Pith. sign in

REVIEW 5 major objections 7 minor 41 references

D-UNet: a dimension-fusion U shape network for chronic stroke lesion segmentation

T0 review · 5 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A 2D-3D fusion encoder lifts stroke lesion DSC to 0.535.

desk verdict Incremental 2D/3D fusion UNet with a useful ablation, but the headline gain over UNet is confounded by an uncontrolled loss function and a single noisy split. read the letter →

arxiv 1908.05104 v1 pith:3WT5VGV6 submitted 2019-08-14 eess.IV cs.CV

Yongjin Zhou , Weijian Huang , Pei Dong , Yong Xia , Shanshan Wang This is my paper · ORCID
classification eess.IVcs.CV
keywords D-UNetdimensionfusion2D-3DhybridCNNEnhancedMixingLossstrokelesionsegmentationATLASdatasetMRI
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 proposes a U-shaped segmentation network, D-UNet, that mixes 2D and 3D convolutions in the encoder so that a mostly 2D model can still exploit the volumetric context of MRI scans. On the ATLAS chronic stroke dataset, the authors report a Dice similarity coefficient of 0.5349±0.2763 and a precision of 0.6331±0.2958, outperforming a tuned 2D UNet (DSC 0.4966) and a 3D UNet (DSC 0.4710) while adding only about 2% more parameters than the 2D network. The paper also introduces an Enhanced Mixing Loss that combines focal loss with a log-transformed Dice loss, which they report converges faster than either loss alone. If the results hold, the practical message is that volumetric context does not require a full 3D network; a light 2D-plus-early-3D fusion can capture enough 3D information for small, irregular lesions at low computational cost.

What carries the argument

The dimension-fusion block is the load-bearing component: a shallow 3D branch runs 3D convolutions over a stack of four consecutive slices in the early encoder, and its output is channel-compressed with a $1\times1\times1$ convolution, spatially squeezed, and projected with a 2D $3\times3$ convolution to match the 2D branch; both branches are then reweighted by squeeze-and-excitation blocks and added. Fusion blocks are placed after the second and third downsampling stages, and the paper finds that adding a third fusion stage degrades performance. The second component, Enhanced Mixing Loss, combines a voxel-averaged focal loss with the negative logarithm of the Dice loss, $EML = \frac{1}{N}FL - \log(DL)$, which the paper reports accelerates and smooths convergence.

What would settle it

Rerun D-UNet and the tuned 2D UNet on ATLAS across several random train/validation splits with the loss hyperparameters held at their default values; if the mean DSC gap (reported as 0.038) falls within the per-case standard deviation of about 0.28, the architecture's benefit over the 2D baseline would not be established.

Watch

Extended reading notes

Core claim

The central claim is that fusing 3D feature maps into the early downsampling stages of a 2D U-Net, through a dimension-fusion block, improves chronic stroke lesion segmentation relative to both pure 2D and pure 3D U-Nets on the ATLAS dataset. The D-UNet reaches a per-case DSC of 0.5349±0.2763 and precision of 0.6331±0.2958, beating the tuned 2D UNet (DSC 0.4966) and the 3D UNet (DSC 0.4710) with only about 2% more parameters than the 2D baseline. The authors further claim that the Enhanced Mixing Loss, $EML(p,g)=\frac{1}{N}FL(p,g)-\log(DL(p,g))$, produces faster and smoother training convergence than focal loss or Dice loss alone, while slightly improving precision at a small cost in DSC.

Load-bearing premise

The reported advantage of D-UNet over the 2D UNet is attributed to the architecture itself, rather than to the specific random train/validation split or the loss parameters tuned on that split.

Editorial extensions

If this is right

  • On the ATLAS dataset, D-UNet improves per-case DSC by about 3.8 percentage points over a tuned 2D UNet, while increasing total parameters by only about 2%.
  • The hybrid encoder reaches a higher per-case DSC and precision than a pure 3D UNet while avoiding the 3D network's heavy memory and training-time burden.
  • Fusion after the second and third downsampling stages (SEAdd-23) gives the best DSC; fusing at all three stages (SEAdd-123) lowers DSC, indicating the fusion depth has an optimum.
  • The Enhanced Mixing Loss converges faster on the training curve than focal loss or Dice loss alone, and it achieves the highest precision among the three losses tested.

