Pith. sign in

REVIEW 5 major objections 7 minor 82 references

LVS-Net: A Lightweight Vessels Segmentation Network for Retinal Image Analysis

T0 review · 5 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read A 0.71-million-parameter network is claimed to out-segment larger retinal vessel models.

desk verdict Plausible lightweight architecture with a useful ablation, but inconsistent numbers and a suspicious test protocol leave the performance claim unverified. read the letter →

arxiv 2412.05968 v1 pith:2UMVNVYF submitted 2024-12-08 eess.IV cs.AIcs.CV

classification eess.IVcs.AIcs.CV
keywords retinalvesselsegmentationlightweightencoder-decoderfocalmodulationattentionspatialfeaturerefinementblockfundusimageanalysisarteryveinDRIVEdatasetSTARE
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper introduces LVS-Net, a lightweight encoder-decoder network for segmenting retinal blood vessels in fundus images, aimed at early disease screening in settings with limited hardware. The authors claim that with only 0.71 million parameters, 2.74 MB of memory, and 29.60 GFLOPs, the model outperforms larger published networks on three standard public datasets, reaching dice scores of 86.44% on DRIVE, 84.22% on CHASE_DB, and 87.88% on STARE. The design combines multi-scale convolutional blocks in the encoder with focal modulation attention and spatial feature refinement at the bottleneck, plus refinement blocks along skip connections and decoder stages. If the reported evaluation is sound, this would establish that high-accuracy vessel segmentation is feasible at a small enough footprint for portable and real-time retinal screening, including artery/vein classification.

What carries the argument

The central mechanism is the combination of the Focal Modulation Attention Module (FMAM) and the Spatial Feature Refinement Block (SFRB). FMAM aggregates context through depth-wise convolutions at multiple levels plus global average pooling, then modulates each query token by gated element-wise multiplication. SFRB is a residual block that concatenates max-pooled and average-pooled features, weights them with a sigmoid attention coefficient from global average pooling, and adds back the input. These blocks are placed at the bottleneck and, for SFRB, at every skip connection and decoder stage, so that multi-scale vessel details survive downsampling and are refined during upsampling.

What would settle it

Re-running the evaluation with a threshold fixed from training data alone and with the standard test splits for STARE and CHASE_DB would settle the claim; if the dice score then drops below the best compared baseline on any dataset, the 'outperforms' conclusion fails.

Watch

Extended reading notes

Core claim

LVS-Net is a lightweight encoder-decoder network that segments retinal vessels and also separates arteries from veins. The encoder uses 1x1 and 3x3 convolutions at three scales, the bottleneck applies Focal Modulation Attention followed by a Spatial Feature Refinement Block, and every decoder upsampling stage and skip connection passes through SFRB before concatenation. The final sigmoid output is thresholded by choosing the F1 threshold that maximizes dice score, and training uses dice loss with Adam. On DRIVE, STARE, and CHASE_DB the authors report accuracy, dice, jaccard, sensitivity, and specificity numbers that exceed those of compared baselines, including U-Net, G-Net Light, Attention U-Net, MultiResNet, BCD-UNet, SegNet, U-Net++, FR-UNet, and RetinaLiteNet, while the model has 0.71M parameters, 2.74 MB memory, and 29.60 GFLOPs. On RITE the model reports average dice 81.34% for arteries/veins/background with higher individual artery and vein dice than listed baselines.

Load-bearing premise

The central claim rests on the evaluation protocol being valid: the dice-maximizing threshold applied after the sigmoid must be selected on validation data rather than test ground truth, and the 80/20 image-level splits used for CHASE_DB and STARE must yield test sets comparable to those used by the published baselines.

Editorial extensions

If this is right

  • If the reported numbers hold, accurate vessel segmentation no longer requires a heavy model: a network under 3 MB can reach dice scores above 84% on all three standard datasets.
  • The same lightweight architecture also separates arteries from veins on RITE, so a single small model can support multi-feature retinal screening.
  • The ablation study's stepwise gains indicate that multiscale convolutions, SFRB at skip connections and bottleneck, and FMAM at the bottleneck each contribute to capturing thin vessels; the final configuration is the sum of these additions.
  • At 29.60 GFLOPs and 0.71M parameters, the model sits between the smallest lightweight baselines and larger U-Net variants, giving clinicians a concrete footprint target for portable screening devices.

Reading between the lines

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

  • A direct next step the paper does not run is cross-dataset evaluation (train on DRIVE, test on STARE and CHASE_DB), which would test whether the reported dice gains survive domain shift between fundus cameras.
  • The bottleneck attention and refinement blocks could transfer to other elongated-structure segmentation tasks, such as coronary angiography or neural fiber tracing, where thin-structure preservation is the limiting factor.
  • Because the reported memory footprint is 2.74 MB before quantization, 8-bit quantization could plausibly bring the deployed model under 1 MB; the paper does not report quantized performance.
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 manuscript proposes LVS-Net, a lightweight encoder-decoder network for retinal vessel segmentation. The encoder uses multi-scale convolutional blocks; the bottleneck combines a Focal Modulation Attention Module (FMAM) and a Spatial Feature Refinement Block (SFRB); the decoder upsamples with transposed convolutions and skip connections that also pass through SFRB. The model is evaluated for vessel segmentation on DRIVE, CHASE_DB, and STARE and for artery/vein segmentation on RITE. The authors report 0.71 million parameters, 2.74 MB memory, and 29.60 GFLOPs, and claim state-of-the-art dice scores of 86.44%, 84.22%, and 87.88% on DRIVE, CHASE_DB, and STARE, respectively.

