Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

High-Throughput Low-Cost Segmentation of Brightfield Microscopy Live Cell Images

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a compute-light ensemble of U-Net variants, trained mostly on bright-field images, reaches a mean F1 of 0.89 on the phase-contrast LIVECell benchmark and beats generalist cell-segmentation tools on bright-field frames.

desk verdict The independent LIVECell validation is real; the brightfield "outperforms all competitors" claim rests on a 10-image comparison that isn't secured. read the letter →

arxiv 2508.14106 v2 pith:FSPTDFKV submitted 2025-08-17 q-bio.QM cs.AIcs.CVeess.IV

Surajit Das , Gourav Roy , Pavel Zun This is my paper
classification q-bio.QMcs.AIcs.CVeess.IV
keywords bright-fieldmicroscopylivecellsegmentationU-Netensemblefrozenencodersphase-contrastgeneralizationLIVECellbenchmarklow-computedeeplearninglabel-freeimaging
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 sets out to show that segmenting unstained live cells in bright-field microscopy does not require heavy generalist models, fluorescent labels, or high-end GPUs. It claims a U-Net pipeline—frozen pretrained encoders, attention-gated decoders, composite losses, hard-example retraining, and majority-vote ensembling—reaches a mean F1 of 0.89 on the phase-contrast LIVECell benchmark while trained predominantly on in-house bright-field images. If true, routine bright-field quantification could be automated in labs with modest compute, and the same model could serve multiple microscopy modalities.

What carries the argument

The load-bearing mechanism is a majority-vote ensemble of U-Net variants that share frozen pretrained encoders and attention-gated decoders. The paper reports that the ensemble beats its best individual model on recall, compensating for single-model failure modes; the models are trained with a composite focal–Dice–boundary loss plus hard-example retraining, and 14 checkpoints are selected by validation F1 for the final vote.

What would settle it

Have a second expert independently redraw masks on a random subset of the same 857 images and compute inter-annotator agreement; then rerun the 10-image comparison with images selected by a blinded protocol. If expert disagreement approaches the reported performance gap over the strongest competitor, the advantage is mostly label noise; if agreement is high and the ranking holds, the claim stands.

Watch

Extended reading notes

Core claim

The authors claim that a U-Net-style pipeline built from frozen pretrained encoders, attention-gated decoders, composite losses, hard-example retraining, and a 14-model majority-vote ensemble can segment unstained live cells in bright-field microscopy at a level competitive with or better than generalist tools. Trained on 857 in-house images (697 bright-field, 160 phase-contrast), the pipeline reports mean Dice/F1 of 0.89 ± 0.07, IoU 0.81 ± 0.10, and pixel accuracy 0.93 ± 0.06 on 3,180 images of the phase-contrast LIVECell benchmark, and reports consistently higher Dice than four state-of-the-art competitors on 10 bright-field test images. The authors interpret this as evidence that the mode

Load-bearing premise

The headline numbers depend on the manual ground-truth masks being accurate and on the 10 bright-field comparison images being representative; if labels are noisy or those 10 images are unrepresentative, the reported F1 and the stated superiority over competitors would shrink.

Editorial extensions

If this is right

  • If the 0.89 mean F1 reproduces, manual masking for bright-field live-cell quantification can be replaced in high-throughput screening, at roughly 3–4 s per 704×520 image.
  • The reported transfer from bright-field-dominated training to phase-contrast implies a single retrainable model could cover several microscope modalities in one lab.
  • At about 6.5 hours of training on a 16 GB T4 GPU and sub-$5k workstations, the recipe lowers the practical cost floor for cell segmentation in teaching and resource-limited settings.
  • The high recall (0.96) and lower precision (0.84) profile makes the model suitable for detection-sensitive experiments, while boundary outliers captured by Hausdorff distance mean it is not yet a turnkey edge-exact tool.