Reading between the lines

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

  • The design suggests a general recipe for other volumetric segmentation tasks: capture local 3D context in a shallow, cheap branch, then let a 2D network refine boundaries; this is a testable extension to datasets beyond ATLAS.
  • The drop in DSC when fusion is added at every downsampling stage hints that deep 2D/3D fusion may create gradient conflicts; ablating fusion depth on other datasets could determine whether the optimum is dataset-specific or general.
  • Because the loss hyperparameters ($\alpha=1.1$, $\gamma=0.48$, $\delta=1$) were chosen to fit the single random split, the reported convergence advantage of EML over Dice loss may depend on that tuning; fixing these parameters across multiple splits is a straightforward robustness check.
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

5 major / 7 minor

Summary. The paper proposes D-UNet, a hybrid 2D/3D U-shaped network for chronic stroke lesion segmentation from T1-weighted MRI. The encoder runs parallel 2D and 3D convolutional streams in the early downsampling stages and fuses them through a dimension-transform block that includes squeeze-and-excitation weighting. The paper also proposes an Enhanced Mixing Loss (EML) that combines focal loss and a log Dice loss. The method is evaluated on the public ATLAS dataset with a single random 183/46 train/validation split, reporting DSC = 0.5349±0.2763 and precision = 0.6331±0.2958, which the authors claim is better than 2D and 3D UNet baselines and several other segmentation frameworks while adding few parameters.

Significance. If the central claim were fully established, the contribution would be practically useful: the architecture is comparatively simple, the parameter overhead over a 2D UNet is modest, and the evaluation on a public dataset (ATLAS) supports reproducibility. The paper also provides useful ablations in Table 5 on the placement of the fusion block and the value of SE weighting, and a loss comparison in Table 4/Fig. 5. However, the main quantitative claim is currently not supported by the experiments as reported: the comparison in Table 2 does not control for the loss function, the reported gains are within the per-case standard deviation, the hyperparameters were tuned to the validation split, and the parameter-efficiency claim is arithmetically inconsistent. These issues are fixable with additional controlled experiments and statistical reporting, so the work is a reasonable candidate for major revision rather than rejection.