Significance. If the evaluation protocol is sound and the reported numbers are reproducible, LVS-Net would be a practically useful lightweight baseline for retinal feature segmentation; the 0.71M parameter count is genuinely low, and the ablation study in Table IV gives a useful decomposition of the architectural components. The paper does not ship code, weights, or a precise test-set definition, and the reported metrics contain internal inconsistencies, so the central 'outperforms existing models' claim is not currently supported. The architectural idea is defensible, but the quantitative evidence needs to be corrected and the protocol made explicit before the claims can be assessed.

major comments (5)
  1. [Section III-A] The sentence 'F1-thresholding is used that maximizes the dice score' does not state whether the threshold is chosen on a validation set disjoint from the test images or on the test ground truth itself. If the threshold is optimized on the test labels, the reported dice values are optimistically biased, especially on CHASE_DB (28 images) and STARE (20 images), where a few threshold choices can change the score by several percentage points. The authors must specify the threshold-selection protocol and confirm that test labels were not used for any model or threshold selection.
  2. [Section IV-B and Table I] For CHASE_DB and STARE, the text states only that 80% of images were used for training and 20% for validation, and Table I lists no testing images for these datasets. Without a defined held-out test split, the reported test-set numbers and the comparison to published baselines are ill-defined. The authors must report the exact split, the number of test images, and how the published baselines were evaluated under the same protocol.
  3. [Abstract, Section IV-D, Table II, Conclusion] The reported CHASE_DB dice score appears as 84.22% in the abstract, 84.78% in the body text and Table II, and 82.10% in the conclusion; the STARE dice score appears as 84.78% in Section IV-D and 87.88% in the abstract and Table II. These internal inconsistencies mean the headline quantitative claims are not reliable as stated and must be reconciled with a single, corrected set of results.
  4. [Section IV-C, Eq. (26)] The 'AUC' formula in Eq. (26) is not the standard ROC-AUC; it is dimensionally inconsistent and cannot be interpreted as a probability because it multiplies TP and TN and divides by nested sums. Since Section IV-D uses AUC values (0.993, 0.997, 0.998) as evidence of superiority over other models, the metric must be defined correctly and all AUC values must be recomputed with the standard definition.
  5. [Table III] The RITE results are internally inconsistent: the average accuracy of 98.44% cannot be reconciled with artery accuracy 97.13% and vein accuracy 99.75%, and the average dice of 81.34% cannot be reconciled with artery dice 75.46% and vein dice 71.18% unless the background class is included in the average, which is not stated. The table needs a clear definition of how the 'Average' column is computed and corrected row values.
minor comments (7)
  1. [Section I] The first contribution bullet says 'Introducing LA V-Net' but the model is named LVS-Net elsewhere; this appears to be a typo.
  2. [Section IV-C, Eq. (22)] The dice formula is written with 'TP+TP' in the numerator and denominator; it is mathematically equivalent to 2TP/(2TP+FP+FN) but should be simplified for readability and to avoid confusion.
  3. [Section IV-A] The CHASE_DB dataset is described as 'CHASE DB-DB' in the first sentence of Section IV-A and as 'CHASE DB' elsewhere; the name should be standardized.
  4. [Eqs. (4)-(6) and Fig. 1] Equation (4) uses 'Re' where 'ReLU' is intended, and the notation for convolution (C), transposed convolution (T), and the SFRB/FMAM operations G and F is not consistently defined at first use.
  5. [Table I] The dash in the 'Testing' column for CHASE_DB and STARE is unexplained; if no separate testing split is used, this should be stated explicitly in the table caption or text.
  6. [Section IV-F, Table IV] The ablation row 'MLU + CBAM in Skip Connections' reports a lower dice (80.86%) than the Lightweight U-Net baseline (82.06%), while the text says CBAM 'substantially improves' performance; a brief explanation would help the reader interpret the ablation.
  7. [Availability] No code, trained weights, or public implementation are provided, which, combined with the protocol ambiguities, prevents independent verification of the reported results.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported dice scores are obtained by thresholding that maximizes dice on the evaluation split, making the headline 'outperforms' claim partly fitted; otherwise the empirical benchmark is external and self-contained.

  1. fitted input called prediction [Section III-A (after Eq. 12) with Section IV-B and Table I]
    "To convert the predicted map from the decoder to a segmentation mask, F1-thresholding is used that maximizes the dice score. ... We employed 80% images for model training and 20% validation from each dataset."

    The dice values reported in Table II and the abstract are not produced at a fixed decision rule. The binarization threshold is explicitly chosen to maximize the dice score, and the only split described is 80% training / 20% validation, with Table I listing no testing images for CHASE_DB and STARE. As written, the reported dice is the maximum of dice over thresholds evaluated on the same images whose dice are reported, i.e., a fitted statistic rather than an independent prediction. The 'outperforms existing models' claim compares these threshold-optimized numbers to published baselines, so part of the claimed advantage is an artifact of optimizing a free parameter against the evaluation labels.

full rationale

The paper is an empirical architecture paper rather than a derivation, and most of its evidence is external: performance on public datasets DRIVE, CHASE_DB, STARE, and RITE is an externally falsifiable benchmark, so the bulk of the work is self-contained. The one identified fitted-input step is the thresholding protocol: 'F1-thresholding is used that maximizes the dice score' combined with the 80/20 training/validation description and Table I's absent test splits for CHASE_DB and STARE. Under the protocol as written, the reported dice is a threshold-maximized value on the reporting set, which makes the headline performance comparison partly a fit to the evaluation labels. Other aspects of the central claim, such as 0.71M parameters, 2.74 MB memory, and 29.60 GFLOPs, are architectural and not circular. The paper's heavy citation of the authors' own prior lightweight networks is not load-bearing: LVS-Net is defined and evaluated independently of those works, and no uniqueness theorem or ansatz is imported through self-citation. Internal numerical inconsistencies (CHASE_DB dice 84.22 in the abstract, 84.78 in Table II, 82.10 in the conclusion; STARE dice 84.78 in the text versus 87.88 in Table II) are correctness risks, not circularity.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