Reading between the lines

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

  • Because the training set already contained 20% phase-contrast images, the modality-transfer claim is not a clean test; training with 0% phase-contrast exposure would isolate whether robustness comes from the mixture or from the bright-field augmentation alone.
  • The same ensemble recipe could plausibly carry over to other label-free modalities such as DIC or low-cost holographic microscopy, which share the low-contrast, noisy-image problem; that extension is implied but not tested in the paper.
  • The 14-model ensemble is the main speed bottleneck, so a distilled single model or shared-backbone inference would be a natural route toward real-time (under 1 s per image) segmentation.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The manuscript proposes a low-cost CNN ensemble for segmenting unstained live cells in bright-field microscopy. The pipeline combines frozen ImageNet encoders (DenseNet121, VGG16) in U-Net decoders, attention gates, composite losses, hard-example mining, progressive dropout, and majority-voting ensembles. On the authors' own bright-field/phase-contrast dataset (857 images, 80:20 train/validation split), an ensemble of 14 models is claimed to outperform Cellpose-SAM, Cellpose3, StarDist, and an SSL baseline on 10 bright-field images (Table 2). On the public LIVECell dataset, the model reports mean F1 = 0.89 ± 0.07, IoU = 0.81, accuracy = 0.93. The paper concludes that the pipeline outperforms existing methods for bright-field segmentation and is practical for resource-limited deployment.

Significance. If the LIVECell result is robust, it is a genuinely useful external validation and supports a practical claim of cross-modality generalization from a model trained with only ~20% phase-contrast data. The low-compute training (Google Colab, 6.5 h) and explicit compute reporting are valuable for the intended users. However, the central scientific claim of consistent superiority over existing methods is currently supported only by a 10-image comparison whose selection protocol, test-set separation, and baseline fairness are not established. The internal ablation statistics also contain numerical inconsistencies. Thus the contribution's significance as a method-comparison paper is not yet secured, although the engineering contribution and the LIVECell generalization result are meaningful.

major comments (4)
  1. [Sections 3.3.3, 3.5, 4.1/Table 2] The claim of consistent outperformance over all competitors rests on 10 images with no stated selection protocol and no separate test set. The paper defines only an 80:20 train/validation split, and Section 3.5 says the 14 ensemble members were 'selected for voting based on the Best Validation Metrics & Stable Performance.' If the 10 images in Table 2 come from that same validation pool, the comparison is on the model-selection set and is optimistically biased. Please specify how the 10 images were chosen, whether they overlap with model/ensemble selection, and provide a hold-out bright-field test set with pre-specified evaluation.
  2. [Section 4.1/Table 2] The comparison is unfair as stated. The proposed pipeline is trained and tuned on the authors' bright-field/phase-contrast dataset, while the baselines (CellPose-SAM, Cellpose3, StarDist, SSL) are used only as off-the-shelf pretrained models with no fine-tuning or hyperparameter optimization on that data. This may support a practical claim that off-the-shelf generalists fail on this domain, but it does not support the scientific claim of algorithmic superiority. Please fine-tune the baselines with comparable effort or explicitly restrict the claim to 'off-the-shelf baselines.'
  3. [Section 4.5/Table 6] The one-sample t-test on per-image F1 scores is inappropriate. F1 scores are bounded, non-Gaussian, and clustered by cell line and experimental batch; the observations are not i.i.d. The test only compares to an arbitrary threshold of 0.75, not to any competitor, so it does not support the 'outperforms' conclusion. The astronomical t value and p < 10^-280 are artifacts of the large sample size. Use clustered bootstrap or non-parametric tests and a pre-specified comparison against competing methods.
  4. [Section 4.3/Table 2] The ablation numbers are internally inconsistent. Table 2's 'OurModel' entries average to F1 ≈ 0.776 and recall ≈ 0.840, which matches the ensemble recall of 0.8400 in Section 4.3. Yet Section 4.3 states that the ensemble differs from Model-3 (F1 = 0.7964) by only -0.012, implying ensemble F1 ≈ 0.784, and it calls the ensemble 'superior' despite lower F1. Also, Section 3.4.1 says VGG16 is the most robust backbone while Table 1/Figure 4 show DenseNet121 with the highest validation F1. Please reconcile these numbers and clarify which model produced Table 2.
minor comments (5)
  1. [Abstract / footnote 1] The abstract states 'The code and dataset are available for reproducibility,' but footnote 1 says they 'may be provided upon reasonable request.' No repository or dataset identifier is given. Please make the availability statement concrete.
  2. [Table 1 / general text] The column header 'F eaturesKey Feat.' is malformed. Also, 'CV AT' should be 'CVAT', 'Nature Communication' should be 'Nature Communications', and 'brought out by' is informal.
  3. [Table 5 / Section 4.2] The text reports F1 std as ±0.08, while Table 5(a) lists 0.07 for F1 (and Table 5(b) lists 0.07). Please reconcile.
  4. [Section 4.2] Eight LIVECell images are listed as 'corrupted' and excluded, but no criterion is given. State why these images are invalid and whether results are robust to their inclusion.
  5. [References] References [42] and [47] both cite the same Cellpose-SAM preprint with different author lists. Please consolidate.