major comments (5)
  1. [Section 4.1, Table 2] The comparison confounds architecture and loss. The manuscript never states which loss function was used to train 2D UNet(original), 2D UNet(transform), or 3D UNet(transform), whereas the proposed method is explicitly trained with EML. This matters because Table 4 shows EML is not uniformly better than plain Dice loss on D-UNet itself (global DSC 0.7231 vs. 0.7346). The reported +0.038 DSC gain over the 2D UNet(transform) cannot therefore be attributed to the dimension-fusion block. Please re-run all baselines with the same loss function (e.g., EML or Dice loss) and report those results.
  2. [Section 3.3, Section 4.1, Table 2] The statistical basis for the main claim is weak. The evaluation uses a single random split of 183/46 cases with no cross-validation or repeated runs, and no significance tests are reported. The per-case DSC standard deviation is about 0.28, which is much larger than the reported improvement of 0.038. Additionally, the loss parameters α=1.1, γ=0.48, δ=1 are stated in Section 3.3 to be set "to fit our randomly selected dataset," which risks overfitting the validation split. Please report results over multiple splits or seeds, with paired significance tests and confidence intervals on per-case DSC.
  3. [Section 4.1, Table 2] The claim that the proposed method increases parameters by only "2%" is arithmetically incorrect. Comparing 8,640,163 parameters with the 7,771,297 parameters of 2D UNet(transform) gives an increase of roughly 11.2%, not 2%. Please correct this number or adjust the architecture description so the efficiency claim is accurate.
  4. [Section 4.3, Table 4] The loss-function validity experiment does not support the role of EML in the main result. On the proposed architecture, EML achieves lower global DSC than Dice loss (0.7231 vs. 0.7346) and lower recall, with only precision improved. The paper characterizes this as acceptable because the goal is faster convergence, but the convergence claim is supported only by the training-curve figure, not by quantitative convergence-time measurements. Please provide a criterion for the claimed advantage of EML and, if EML is to be credited for the Table 2 gains, demonstrate this in the controlled comparison requested above.
  5. [Section 4.4, Table 5] The internal ablation is informative but still statistically fragile. The best configuration, SEAdd-23, differs from Add-23 by only 0.0101 DSC (0.5349 vs. 0.5248), well within the reported per-case standard deviations, and all results come from the same single split. Please report repeated-run variability or a statistical test to support the conclusion that SE weighting and fusion depth reliably improve performance.
minor comments (7)
  1. [Abstract] There is a typographical error in the results: "DSC = 0.5349+0.2763" should be "0.5349±0.2763", and the final metric "precision = 0.6331±0.295" is missing the closing parenthesis or trailing digits.
  2. [Introduction] The word "infraction" in the first paragraph should be "infarction".
  3. [Section 3.3] The initialization method "Hes method" should be "He method" (after He et al. [35]).
  4. [Section 4.2, text after Table 3] The text states that the proposed method has the highest precision score of "0.6631", but Table 3 reports 0.6331. Please correct the inconsistency.
  5. [Section 4.2] The sentence "This is 0.30 higher than the Clusterize method in DSC" is numerically inconsistent with Table 3: the difference between 0.7231 and 0.23 is 0.4931, not 0.30.
  6. [Fig. 3 caption vs. Section 4.2] The caption lists "Baseline, DenseUnet, DeepLabv3+, PSPNet, and FCN-8s" while the text and Table 3 mention SegNet, PSP, and DeepLab v3 plus; the naming should be unified.
  7. [Section 4.1] The sentence "The results we reported is also consistent with that of [29]" contains a subject-verb agreement error; please rephrase.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the proposed D-UNet architecture and EML loss are original constructions evaluated on the external ATLAS benchmark, and no load-bearing step reduces to an input by definition or self-citation.

full rationale

The paper's central claims are empirical: a new architecture (D-UNet) and a new loss (EML) are defined explicitly in Sections 3.1 and 3.2, then evaluated on the public ATLAS dataset against reimplemented baselines and published methods. There is no derivation chain in which an output quantity is defined in terms of the quantity it is supposed to predict, and no fitted parameter is renamed as a prediction. The only parameter-tuning statement is in Section 3.3: 'α,γ,δ in the loss function is set to 1.1, 0.48, 1 respectly to fit our randomly selected dataset.' That is hyperparameter selection on the evaluation dataset, which raises overfitting and generalizability concerns but is not circularity: the reported DSC values are measured outcomes, not consequences of the tuning formula by construction. Likewise, the fact that Table 2 does not state which loss was used for the UNet baselines and Table 4 shows EML is not uniformly better than Dice loss is a confounding/control weakness that undermines attribution of the gain to the dimension-fusion block; it is an experimental-design issue, not a self-referential derivation. The paper also does not rely on self-citations: the cited blocks and losses (UNet, SE, focal loss, Dice loss) are external standard works, and no 'uniqueness theorem' or prior result by the authors is invoked to force the proposed choice. The claims are thus self-contained against an external benchmark, so the appropriate circularity score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several hand-chosen hyperparameters (loss α, γ, δ, SE ratio, crop region, slice stacking) and on the empirical assumption that the 2D+3D fusion and the new loss generalize from a single split of the ATLAS dataset. No new physical or mathematical entities are introduced.