LVS-Net is an empirical architecture paper; the ledger lists hand-chosen hyperparameters and evaluation assumptions that the headline numbers depend on. There are no invented physical entities. The most consequential unstated choices are the threshold selection, the non-standard data split, and the unreported class weights.

free parameters (5)
  • Encoder channel widths = 24, 48, 96 channels
    The number of feature channels at each encoder stage is chosen by hand in Eqs. 1-7. This determines capacity and GFLOPs, and no search or derivation is reported.
  • Dropout probability = 0.5
    Dropout after the last encoder block is set to 0.5 in Eq. 7. This is a design choice that can affect test performance.
  • Augmentation rotation angle and contrast factor = 20 degrees; contrast factor not reported
    Section III states augmentation by rotating 20 degrees and adjusting contrast, but the contrast adjustment parameters are not specified.
  • Binary mask threshold = Not reported; selected to maximize dice
    Section III-A applies thresholding that 'maximizes the dice score'. The threshold value and the data it is chosen on are not reported.
  • Dice loss per-class weights w_k = Not reported
    Eq. 13 includes class weights w_k; the values are not specified, which matters for multi-class artery/vein training on RITE.
assumptions (5)
  • domain assumption Public ground-truth annotations in DRIVE, CHASE_DB, STARE, and RITE are reliable training and evaluation labels.
    The benchmark conclusions depend entirely on these manual annotations; the paper does not audit inter-observer variability beyond noting that two graders exist.
  • domain assumption The 80/20 image-level split is representative and comparable to the splits used by baseline methods.
    Section IV-B states this split; for CHASE_DB and STARE it is not the standard published protocol, and exact image identities are not provided.
  • domain assumption The dice-maximizing threshold is not fitted to test labels.
    Section III-A does not specify validation vs test for threshold selection; if test labels are used, all reported dice scores are optimistically biased.
  • standard math The standard formulas for accuracy, dice, Jaccard, sensitivity, specificity, and AUC are correctly implemented.
    The paper states standard definitions, but Eq. 26 for AUC is not the standard ROC-AUC formula, so metric implementation is in doubt.
  • domain assumption The GFLOP and memory estimates are computed with a defined input size and a standard profiling tool.
    Sections I and V report 29.60 GFLOPs and 2.74 MB without providing a calculation method or input resolution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LVS-Net: A Lightweight Vessels Segmentation Network for Retinal Image Analysis." pith.science (2026). https://pith.science/paper/2UMVNVYF

@misc{pith2026241205968,
  author       = {Pith},
  title        = {Pith review of: LVS-Net: A Lightweight Vessels Segmentation Network for Retinal Image Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2UMVNVYF}},
  note         = {Machine review of arXiv:2412.05968}
}
read the original abstract

The analysis of retinal images for the diagnosis of various diseases is one of the emerging areas of research. Recently, the research direction has been inclined towards investigating several changes in retinal blood vessels in subjects with many neurological disorders, including dementia. This research focuses on detecting diseases early by improving the performance of models for segmentation of retinal vessels with fewer parameters, which reduces computational costs and supports faster processing. This paper presents a novel lightweight encoder-decoder model that segments retinal vessels to improve the efficiency of disease detection. It incorporates multi-scale convolutional blocks in the encoder to accurately identify vessels of various sizes and thicknesses. The bottleneck of the model integrates the Focal Modulation Attention and Spatial Feature Refinement Blocks to refine and enhance essential features for efficient segmentation. The decoder upsamples features and integrates them with the corresponding feature in the encoder using skip connections and the spatial feature refinement block at every upsampling stage to enhance feature representation at various scales. The estimated computation complexity of our proposed model is around 29.60 GFLOP with 0.71 million parameters and 2.74 MB of memory size, and it is evaluated using public datasets, that is, DRIVE, CHASE\_DB, and STARE. It outperforms existing models with dice scores of 86.44\%, 84.22\%, and 87.88\%, respectively.

Figures

Figures reproduced from arXiv: 2412.05968 by the authors.

Figure 1
Figure 1. Architecture of the proposed LVS-Net: begins with convolutional operations followed by a decoding path using transposed convolutions. Key elements [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Schematics of the proposed blocks: (a) Focal modulation attention module with context aggregation, (b) Spatial feature refinement block. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 5
Figure 5. In [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figures from the paper (5 more)
Figure 3
Figure 3. Figure 3: Segmentation outcomes of selected test images from the DRIVE dataset. Arranged in a left-to-right sequence are the input images, specifically [PITH_FULL_IMAGE:figures/full_fig_p008_3.png]
Figure 4
Figure 4. Figure 4: Segmentation outcomes of selected test images from the STARE dataset. The images displayed in the following order are: input images (specifically [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Segmentation outcomes of selected test images from the CHASE [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Comparative Analysis of Receiver Operating Characteristic (ROC) Curves for Different Models Evaluated on Three Retinal Datasets: (a) DRIVE, (b) [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Illustration of the visual results obtained by employing different components of the proposed architecture: (a) Input RGB image, (b) Corresponding [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

82 extracted references · 73 canonical work pages

  1. [1]

    Automatic retinal vessel extraction algorithm,

    T. A. Soomro, M. A. Khan, J. Gao, T. M. Khan, M. Paul, and N. Mir, “Automatic retinal vessel extraction algorithm,” in 2016 International Conference on Digital Image Computing: Techniques and Applications (DICTA). IEEE, 2016, pp. 1–8

  2. [2]