Circularity Check

1 steps flagged · score 6.0 of 10

Bright-field 'outperforms all competitors' claim is evaluated on the same validation set used to select the ensemble; LIVECell result is independent.

  1. fitted input called prediction [Sections 3.3.3, 3.5, 4.1 (Table 2)]
    "The dataset is split into two subsets namely, Train and Validation with the proportion 80:20. ... Finally 14 classifiers were selected for voting based on the Best Validation Metrics & Stable Performance. ... Table 2 compares the segmentation performance of the proposed model against four state-of-the-art approaches ... across 10 diverse bright-field microscopy images. Our model consistently outperformed all competitors."

    The paper defines no separate test set; the only split is 80:20 Train/Validation. The ensemble is explicitly selected using 'Best Validation Metrics.' The 10-image comparison in Table 2, which supports the claim to 'consistently outperformed all competitors' on bright-field data, therefore evaluates the model on the same validation metrics used to choose the 14 classifiers. The reported F1/Dice on these images is the selection criterion, not an independent prediction; the bright-field outperformance claim reduces to the validation fit. The LIVECell evaluation (Sec. 4.2) is independent and not circular, so circularity is partial.

full rationale

The paper's central LIVECell result (mean F1 0.89 ± 0.07 on a public dataset) is an external, independent check and does not reduce to the model's training or selection inputs; that part of the derivation is self-contained. However, the bright-field 'outperforms all competitors' claim rests entirely on Table 2, which uses 10 images from the authors' own dataset. Since Section 3.3.3 defines only Train and Validation, and Section 3.5 selects ensemble members on validation metrics, the Table 2 comparison is effectively a report of the selection criterion. The claimed superiority over SOTA on bright-field is therefore statistically forced by the ensemble-selection procedure, not an unbiased prediction. The absence of any self-citation or imported uniqueness theorem keeps this from being a full 8-10; the independent LIVECell validation prevents complete circularity. The score of 6 reflects that one of the paper's headline claims reduces by construction to its own validation fit, while another major claim is genuinely external.

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

The paper introduces no new physical entities or mathematical objects. Its load-bearing commitments are empirical: manual label quality, transferability of frozen encoders, validity of the LIVECell benchmark, and the representativeness of a 10-image comparison. The hand-chosen hyperparameters and post hoc exclusions listed above are the adjustable inputs on which the headline metrics rest.

free parameters (6)
  • Composite loss weights = 0.3 focal, 0.6 dice, 0.1 boundary
    Chosen by hand in Section 3.5; directly shapes the optimization objective.
  • Learning rate schedule = eta0=1e-4, gamma=0.9, T=9000
    Set in Section 3.5; typical but not derived.
  • Progressive dropout rates = 0.25-0.5 (MODEL-1), 0.1 (MODEL-2), 0.2-0.4 (MODEL-3)
    Hand-selected per model in Section 3.4; affects regularization.
  • Ensemble size and selection = 14 classifiers
    Selected by best validation metrics in Section 3.5; the choice is data-dependent and affects all reported results.
  • LIVECell exclusion list = 8 images
    Post hoc removal of 'corrupted' samples in Section 4.2; no automated criterion given.
  • Statistical baseline threshold = 0.75
    Arbitrary threshold in Section 4.5 for the t-test; not motivated by clinical or practical requirements.
assumptions (6)
  • domain assumption Manual annotations by subject matter experts are accurate ground truth
    Section 3.3.1 states masks were generated manually with CVAT; no inter-annotator agreement is reported, and all metrics inherit this label quality.
  • domain assumption Frozen ImageNet-pretrained encoders transfer to brightfield microscopy
    Section 3.4 freezes encoders to 'retain the generalization capacity of the pre-trained features'; this transferability is assumed, not demonstrated.
  • domain assumption LIVECell dataset is a valid external benchmark
    Section 4.2 uses LIVECell as the external test; the paper assumes its annotations and image distribution are a fair measure of generalization.
  • domain assumption The 80:20 split is sufficient and no separate test set is needed
    Section 3.3.3 describes only a Train/Validation split; the paper still reports 'test accuracy 93%', implying validation is treated as test.
  • domain assumption The 10 brightfield images are representative
    Table 2 reports SOTA comparison on 10 images with no sampling protocol; the paper generalizes from this sample to 'outperforms existing methods'.
  • domain assumption F1 scores across LIVECell images are i.i.d. samples suitable for t-tests
    Section 4.5 runs a one-sample t-test and Cohen's d on F1 scores, ignoring spatial correlation and non-normality.