free parameters (6)
  • alpha (focal loss balance factor) = 1.1
    Set 'to fit our randomly selected dataset' (Section 3.3); tunes the loss for the specific data split, a post-hoc choice that affects the reported loss comparison.
  • gamma (focal loss modulating factor) = 0.48
    Same as alpha, chosen to fit the dataset (Section 3.3).
  • delta (Dice loss smoothing term) = 1
    Same, set to 1 to fit the dataset (Section 3.3).
  • SE reduction ratio r = 16
    Hyperparameter in the dimension transform block (Section 3.1, 3.3), chosen by hand to balance capacity and cost.
  • Crop region coordinates = (10,40)-(190,220)
    Manual preprocessing choice that discards all image content outside a fixed square (Section 3.3); if a lesion lies outside this region, segmentation will fail.
  • Input slice stacking pattern = 2 upper + 1 lower slices (4 total)
    Empirically chosen arrangement to form the 192x192x4 network input (Section 3.3); no justification given for why this stacking preserves 3D context.
assumptions (4)
  • domain assumption Lesions are visible in T1-weighted MRI and located within the fixed cropped square region.
    The preprocessing in Section 3.3 crops every image to coordinates (10,40)-(190,220). The method cannot segment lesions outside this region, so the assumption that all relevant lesions fall inside is load-bearing.
  • domain assumption The 2+1 slice stacking arrangement captures enough 3D context to improve segmentation.
    The claimed benefit of the 3D branch relies on this heuristic input construction (Section 3.3). The paper does not test alternative stacking patterns.
  • ad hoc to paper The SE block weighting improves the fusion of 2D and 3D features in the dimension transform block.
    The authors assert in Section 4.4 that SE weighting enhances fusion because SE variants score higher than Add variants, but no derivation or theoretical justification is given.
  • ad hoc to paper The background/foreground voxel imbalance is best handled by the proposed Enhanced Mixing Loss formula.
    EML is proposed in Section 3.2.3 as a combination of focal and log-Dice losses. The paper provides no derivation for why equal weighting (via 1/N) is optimal, and Table 4 shows EML has lower global DSC than plain Dice loss.

how reviews work

0 comments
Cite this review

Pith. "Pith review of D-UNet: a dimension-fusion U shape network for chronic stroke lesion segmentation." pith.science (2026). https://pith.science/paper/3WT5VGV6

@misc{pith2026190805104,
  author       = {Pith},
  title        = {Pith review of: D-UNet: a dimension-fusion U shape network for chronic stroke lesion segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3WT5VGV6}},
  note         = {Machine review of arXiv:1908.05104}
}
read the original abstract

Assessing the location and extent of lesions caused by chronic stroke is critical for medical diagnosis, surgical planning, and prognosis. In recent years, with the rapid development of 2D and 3D convolutional neural networks (CNN), the encoder-decoder structure has shown great potential in the field of medical image segmentation. However, the 2D CNN ignores the 3D information of medical images, while the 3D CNN suffers from high computational resource demands. This paper proposes a new architecture called dimension-fusion-UNet (D-UNet), which combines 2D and 3D convolution innovatively in the encoding stage. The proposed architecture achieves a better segmentation performance than 2D networks, while requiring significantly less computation time in comparison to 3D networks. Furthermore, to alleviate the data imbalance issue between positive and negative samples for the network training, we propose a new loss function called Enhance Mixing Loss (EML). This function adds a weighted focal coefficient and combines two traditional loss functions. The proposed method has been tested on the ATLAS dataset and compared to three state-of-the-art methods. The results demonstrate that the proposed method achieves the best quality performance in terms of DSC = 0.5349+0.2763 and precision = 0.6331+0.295).

Figures

Figures reproduced from arXiv: 1908.05104 by the authors.

Figure 1
Figure 1. The MRI T1 sequence stroke image from the ATLAS dataset. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The entire D-UNet architecture is shown in (a). This network improves 2D UNet, which combines 3D convolution in the downsampling phase [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparisons of our method, Baseline, DenseUnet, DeepLabv3+, PSPNet, and FCN-8s on four different patients. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Box plots of DSC score results for different methods. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: The DSC score curve of the training set during the training [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 40 canonical work pages

  1. [1]

    Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008[J]

    Ferlay J, Shin H R, Bray F, et al. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008[J]. International journal of cancer, 2010, 127(12): 2893-2917

    work page 2008

  2. [2]

    Liver tumor volume estimation by semi-automatic segmentation method[C]