    Automatic retinal vessel extraction algorithm based on contrast- sensitive schemes,

    M. A. Khan, T. A. Soomro, T. M. Khan, D. G. Bailey, J. Gao, and N. Mir, “Automatic retinal vessel extraction algorithm based on contrast- sensitive schemes,” in 2016 International conference on image and vision computing New Zealand (IVCNZ) . IEEE, 2016, pp. 1–5

  3. [3]

    Boosting sensitivity of a retinal vessel segmentation algorithm,

    M. A. Khan, T. M. Khan, T. A. Soomro, N. Mir, and J. Gao, “Boosting sensitivity of a retinal vessel segmentation algorithm,” Pattern Analysis and Applications, vol. 22, pp. 583–599, 2019

  4. [4]

    Impact of ica-based image enhancement technique on retinal blood vessels segmentation,

    T. A. Soomro, T. M. Khan, M. A. Khan, J. Gao, M. Paul, and L. Zheng, “Impact of ica-based image enhancement technique on retinal blood vessels segmentation,” IEEE Access, vol. 6, pp. 3524–3538, 2018

  5. [5]

    A generalized multi-scale line-detection method to boost retinal vessel segmentation sensitivity,

    M. A. Khan, T. M. Khan, D. G. Bailey, and T. A. Soomro, “A generalized multi-scale line-detection method to boost retinal vessel segmentation sensitivity,” Pattern Analysis and Applications , vol. 22, pp. 1177–1196, 2019

  6. [6]

    Ggm classifier with multi-scale line detectors for retinal vessel segmentation,

    M. A. Khan, T. M. Khan, S. S. Naqvi, and M. Aurangzeb Khan, “Ggm classifier with multi-scale line detectors for retinal vessel segmentation,” Signal, Image and Video Processing , vol. 13, pp. 1667–1675, 2019

  7. [7]

    An improved retinal vessel segmentation framework using frangi filter coupled with the probabilistic patch based denoiser,

    A. Khawaja, T. M. Khan, K. Naveed, S. S. Naqvi, N. U. Rehman, and S. J. Nawaz, “An improved retinal vessel segmentation framework using frangi filter coupled with the probabilistic patch based denoiser,” IEEE Access, vol. 7, pp. 164 344–164 361, 2019

  8. [8]

    A multi- scale directional line detector for retinal vessel segmentation,

    A. Khawaja, T. M. Khan, M. A. Khan, and J. Nawaz, “A multi- scale directional line detector for retinal vessel segmentation,” Sensors, vol. 19, no. 22, 2019

Show all 82 references
  1. [9]

    The use of fourier phase symmetry for thin vessel detection in retinal fundus images,

    M. A. Khan, T. M. Khan, K. I. Aziz, S. S. Ahmad, N. Mir, and E. El- bakush, “The use of fourier phase symmetry for thin vessel detection in retinal fundus images,” in 2019 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) . IEEE, 2019, pp. 1–6

  2. [10]

    Shallow vessel segmentation network for automatic retinal vessel seg- mentation,

    T. M. Khan, F. Abdullah, S. S. Naqvi, M. Arsalan, and M. A. Khan, “Shallow vessel segmentation network for automatic retinal vessel seg- mentation,” in 2020 International Joint Conference on Neural Networks (IJCNN). IEEE, 2020, pp. 1–7

  3. [11]

    Exploiting residual edge information in deep fully convolutional neural networks for retinal vessel segmentation,

    T. M. Khan, S. S. Naqvi, M. Arsalan, M. A. Khan, H. A. Khan, and A. Haider, “Exploiting residual edge information in deep fully convolutional neural networks for retinal vessel segmentation,” in 2020 International Joint Conference on Neural Networks (IJCNN) . IEEE, 2020, pp. 1–8

  4. [12]

    A semantically flexible feature fusion network for retinal vessel segmentation,

    T. M. Khan, A. Robles-Kelly, and S. S. Naqvi, “A semantically flexible feature fusion network for retinal vessel segmentation,” in International Conference on Neural Information Processing . Springer, Cham, 2020, pp. 159–167

  5. [13]

    Towards automated eye diagnosis: an improved retinal vessel segmentation framework using ensemble block matching 3d filter,

    K. Naveed, F. Abdullah, H. A. Madni, M. A. Khan, T. M. Khan, and S. S. Naqvi, “Towards automated eye diagnosis: an improved retinal vessel segmentation framework using ensemble block matching 3d filter,” Diagnostics, vol. 11, no. 1, p. 114, 2021

  6. [14]

    A review on glaucoma disease detection using computerized techniques,

    F. Abdullah, R. Imtiaz, H. A. Madni, H. A. Khan, T. M. Khan, M. A. Khan, and S. S. Naqvi, “A review on glaucoma disease detection using computerized techniques,” IEEE Access, vol. 9, pp. 37 311–37 333, 2021

  7. [15]

    Screening of glaucoma disease from retinal vessel images using se- mantic segmentation,

    R. Imtiaz, T. M. Khan, S. S. Naqvi, M. Arsalan, and S. J. Nawaz, “Screening of glaucoma disease from retinal vessel images using se- mantic segmentation,” Computers & Electrical Engineering , vol. 91, p. 107036, 2021

  8. [16]

    Retinal capillary microvessel morphology changes are associated with vascular damage and dysfunction in cerebral small vessel disease,

    S. J. Wiseman, J.-F. Zhang, C. Gray, C. Hamid, M. d. C. Vald´es Hern ´andez, L. Ballerini, M. J. Thrippleton, C. Manning, M. Stringer, E. Sleight et al., “Retinal capillary microvessel morphology changes are associated with vascular damage and dysfunction in cerebral small ves...