how reviews work

0 comments
Cite this review

Pith. "Pith review of High-Throughput Low-Cost Segmentation of Brightfield Microscopy Live Cell Images." pith.science (2026). https://pith.science/paper/FSPTDFKV

@misc{pith2026250814106,
  author       = {Pith},
  title        = {Pith review of: High-Throughput Low-Cost Segmentation of Brightfield Microscopy Live Cell Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FSPTDFKV}},
  note         = {Machine review of arXiv:2508.14106}
}
read the original abstract

Live cell culture is crucial in biomedical studies for analyzing cell properties and dynamics in vitro. This study focuses on segmenting unstained live cells imaged with bright-field microscopy. While many segmentation approaches exist for microscopic images, none consistently address the challenges of bright-field live-cell imaging with high throughput, where temporal phenotype changes, low contrast, noise, and motion-induced blur from cellular movement remain major obstacles. We developed a low-cost CNN-based pipeline incorporating comparative analysis of frozen encoders within a unified U-Net architecture enhanced with attention mechanisms, instance-aware systems, adaptive loss functions, hard instance retraining, dynamic learning rates, progressive mechanisms to mitigate overfitting, and an ensemble technique. The model was validated on a public dataset featuring diverse live cell variants, showing consistent competitiveness with state-of-the-art methods, achieving 93% test accuracy and an average F1-score of 89% (std. 0.07) on low-contrast, noisy, and blurry images. Notably, the model was trained primarily on bright-field images with limited exposure to phase- contrast microscopy (<20%), yet it generalized effectively to the phase-contrast LIVECell dataset, demonstrating modality, robustness and strong performance. This highlights its potential for real- world laboratory deployment across imaging conditions. The model requires minimal compute power and is adaptable using basic deep learning setups such as Google Colab, making it practical for training on other cell variants. Our pipeline outperforms existing methods in robustness and precision for bright-field microscopy segmentation. The code and dataset are available for reproducibility 1.

Figures

Figures reproduced from arXiv: 2508.14106 by the authors.

Figure 1
Figure 1. Instances used in the model pipelines 4 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Different augmentation results of an instance along with its mask. Top row (left to right): [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. From left: Image, Ground truth, Segmentation result produced by Densenet, Efficientnet, Inception, Mobilenet, [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: (Left) Validation F1-score comparison across all models. DenseNet achieves the highest score, followed closely [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Random samples of four images of myoblast (C2C12) and their corresponding segmentation outcomes generated [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Self-supervised Pretraining of Cell Segmentation Models

    cs.CV 2026-04 unverdicted novelty 6.0 of 10

    DINOCell achieves a SEG score of 0.784 on LIVECell by self-supervised domain adaptation of DINOv2, improving 10.42% over SAM-based models and showing strong zero-shot transfer.