    Lu R, Marziliano P , Thng C H. Liver tumor volume estimation by semi-automatic segmentation method[C]. 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference. IEEE, 2006: 3296-3299

    work page 2005

  3. [3]

    Acute ischemic stroke[J]

    Van der Worp H B, van Gijn J. Acute ischemic stroke[J]. New England Journal of Medicine, 2007, 357(6): 572-579

    work page 2007

  4. [4]

    Cost of stroke in the United King- dom[J]

    Saka , McGuire A, Wolfe C. Cost of stroke in the United King- dom[J]. Age and ageing, 2009, 38(1): 27-32

    work page 2009

  5. [5]

    Interrater agreement for final infarct MRI lesion delineation[J]

    Neumann A B, Jonsdottir K Y, Mouridsen K, et al. Interrater agreement for final infarct MRI lesion delineation[J]. Stroke, 2009, 40(12): 3768-3771

    work page 2009

  6. [6]

    Measurement of in- farct volume in stroke patients using adaptive segmentation of diffusion weighted MR images[C]

    Martel A L, Allder S J, Delay G S, et al. Measurement of in- farct volume in stroke patients using adaptive segmentation of diffusion weighted MR images[C]. International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, Berlin, Heidelberg, 1999: 22-31

    work page 1999

  7. [7]

    A large, open source dataset of stroke anatomical brain images and manual lesion segmentations[J]

    Liew S L, Anglin J M, Banks N W, et al. A large, open source dataset of stroke anatomical brain images and manual lesion segmentations[J]. Scientific data, 2018, 5: 180011

    work page 2018

  8. [8]

    Ishemic Stroke Lesion Segmentation by Analyzing MRI Images Using Dilated and Transposed Convolutions in Con- volutional Neural Networks[C]

    Joshi S, Gore S. Ishemic Stroke Lesion Segmentation by Analyzing MRI Images Using Dilated and Transposed Convolutions in Con- volutional Neural Networks[C]. 2018 Fourth International Con- ference on Computing Communication Control and Automation (ICCUBEA). IEEE, 2018: 1-5

    work page 2018

Show all 41 references
  1. [9]

    Towards clinical diagnosis: Automated stroke lesion segmentation on multi-spectral MR image using con- volutional neural network[J]

    Liu Z, Cao C, Ding S, et al. Towards clinical diagnosis: Automated stroke lesion segmentation on multi-spectral MR image using con- volutional neural network[J]. IEEE Access, 2018, 6: 57006-57016. 10

    2018

  2. [10]

    Automatic segmentation of acute ischemic stroke from DWI using 3-D fully convolutional DenseNets[J]

    Zhang R, Zhao L, Lou W, et al. Automatic segmentation of acute ischemic stroke from DWI using 3-D fully convolutional DenseNets[J]. IEEE transactions on medical imaging, 2018, 37(9): 2149-2160

    2018

  3. [11]

    Stroke lesion detection using convolutional neural networks[C]

    Pereira D R, Reboucas Filho P P , de Rosa G H, et al. Stroke lesion detection using convolutional neural networks[C]. 2018 International joint conference on neural networks (IJCNN). IEEE, 2018: 1-6

    2018

  4. [12]

    Optimizing U- Net to Segment Left Ventricle from Magnetic Resonance Imag- ing[C]

    Charmchi S, Punithakumar K, Boulanger P . Optimizing U- Net to Segment Left Ventricle from Magnetic Resonance Imag- ing[C]. 2018 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE, 2018: 327-332

    2018

  5. [13]

    Classification of Atrial Fib- rillation with Pre-Trained Convolutional Neural Network Mod- els[C]

    Qayyum A, Meriaudeau F, Chan G C Y. Classification of Atrial Fib- rillation with Pre-Trained Convolutional Neural Network Mod- els[C]. 2018 IEEE-EMBS Conference on Biomedical Engineering and Sciences (IECBES). IEEE, 2018: 594-599

    2018

  6. [14]

    Drinet for medical image seg- mentation[J]

    Chen L, Bentley P , Mori K, et al. Drinet for medical image seg- mentation[J]. IEEE transactions on medical imaging, 2018, 37(11): 2453-2462