  9. [17]

    Alzheimer’s retinopathy: seeing disease in the eyes,

    N. Mirzaei, H. Shi, M. Oviatt, J. Doustar, A. Rentsendorj, D.-T. Fuchs, J. Sheyn, K. L. Black, Y . Koronyo, and M. Koronyo-Hamaoui, “Alzheimer’s retinopathy: seeing disease in the eyes,” Frontiers in neuroscience, vol. 14, p. 921, 2020

  10. [18]

    Residual multiscale full convolutional network (rm-fcn) for high resolution se- mantic segmentation of retinal vasculature,

    T. M. Khan, A. Robles-Kelly, S. S. Naqvi, and A. Muhammad, “Residual multiscale full convolutional network (rm-fcn) for high resolution se- mantic segmentation of retinal vasculature,” in Structural, Syntactic, and Statistical Pattern Recognition: Joint IAPR International Work...

  11. [19]

    Width-wise vessel bifurcation for improved retinal vessel segmentation,

    T. M. Khan, M. A. Khan, N. U. Rehman, K. Naveed, I. U. Afridi, S. S. Naqvi, and I. Raazak, “Width-wise vessel bifurcation for improved retinal vessel segmentation,” Biomedical Signal Processing and Control, vol. 71, p. 103169, 2022

  12. [20]

    Glan: Gan assisted lightweight attention network for biomedical imaging based diagnostics,

    S. S. Naqvi, Z. A. Langah, H. A. Khan, M. I. Khan, T. Bashir, M. I. Razzak, and T. M. Khan, “Glan: Gan assisted lightweight attention network for biomedical imaging based diagnostics,” Cognitive Compu- tation, vol. 15, no. 3, pp. 932–942, 2023

  13. [21]

    Retinal vessel segmentation via a multi-resolution contextual network and adversarial learning,

    T. M. Khan, S. S. Naqvi, A. Robles-Kelly, and I. Razzak, “Retinal vessel segmentation via a multi-resolution contextual network and adversarial learning,” Neural Networks, vol. 165, pp. 310–320, 2023

  14. [22]

    Lmbf- net: A lightweight multipath bidirectional focal attention network for multifeatures segmentation,

    T. M. Khan, S. Iqbal, S. S. Naqvi, I. Razzak, and E. Meijering, “Lmbf- net: A lightweight multipath bidirectional focal attention network for multifeatures segmentation,” in 2024 IEEE International Conference on Image Processing (ICIP) . IEEE, 2024, pp. 2807–2813

  15. [23]

    Tesl- net: A transformer-enhanced cnn for accurate skin lesion segmentation,

    S. Iqbal, M. Zeeshan, M. Mehmood, T. M. Khan, and I. Razzak, “Tesl- net: A transformer-enhanced cnn for accurate skin lesion segmentation,” arXiv preprint arXiv:2408.09687 , 2024

  16. [24]

    Euis-net: A convolutional neural network for efficient ultrasound image segmentation,

    S. Iqbal, H. Ahmed, M. Sharif, M. Hena, T. M. Khan, and I. Razzak, “Euis-net: A convolutional neural network for efficient ultrasound image segmentation,” arXiv preprint arXiv:2408.12323 , 2024

  17. [25]

    Tbconvl-net: A hybrid deep learning architecture for robust medical image segmentation,

    S. Iqbal, T. M. Khan, S. S. Naqvi, A. Naveed, and E. Meijering, “Tbconvl-net: A hybrid deep learning architecture for robust medical image segmentation,” Pattern Recognition, vol. 158, p. 111028, 2025

  18. [26]

    Ad-net: Attention-based dilated convolutional residual network with guided decoder for robust skin lesion segmentation,

    A. Naveed, S. S. Naqvi, T. M. Khan, S. Iqbal, M. Y . Wani, and H. A. Khan, “Ad-net: Attention-based dilated convolutional residual network with guided decoder for robust skin lesion segmentation,” Neural Computing and Applications , pp. 1–23, 2024

  19. [27]

    Lssf-net: Lightweight segmentation with self-awareness, spatial atten- tion, and focal modulation,

    H. Farooq, Z. Zafar, A. Saadat, T. M. Khan, S. Iqbal, and I. Razzak, “Lssf-net: Lightweight segmentation with self-awareness, spatial atten- tion, and focal modulation,” Artificial Intelligence in Medicine, vol. 158, 2024

  20. [28]

    Advancing metaverse- based healthcare with multimodal neuroimaging fusion via multi-task adversarial variational autoencoder for brain age estimation,

    U. Muhammad, R. Azka, S. Abdullah, R. Abd Ur, G. Sung-Min, L. Aleum, K. Tariq M., and R. Imran, “Advancing metaverse- based healthcare with multimodal neuroimaging fusion via multi-task adversarial variational autoencoder for brain age estimation,” IEEE Journal of Biomedical a...

  21. [29]

    Imaging the retinal vasculature,

    S. A. Burns, A. E. Elsner, and T. J. Gast, “Imaging the retinal vasculature,” Annual review of vision science, vol. 7, pp. 129–153, 2021

  22. [30]

    Robust retinal blood vessel segmentation using a patch-based statistical adaptive multi-scale line detector,

    S. Iqbal, K. Naveed, S. S. Naqvi, A. Naveed, and T. M. Khan, “Robust retinal blood vessel segmentation using a patch-based statistical adaptive multi-scale line detector,”Digital Signal Processing, vol. 139, p. 104075, 2023

  23. [31]

    Semantic segmentation of retinal exudates using a residual encoder–decoder architecture in diabetic retinopathy,

    M. A. Manan, F. Jinchao, T. M. Khan, M. Yaqub, S. Ahmed, and I. s. Chuhan, “Semantic segmentation of retinal exudates using a residual encoder–decoder architecture in diabetic retinopathy,” Microscopy Re- search and Technique, 2023