Reference graph

Works this paper leans on

48 extracted references · 43 canonical work pages · cited by 1 Pith paper

  1. [1]

    Greenwade, The Comprehensive TeX Archive Network (CTAN), TUGBoat 14(3) (1993) 342–351

    G.D. Greenwade, The Comprehensive TeX Archive Network (CTAN), TUGBoat 14(3) (1993) 342–351

    work page 1993

  2. [2]

    R. Zhu, D. Sui, H. Qin, A. Hao, An extended type cell detection and counting method based on FCN, in: 2017 IEEE 17th International Conference on Bioinformatics and Bioengineering (BIBE), IEEE, 2017, pp. 51–56

    work page 2017

  3. [3]

    Ferreira, G.F

    E.K.G.D. Ferreira, G.F. Silveira, Classification and counting of cells in brightfield microscopy images: an application of convolutional neural networks, Scientific Reports 14(1) (2024) 9031

    work page 2024

  4. [4]

    B. Peng, J. Chen, P.B. Githinji, I. Gul, Q. Ye, M. Chen, P. Qin, X. Huang, C. Yan, D. Yu, et al., Practical guidelines for cell segmentation models under optical aberrations in microscopy, Computational and Structural Biotechnology Journal 26 (2024) 23–39

    work page 2024

  5. [5]

    X. Chen, B. Zheng, H. Liu, Optical and digital microscopic imaging techniques and applications in pathology, Analytical Cellular Pathology 34(1-2) (2011) 5–18

    work page 2011

  6. [6]

    M.A.S. Ali, K. Hollo, T. Laasfeld, J. Torp, M.-J. Tahk, A. Rinken, K. Palo, L. Parts, D. Fishman, ArtSeg—Artifact segmentation and removal in brightfield cell microscopy images without manual pixel-level annotations, Scientific Reports 12(1) (2022) 11404

    work page 2022

  7. [7]

    Al-Kofahi, A

    Y. Al-Kofahi, A. Zaltsman, R. Graves, W. Marshall, M. Rusu, A deep learning-based algorithm for 2-D cell segmentation in microscopy images, BMC Bioinformatics 19 (2018) 1–11

    work page 2018

  8. [8]

    Cheng, W

    J. Cheng, W. Xiong, S.C. Chia, J.H. Lim, S. Sankaran, S. Ahmed, Neurosphere segmentation in brightfield images, in: Medical Imaging 2014: Image Processing, SPIE, 2014, Vol. 9034, pp. 1148–1154

    work page 2014

Show all 48 references
  1. [9]

    R. Ali, M. Gooding, M. Christlieb, M. Brady, Advanced phase-based segmentation of multiple cells from brightfield microscopy images, in: 2008 5th IEEE International Symposium on Biomedical Imaging: From Nano to Macro, IEEE, 2008, pp. 181–184

    2008

  2. [10]

    Chen, J.W.L

    Y. Chen, J.W.L. Wan, Bright-field cell image segmentation by principal component pursuit with an Ncut penalization, in: Medical Imaging 2015: Image Processing, SPIE, 2015, Vol. 9413, pp. 912–919

    2015

  3. [11]

    Wiggins, P.J

    L. Wiggins, P.J. O’Toole, W.J. Brackenbury, J. Wilson, Exploring the impact of variability in cell segmentation and tracking approaches, Microscopy Research and Technique 88(3) (2025) 716–731

    2025

  4. [12]

    Bhattarai, J

    A. Bhattarai, J. Meyer, L. Petersilie, S.I. Shah, L.A. Neu, C.R. Rose, G. Ullah, Deep-Learning- Based Segmentation of Cells and Analysis (DL-SCAN), Biomolecules 14(11) (2024) 1348

    2024

  5. [13]

    Ronneberger, P

    O. Ronneberger, P. Fischer, T. Brox, U-net: Convolutional networks for biomedical image segmen- tation, in: Medical image computing and computer-assisted intervention–MICCAI 2015, Springer, 2015, pp. 234–241

    2015

  6. [14]

    Minaee, Y

    S. Minaee, Y. Boykov, F. Porikli, A. Plaza, N. Kehtarnavaz, D. Terzopoulos, Image segmentation using deep learning: A survey, IEEE Transactions on Pattern Analysis and Machine Intelligence 44(7) (2021) 3523–3542. 16

    2021

  7. [15]

    Khalifa, E

    A.F. Khalifa, E. Badr, Deep learning for image segmentation: a focus on medical imaging, Com- puters, Materials and Continua 75(1) (2023) 1995–2024

    2023

  8. [16]

    Kakarwal, P

    S. Kakarwal, P. Paithane, Automatic pancreas segmentation using ResNet-18 deep learning ap- proach, System Research and Information Technologies 2 (2022) 104–116

