mlp architecture for vision,” Advances in neural information processing systems, vol. 34, pp. 24 261–24 272, 2021

    2021

  33. [40]

    Multilayer perceptron (mlp),

    H. Taud and J.-F. Mas, “Multilayer perceptron (mlp),” Geomatic approaches for modeling land change scenar- ios, pp. 451–455, 2018

    2018

  34. [41]

    Resnest: Split-attention networks,

    H. Zhang, C. Wu, Z. Zhang, Y . Zhu, H. Lin, Z. Zhang, Y . Sun, T. He, J. Mueller, R. Manmatha et al. , “Resnest: Split-attention networks,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 2736–2746

    2022

  35. [42]

    Y-net: joint segmentation and classifi- cation for diagnosis of breast biopsy images,

    S. Mehta, E. Mercan, J. Bartlett, D. Weaver, J. G. Elmore, and L. Shapiro, “Y-net: joint segmentation and classifi- cation for diagnosis of breast biopsy images,” in Medical Image Computing and Computer Assisted Intervention– MICCAI 2018: 21st International Conference, Granada...

    2018

  36. [43]

    Unet++: A nested u-net architecture for medical image segmentation,

    Z. Zhou, M. M. Rahman Siddiquee, N. Tajbakhsh, and J. Liang, “Unet++: A nested u-net architecture for medical image segmentation,” in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: 4th International Workshop, DLMIA 2018, and 8th ...

    2018

  37. [44]

    Recurrent residual convolutional neural network based on u-net (r2u-net) for medical image segmentation,

    M. Z. Alom, M. Hasan, C. Yakopcic, T. M. Taha, and V . K. Asari, “Recurrent residual convolutional neural network based on u-net (r2u-net) for medical image segmentation,” arXiv preprint arXiv:1802.06955 , 2018

    2018 arXiv

  38. [46]

    Unet 3+: A full- scale connected unet for medical image segmentation,

    H. Huang, L. Lin, R. Tong, H. Hu, Q. Zhang, Y . Iwamoto, X. Han, Y .-W. Chen, and J. Wu, “Unet 3+: A full- scale connected unet for medical image segmentation,” in ICASSP 2020-2020 IEEE international conference on acoustics, speech and signal processing (ICASSP) . IEEE, 2020, ...

    2020

  39. [47]

    Sa-unet: Spatial attention u-net for retinal vessel segmentation,

    C. Guo, M. Szemenyei, Y . Yi, W. Wang, B. Chen, and C. Fan, “Sa-unet: Spatial attention u-net for retinal vessel segmentation,” in 2020 25th international conference on pattern recognition (ICPR). IEEE, 2021, pp. 1236–1242

    2020

  40. [48]

    Uncertainty- guided lung nodule segmentation with feature-aware attention,

    H. Yang, L. Shen, M. Zhang, and Q. Wang, “Uncertainty- guided lung nodule segmentation with feature-aware attention,” in International Conference on Medical Im- age Computing and Computer-Assisted Intervention . Springer, 2022, pp. 44–54

    2022

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.