  24. [32]

    Retinalitenet: A lightweight transformer based cnn for retinal feature segmentation,

    M. Mehmood, M. Alsharari, S. Iqbal, I. Spence, and M. Fahim, “Retinalitenet: A lightweight transformer based cnn for retinal feature segmentation,” in Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR) Workshops , June 2024, pp. 2454–2463

  25. [33]

    Feature enhancer segmentation network (fes-net) for vessel segmentation,

    T. M. Khan, M. Arsalan, S. Iqbal, I. Razzak, and E. Meijering, “Feature enhancer segmentation network (fes-net) for vessel segmentation,” in 2023 International Conference on Digital Image Computing: Techniques and Applications (DICTA) . IEEE, 2023, pp. 160–167

  26. [34]

    Lmbis-net: A lightweight multipath bidirectional skip con- nection based cnn for retinal blood vessel segmentation,

    M. M. Abbasi, S. Iqbal, A. Naveed, T. M. Khan, S. S. Naqvi, and W. Khalid, “Lmbis-net: A lightweight multipath bidirectional skip con- nection based cnn for retinal blood vessel segmentation,” arXiv preprint arXiv:2309.04968, 2023

  27. [35]

    Ldmres-net: A lightweight neural network for efficient medical image segmentation on iot and edge devices,

    S. Iqbal, T. M. Khan, S. S. Naqvi, A. Naveed, M. Usman, H. A. Khan, and I. Razzak, “Ldmres-net: A lightweight neural network for efficient medical image segmentation on iot and edge devices,” IEEE journal of biomedical and health informatics , 2023

  28. [36]

    Vessel intensity profile uniformity improvement for retinal vessel seg- mentation,

    M. Mehmood, T. M. Khan, M. A. Khan, S. S. Naqvi, and W. Alhalabi, “Vessel intensity profile uniformity improvement for retinal vessel seg- mentation,” Procedia Computer Science , vol. 163, pp. 370–380, 2019

  29. [37]

    Lmbis-net: A lightweight bidirectional skip connection based multipath cnn for retinal blood vessel segmentation,

    M. Matloob Abbasi, S. Iqbal, K. Aurangzeb, M. Alhussein, and T. M. Khan, “Lmbis-net: A lightweight bidirectional skip connection based multipath cnn for retinal blood vessel segmentation,” Scientific Reports, vol. 14, no. 1, p. 15219, 2024

  30. [38]

    Region guided attention network for retinal vessel segmentation,

    S. Javed, T. M. Khan, A. Qayyum, A. Sowmya, and I. Razzak, “Region guided attention network for retinal vessel segmentation,” arXiv preprint arXiv:2407.18970, 2024. 12

  31. [39]

    Recent trends and advances in fundus image analysis: A review,

    S. Iqbal, T. M. Khan, K. Naveed, S. S. Naqvi, and S. J. Nawaz, “Recent trends and advances in fundus image analysis: A review,” Computers in Biology and Medicine , vol. 151, p. 106277, 2022

  32. [40]

    Rc-net: A convolutional neural network for retinal vessel segmentation,

    T. M. Khan, A. Robles-Kelly, and S. S. Naqvi, “Rc-net: A convolutional neural network for retinal vessel segmentation,” in 2021 Digital Image Computing: Techniques and Applications (DICTA) . IEEE, 2021, pp. 01–07

  33. [41]

    Leveraging image complexity in macro-level neural network design for medical image segmentation,

    T. M. Khan, S. S. Naqvi, and E. Meijering, “Leveraging image complexity in macro-level neural network design for medical image segmentation,” Scientific Reports, vol. 12, no. 1, p. 22286, 2022

  34. [42]

    T-net: A resource- constrained tiny convolutional neural network for medical image seg- mentation,

    T. M. Khan, A. Robles-Kelly, and S. S. Naqvi, “T-net: A resource- constrained tiny convolutional neural network for medical image seg- mentation,” in Proceedings of the IEEE/CVF winter conference on applications of computer vision , 2022, pp. 644–653

  35. [43]

    G-net light: A lightweight modified google net for retinal vessel segmentation,

    S. Iqbal, S. Naqvi, H. Ahmed, A. Saadat, and T. M. Khan, “G-net light: A lightweight modified google net for retinal vessel segmentation,” in Photonics, vol. 9, no. 12. MDPI, 2022, pp. 923–936

  36. [44]

    Prompt deep light-weight vessel segmentation network (plvs-net),

    M. Arsalan, T. M. Khan, S. S. Naqvi, M. Nawaz, and I. Razzak, “Prompt deep light-weight vessel segmentation network (plvs-net),” IEEE/ACM Transactions on Computational Biology and Bioinformatics , vol. 20, no. 2, pp. 1363–1371, 2022

  37. [45]

    Mkis-net: a light-weight multi-kernel network for medical image segmentation,

    T. M. Khan, M. Arsalan, A. Robles-Kelly, and E. Meijering, “Mkis-net: a light-weight multi-kernel network for medical image segmentation,” in International Conference on Digital Image Computing: Techniques and Applications (DICTA) . 10.1109/DICTA56598.2022.10034573, 2022, pp. 1–8

  38. [46]

    Focal modulation networks,

    J. Yang, C. Li, X. Dai, and J. Gao, “Focal modulation networks,” Advances in Neural Information Processing Systems , vol. 35, pp. 4203– 4217, 2022

  39. [47]

    Spatial attention for multi-scale feature refinement for object detection,

    H. Wang, Z. Wang, M. Jia, A. Li, T. Feng, W. Zhang, and L. Jiao, “Spatial attention for multi-scale feature refinement for object detection,” in Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops, 2019, pp. 0–0