    2022

  9. [17]

    Chaurasia, E

    A. Chaurasia, E. Culurciello, Linknet: Exploiting encoder representations for efficient semantic segmentation, in: 2017 IEEE Visual Communications and Image Processing (VCIP), IEEE, 2017, pp. 1–4

    2017

  10. [18]

    Rajesh, B.S

    M.N. Rajesh, B.S. Chandrasekar, Prostate gland segmentation using semantic segmentation mod- els u-net and linknet, Int J Eng Trends Technol 70(20) (2022) 252–271

    2022

  11. [19]

    Chen, Application of Resnet18-Unet in separating tumors from brain MRI images, J

    Y. Chen, Application of Resnet18-Unet in separating tumors from brain MRI images, J. Phys.: Conf. Ser. 2580(1) (2023) 012057

    2023

  12. [20]

    S.Y. Wu, N. Dugan, B.M. Hennelly, Investigation of autofocus algorithms for brightfield mi- croscopy of unstained cells, in: Optical Modelling and Design III, SPIE, 2014, Vol. 9131, pp. 205–216

    2014

  13. [21]

    Bradbury, J.W.L

    L. Bradbury, J.W.L. Wan, A spectral k-means approach to bright-field cell image segmentation, in: 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology, IEEE, 2010, pp. 4748–4751

    2010

  14. [22]

    Anwar, M

    S.M. Anwar, M. Majid, A. Qayyum, M. Awais, M. Alnowami, M.K. Khan, Medical image analysis using convolutional neural networks: a review, Journal of Medical Systems 42 (2018) 1–13

    2018

  15. [23]

    LeCun, Y

    Y. LeCun, Y. Bengio, G. Hinton, Deep learning, Nature 521(7553) (2015) 436–444

    2015

  16. [24]

    Awasthi, S.B

    A. Awasthi, S.B. Rao, K. Acharjya, Efficient Image Feature Extraction using Convolutional Neural Networks, in: 2024 3rd International Conference for Advancement in Technology (ICONAT), IEEE, 2024, pp. 1–5

    2024

  17. [25]

    Paoletti, J.M

    M.E. Paoletti, J.M. Haut, J. Plaza, A. Plaza, Deep learning classifiers for hyperspectral imaging: A review, ISPRS Journal of Photogrammetry and Remote Sensing 158 (2019) 279–317

    2019

  18. [26]

    Y. Guo, Y. Liu, A. Oerlemans, S. Lao, S. Wu, M.S. Lew, Deep learning for visual understanding: A review, Neurocomputing 187 (2016) 27–48

    2016

  19. [27]

    A. Khan, A. Sohail, U. Zahoora, A.S. Qureshi, A survey of the recent architectures of deep convolutional neural networks, Artificial Intelligence Review 53 (2020) 5455–5516

    2020

  20. [28]

    Scherer, A

    D. Scherer, A. M¨ uller, S. Behnke, Evaluation of pooling operations in convolutional architectures for object recognition, in: International Conference on Artificial Neural Networks, Springer, 2010, pp. 92–101

    2010

  21. [29]

    Roopa, K.N

    R.B.S. Roopa, K.N. Prema, S.M. Smitha, Exploring Convolutional Neural Networks: Architec- tures, Training Strategies, and Applications, Int. J. Future Multidisciplinary Research 6(5) (2024). doi:10.36948/ijfmr.2024.v06i05.28671

    2024 doi

  22. [30]

    Lulevich, Y.-P

    V. Lulevich, Y.-P. Shih, S.H. Lo, G. Liu, Cell tracing dyes significantly change single cell mechan- ics, J. Phys. Chem. B 113(18) (2009) 6511–6519

    2009

  23. [31]

    Mualla, S

    F. Mualla, S. Sch¨ oll, B. Sommerfeldt, A. Maier, J. Hornegger, Automatic cell detection in bright- field microscope images using SIFT, random forests, and hierarchical clustering, IEEE Trans. Med. Imaging 32(12) (2013) 2274–2286

    2013

  24. [32]