    2018

  7. [15]

    Automatic real-time CNN- based neonatal brain ventricles segmentation[C]

    Wang P , Cuccolo N G, Tyagi R, et al. Automatic real-time CNN- based neonatal brain ventricles segmentation[C]. 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018). IEEE, 2018: 716-719

    2018

  8. [16]

    3D fully convolutional networks for co-segmentation of tumors on PET-CT images[C]

    Zhong Z, Kim Y, Zhou L, et al. 3D fully convolutional networks for co-segmentation of tumors on PET-CT images[C]. 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018). IEEE, 2018: 228-231

    2018

  9. [17]

    Image segmentation of liver CT based on fully convolutional network[C]

    Jin X, Ye H, Li L, et al. Image segmentation of liver CT based on fully convolutional network[C]. 2017 10th International Sym- posium on Computational Intelligence and Design (ISCID). IEEE, 2017, 1: 210-213

    2017

  10. [18]

    Querying Representative and Informative Super-pixels for Filament Segmentation in Bioim- ages[J]

    Shao W, Huang S J, Liu M X, et al. Querying Representative and Informative Super-pixels for Filament Segmentation in Bioim- ages[J]. IEEE/ACM transactions on computational biology and bioinformatics, 2019

    2019

  11. [19]

    Multivariate dy- namic prediction of ischemic infarction and tissue salvage as a function of time and degree of recanalization[J]

    Kemmling A, Flottmann F, Forkert N D, et al. Multivariate dy- namic prediction of ischemic infarction and tissue salvage as a function of time and degree of recanalization[J]. Journal of Cerebral Blood Flow & Metabolism, 2015, 35(9): 1397-1405

    2015

  12. [20]

    Texture-based treatment prediction by automatic liver tumor segmentation on computed tomography

    Kuo, C.-L., et al. Texture-based treatment prediction by automatic liver tumor segmentation on computed tomography. in Computer, Information and Telecommunication Systems (CITS), 2017 Interna- tional Conference on. 2017. IEEE

    2017

  13. [21]

    An active learn- ing approach for stroke lesion segmentation on multimodal MRI data[J]

    Chyzhyk D, Dacosta-Aguayo R, Matar M, et al. An active learn- ing approach for stroke lesion segmentation on multimodal MRI data[J]. Neurocomputing, 2015, 150: 26-36

    2015

  14. [22]

    An efficient automated methodol- ogy for detecting and segmenting the ischemic stroke in brain MRI images[J]

    Sivakumar P , Ganeshkumar P . An efficient automated methodol- ogy for detecting and segmenting the ischemic stroke in brain MRI images[J]. International Journal of Imaging Systems and Technology, 2017, 27(3): 265-272

    2017

  15. [23]

    Fully automatic acute ischemic le- sion segmentation in DWI using convolutional neural networks[J]

    Chen L, Bentley P , Rueckert D. Fully automatic acute ischemic le- sion segmentation in DWI using convolutional neural networks[J]. NeuroImage: Clinical, 2017, 15: 633-643

    2017

  16. [24]

    Automatic semantic segmentation of brain gliomas from MRI images using a deep cascaded neural network[J]

    Cui S, Mao L, Jiang J, et al. Automatic semantic segmentation of brain gliomas from MRI images using a deep cascaded neural network[J]. Journal of healthcare engineering, 2018

    2018

  17. [25]

    Efficient multi- scale 3D CNN with fully connected CRF for accurate brain lesion segmentation[J]

    Kamnitsas K, Ledig C, Newcombe V F J, et al. Efficient multi- scale 3D CNN with fully connected CRF for accurate brain lesion segmentation[J]. Medical image analysis, 2017, 36: 61-78

    2017

  18. [26]

    A Deep Learning Approach for Targeted Contrast-Enhanced Ultrasound Based Prostate Cancer Detection[J]

    Feng Y, Yang F, Zhou X, et al. A Deep Learning Approach for Targeted Contrast-Enhanced Ultrasound Based Prostate Cancer Detection[J]. IEEE/ACM transactions on computational biology and bioinformatics, 2018