  40. [48]

    A manually- labeled, artery/vein classified benchmark for the drive dataset,

    T. A. Qureshi, M. Habib, A. Hunter, and B. Al-Diri, “A manually- labeled, artery/vein classified benchmark for the drive dataset,” in Proceedings of the 26th IEEE international symposium on computer- based medical systems . IEEE, 2013, pp. 485–488

  41. [49]

    Robust retinal vessel segmentation via locally adaptive derivative frames in orientation scores,

    J. Zhang, B. Dashtbozorg, E. Bekkers, J. P. W. Pluim, R. Duits, and B. M. ter Haar Romeny, “Robust retinal vessel segmentation via locally adaptive derivative frames in orientation scores,” IEEE Transactions on Medical Imaging, vol. 35, no. 12, pp. 2631–2644, Dec 2016

  42. [50]

    Locating blood vessels in retinal images by piecewise threshold probing of a matched filter response,

    A. Hoover, V . Kouznetsova, and M. Goldbaum, “Locating blood vessels in retinal images by piecewise threshold probing of a matched filter response,” IEEE Transactions Medical Imaging, vol. 19, no. 3, pp. 203– 210, 2000

  43. [51]

    Msganet-rav: A multiscale guided attention network for artery-vein segmentation and classification from optic disc and retinal images,

    A. E. Chowdhury, G. Mann, W. H. Morgan, A. Vukmirovic, A. Mehnert, and F. Sohel, “Msganet-rav: A multiscale guided attention network for artery-vein segmentation and classification from optic disc and retinal images,” Journal of optometry , vol. 15, pp. S58–S69, 2022

  44. [52]

    Artery–vein segmentation in fundus images using a fully convolutional network,

    R. Hemelings, B. Elen, I. Stalmans, K. Van Keer, P. De Boever, and M. B. Blaschko, “Artery–vein segmentation in fundus images using a fully convolutional network,” Computerized Medical Imaging and Graphics, vol. 76, p. 101636, 2019

  45. [53]

    Fractal dimension of retinal vasculature as an image quality metric for automated fundus image analysis systems,

    X. Lyu, P. Jajal, M. Z. Tahir, and S. Zhang, “Fractal dimension of retinal vasculature as an image quality metric for automated fundus image analysis systems,” Scientific Reports, vol. 12, p. 11868, 2022

  46. [54]

    Dual-channel asymmetric convolutional neural network for an efficient retinal blood vessel segmentation in eye fundus images,

    Y . Xu and Y . Fan, “Dual-channel asymmetric convolutional neural network for an efficient retinal blood vessel segmentation in eye fundus images,” Biocybernetics and Biomedical Engineering, vol. 42, no. 2, pp. 695–706, 2022

  47. [55]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” arXiv:2010.11929, 2020

  48. [56]

    SegViT: Semantic segmentation with plain vision transformers,

    B. Zhang, Z. Tian, Q. Tang, X. Chu, X. Wei, C. Shen, and Y . Liu, “SegViT: Semantic segmentation with plain vision transformers,” arXiv:2210.05844, 2022

  49. [57]

    Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers,

    S. Zheng, J. Lu, H. Zhao, X. Zhu, Z. Luo, Y . Wang, Y . Fu, J. Feng, T. Xi- ang, P. H. Torr, and L. Zhang, “Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 20...

  50. [58]

    Segmenter: Trans- former for semantic segmentation,

    R. Strudel, R. Garcia, I. Laptev, and C. Schmid, “Segmenter: Trans- former for semantic segmentation,” in IEEE/CVF International Confer- ence on Computer Vision (ICCV) , 2021, pp. 7262–7272

  51. [59]

    Vision transformers for dense prediction,

    R. Ranftl, A. Bochkovskiy, and V . Koltun, “Vision transformers for dense prediction,” in IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 12 179–12 188

  52. [60]

    Full-resolution network and dual-threshold iteration for retinal vessel and coronary angiograph segmentation,

    W. Liu, H. Yang, T. Tian, Z. Cao, X. Pan, W. Xu, Y . Jin, and F. Gao, “Full-resolution network and dual-threshold iteration for retinal vessel and coronary angiograph segmentation,” IEEE journal of biomedical and health informatics, vol. 26, no. 9, pp. 4623–4634, 2022

  53. [61]

    DENSE-INception U-net for medical image segmentation,

    Z. Zhang, C. Wu, S. Coleman, and D. Kerr, “DENSE-INception U-net for medical image segmentation,” Computer Methods and Programs in Biomedicine, vol. 192, p. 105395, 2020

  54. [62]

    nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation,

    F. Isensee, P. F. Jaeger, S. A. Kohl, J. Petersen, and K. H. Maier-Hein, “nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation,” Nature Methods, vol. 18, no. 2, pp. 203–211, 2021

  55. [63]

    Lightweight V-Net for liver segmentation,

    T. Lei, W. Zhou, Y . Zhang, R. Wang, H. Meng, and A. K. Nandi, “Lightweight V-Net for liver segmentation,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 1379–1383

  56. [64]

    Lightweight U-Nets for brain tumor segmentation,

    T. Tarasiewicz, M. Kawulok, and J. Nalepa, “Lightweight U-Nets for brain tumor segmentation,” in International MICCAI Brain Lesion Workshop, 2020, pp. 3–14

  57. [65]

    PyConvU-Net: A lightweight and multiscale network for biomedical image segmentation,

    C. Li, Y . Fan, and X. Cai, “PyConvU-Net: A lightweight and multiscale network for biomedical image segmentation,” BMC Bioinformatics , vol. 22, no. 1, pp. 1–11, 2021

  58. [66]

    MobileNets: Efficient convolutional neural networks for mobile vision applications,

    A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto, and H. Adam, “MobileNets: Efficient convolutional neural networks for mobile vision applications,” arXiv:1704.04861, 2017

  59. [67]

    Simultaneous segmentation and classification of the retinal arteries and veins from color fundus images,

    J. Morano, ´A. S. Hervella, J. Novo, and J. Rouco, “Simultaneous segmentation and classification of the retinal arteries and veins from color fundus images,” Artificial Intelligence in Medicine , vol. 118, p. 102116, 2021

  60. [68]

    One-shot retinal artery and vein segmentation via cross-modality pretraining,

    D. Shi, S. He, J. Yang, Y . Zheng, and M. He, “One-shot retinal artery and vein segmentation via cross-modality pretraining,” Ophthalmology Science, vol. 4, no. 2, p. 100363, 2024

  61. [69]

    Lunet: deep learning for the segmentation of arterioles and venules in high resolution fundus images,

    J. Fhima, J. Van Eijgen, H. Kulenovic, V . Debeuf, M. Vangilbergen, M.-I. Billen, H. Brackenier, M. Freiman, I. Stalmans, and J. A. Behar, “Lunet: deep learning for the segmentation of arterioles and venules in high resolution fundus images,” Physiological Measurement, 2024

  62. [70]

    Multi- scale interactive network with artery/vein discriminator for retinal vessel classification,

    J. Hu, H. Wang, G. Wu, Z. Cao, L. Mou, Y . Zhao, and J. Zhang, “Multi- scale interactive network with artery/vein discriminator for retinal vessel classification,” IEEE Journal of Biomedical and Health Informatics , vol. 26, no. 8, pp. 3896–3905, 2022

  63. [71]

    Multi- task neural networks with spatial activation for retinal vessel seg- mentation and artery/vein classification,

    W. Ma, S. Yu, K. Ma, J. Wang, X. Ding, and Y . Zheng, “Multi- task neural networks with spatial activation for retinal vessel seg- mentation and artery/vein classification,” in Medical Image Computing and Computer Assisted Intervention–MICCAI 2019: 22nd International Conferenc...

  64. [72]

    Lightweight attention convolutional neural network for retinal vessel image segmentation,

    X. Li, Y . Jiang, M. Li, and S. Yin, “Lightweight attention convolutional neural network for retinal vessel image segmentation,” IEEE Transac- tions on Industrial Informatics , vol. 17, no. 3, pp. 1958–1967, 2020

  65. [73]

    Retinal vessel segmentation using deep learning: a review,

    C. Chen, J. H. Chuah, R. Ali, and Y . Wang, “Retinal vessel segmentation using deep learning: a review,” IEEE Access , vol. 9, pp. 111 985– 112 004, 2021

  66. [74]

    Automated separation of binary overlapping trees in low-contrast color retinal images,

    Q. Hu, M. D. Abr `amoff, and M. K. Garvin, “Automated separation of binary overlapping trees in low-contrast color retinal images,” inMedical Image Computing and Computer-Assisted Intervention–MICCAI 2013: 16th International Conference, Nagoya, Japan, September 22-26, 2013, Pr...

  67. [75]

    U-Net: Convolutional net- works for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional net- works for biomedical image segmentation,” in International Confer- ence on Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2015, pp. 234–241

  68. [76]

    Atten- tion u-net: Learning where to look for the pancreas,

    O. Oktay, J. Schlemper, L. L. Folgoc, M. Lee, M. Heinrich, K. Misawa, K. Mori, S. McDonagh, N. Y . Hammerla, B. Kainz et al. , “Atten- tion u-net: Learning where to look for the pancreas,” arXiv preprint arXiv:1804.03999, 2018

  69. [77]

    Multiresunet: Rethinking the u-net architecture for multimodal biomedical image segmentation,

    N. Ibtehaz and M. S. Rahman, “Multiresunet: Rethinking the u-net architecture for multimodal biomedical image segmentation,” Neural networks, vol. 121, pp. 74–87, 2020

  70. [78]

    Bi- directional convlstm u-net with densley connected convolutions,

    R. Azad, M. Asadi-Aghbolaghi, M. Fathy, and S. Escalera, “Bi- directional convlstm u-net with densley connected convolutions,” in Proceedings of the IEEE/CVF international conference on computer vision workshops, 2019, pp. 0–0

  71. [79]

    Segnet: A deep con- volutional encoder-decoder architecture for image segmentation,

    V . Badrinarayanan, A. Kendall, and R. Cipolla, “Segnet: A deep con- volutional encoder-decoder architecture for image segmentation,” IEEE transactions on pattern analysis and machine intelligence , vol. 39, no. 12, pp. 2481–2495, 2017. 13

  72. [80]

    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 International Workshops on Deep Learning in Medical Image Analysis (DLMIA) and Multimodal Learning for Clinical Decision Support (ML- CDS) Held...

  73. [81]

    Retinalitenet: A lightweight transformer based cnn for retinal feature segmentation,

    M. Mehmood, M. Alsharari, S. Iqbal, I. Spence, and M. Fahim, “Retinalitenet: A lightweight transformer based cnn for retinal feature segmentation,” in Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , 2024, pp. 2454–2463

  74. [82]

    An ensemble classification-based ap- proach applied to retinal blood vessel segmentation,

    M. M. Fraz, P. Remagnino, A. Hoppe, B. Uyyanonvara, A. R. Rudnicka, C. G. Owen, and S. A. Barman, “An ensemble classification-based ap- proach applied to retinal blood vessel segmentation,” IEEE Transactions on Biomedical Engineering , vol. 59, no. 9, pp. 2538–2548, 2012

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.