    2018

  19. [27]

    Deep convolutional neu- ral networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning[J]

    Shin H C, Roth H R, Gao M, et al. Deep convolutional neu- ral networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning[J]. IEEE transactions on medical imaging, 2016, 35(5): 1285-1298

    2016

  20. [28]

    3D convolutional neural networks for human action recognition[J]

    Ji S, Xu W, Yang M, et al. 3D convolutional neural networks for human action recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2012, 35(1): 221-231

    2012

  21. [29]

    H-DenseUNet: hybrid densely connected UNet for liver and tumor segmentation from CT volumes[J]

    Li X, Chen H, Qi X, et al. H-DenseUNet: hybrid densely connected UNet for liver and tumor segmentation from CT volumes[J]. IEEE transactions on medical imaging, 2018, 37(12): 2663-2674

    2018

  22. [30]

    Axiomatic derivation of the principle of maximum entropy and the principle of minimum cross-entropy[J]

    Shore J, Johnson R. Axiomatic derivation of the principle of maximum entropy and the principle of minimum cross-entropy[J]. IEEE Transactions on information theory, 1980, 26(1): 26-37

    1980

  23. [31]

    Focal loss for dense object detection[C]

    Lin T Y, Goyal P , Girshick R, et al. Focal loss for dense object detection[C]. Proceedings of the IEEE international conference on computer vision. 2017: 2980-2988

    2017

  24. [32]

    V-net: Fully convolutional neural networks for volumetric medical image segmentation[C]

    Milletari F, Navab N, Ahmadi S A. V-net: Fully convolutional neural networks for volumetric medical image segmentation[C]. 2016 Fourth International Conference on 3D Vision (3DV). IEEE, 2016: 565-571

    2016

  25. [33]

    U-net: Convolutional networks for biomedical image segmentation[C]

    Ronneberger O, Fischer P , Brox T. U-net: Convolutional networks for biomedical image segmentation[C]. International Conference on Medical image computing and computer-assisted intervention. Springer, Cham, 2015: 234-241

    2015

  26. [34]

    Squeeze-and-excitation networks[C]

    Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]. Pro- ceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141

    2018

  27. [35]

    Delving deep into rectifiers: Sur- passing human-level performance on imagenet classification[C]

    He K, Zhang X, Ren S, et al. Delving deep into rectifiers: Sur- passing human-level performance on imagenet classification[C]. Proceedings of the IEEE international conference on computer vision. 2015: 1026-1034

    2015

  28. [36]

    Adam: A method for stochastic optimization[J]

    Kingma D P , Ba J. Adam: A method for stochastic optimization[J]. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  29. [37]

    A comparison of automated lesion segmentation approaches for chronic stroke T1-weighted MRI data[J]

    Ito K L, Kim H, Liew S L. A comparison of automated lesion segmentation approaches for chronic stroke T1-weighted MRI data[J]. bioRxiv, 2018: 441451

    2018

  30. [38]

    Segnet: A deep convo- lutional encoder-decoder architecture for image segmentation[J]

    Badrinarayanan V , Kendall A, Cipolla R. Segnet: A deep convo- lutional encoder-decoder architecture for image segmentation[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(12): 2481-2495

    2017

  31. [39]

    Pyramid scene parsing network[C]

    Zhao H, Shi J, Qi X, et al. Pyramid scene parsing network[C]. Pro- ceedings of the IEEE conference on computer vision and pattern recognition. 2017: 2881-2890

    2017

  32. [40]

    Encoder-decoder with atrous separable convolution for semantic image segmentation[C]

    Chen L C, Zhu Y, Papandreou G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]. Proceedings of the European conference on computer vision (ECCV). 2018: 801-818

    2018

  33. [41]

    Beyond the pixel- wise loss for topology-aware delineation[C]

    Mosinska A, Marquez-Neila P , Koziski M, et al. Beyond the pixel- wise loss for topology-aware delineation[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 3136-3145. Y ongjin ZhouY ongjin Zhou (M’13) received the B.Sc., M.Eng and Ph.D. ...

    2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.