    X. Ma, Z. Zhang, M. Yao, J. Peng, J. Zhong, Spatially-incoherent annular illumination microscopy for bright-field optical sectioning, Ultramicroscopy 195 (2018) 74–84

    2018

  25. [33]

    ˇCepa, Segmentation of total cell area in brightfield microscopy images, Methods and Protocols 1(4) (2018) 43

    M. ˇCepa, Segmentation of total cell area in brightfield microscopy images, Methods and Protocols 1(4) (2018) 43. 17

    2018

  26. [34]

    Y. Wang, W. Wang, D. Liu, W. Hou, T. Zhou, Z. Ji, GeneSegNet: a deep learning framework for cell segmentation by integrating gene expression and imaging, Genome Biology 24(1) (2023) 235

    2023

  27. [35]

    Greenwald, G

    N.F. Greenwald, G. Miller, E. Moen, A. Kong, A. Kagel, T. Dougherty, C.C. Fullaway, B.J. McIntosh, K.X. Leow, M.S. Schwartz, et al., Whole-cell segmentation of tissue images with human- level performance using large-scale data annotation and deep learning, Nature Biotechnology...

    2022

  28. [36]

    Sahraeian, R

    S.M.E. Sahraeian, R. Liu, B. Lau, K. Podesta, M. Mohiyuddin, H.Y.K. Lam, Deep convolutional neural networks for accurate somatic mutation detection, Nature Communications 10(1) (2019) 1041

    2019

  29. [37]

    Lam, J.M

    V.K. Lam, J.M. Byers, M.C. Robitaille, L. Kaler, J.A. Christodoulides, M.P. Raphael, A self- supervised learning approach for high throughput and high content cell segmentation, Communi- cations Biology 8 (2025) 780. doi:10.1038/s42003-025-08190-w

    2025 doi

  30. [38]

    A.A. Taha, A. Hanbury, Image segmentation evaluation: A survey of methods, Pattern Recogni- tion 73 (2018) 359–377

    2018

  31. [39]

    Simonyan, A

    K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, arXiv preprint arXiv:1409.1556 (2014)

    2014 arXiv

  32. [40]

    Oktay, J

    O. Oktay, J. Schlemper, L.L. Folgoc, M. Lee, M. Heinrich, K. Misawa, K. Mori, S. McDonagh, N. Hammerla, B. Kainz, et al., Attention U-Net: Learning where to look for the pancreas, arXiv preprint arXiv:1804.03999 (2018)

    2018 arXiv

  33. [41]

    Micikevicius, S

    P. Micikevicius, S. Narang, J. Alben, G. Diamos, E. Elsen, D. Garcia, B. Ginsburg, M. Houston, O. Kuchaiev, G. Venkatesh, et al., Mixed precision training, arXiv preprint arXiv:1710.03740 (2018)

    2018 arXiv

  34. [42]

    Pachitariu, M

    M. Pachitariu, M. Rariden, C. Stringer, Cellpose-SAM: superhuman generalization for cellular segmentation, bioRxiv (2025). doi:10.1101/2025.04.28.651001

    2025 doi

  35. [43]

    Stringer, T

    C. Stringer, T. Wang, M. Michaelos, M. Pachitariu, Cellpose: a generalist algorithm for cellular segmentation, Nature Methods 18 (2021) 100–106

    2021

  36. [44]

    N.U. Din, J. Yu, Unsupervised deep learning method for cell segmentation, bioRxiv (2021). doi:10.1101/2021.05.17.444529

    2021 doi

  37. [45]

    Lugagne, H

    J.-B. Lugagne, H. Lin, M.J. Dunlop, DeLTA: Automated cell segmentation, tracking, and lin- eage reconstruction using deep learning, PLOS Computational Biology 16(4) (2020) e1007673. doi:10.1371/journal.pcbi.1007673

    2020 doi

  38. [46]

    Fotos, P

    A. Fotos, P. Campbell, P. Murray, et al., Deep learning enhanced Watershed for microstructural analysis using a boundary class semantic segmentation, Journal of Materials Science 58(35) (2023) 14390–14410. doi:10.1007/s10853-023-08901-w

    2023 doi

  39. [47]

    Pachitariu, C

    M. Pachitariu, C. Stringer, M. Rariden, Cellpose-sam: superhuman generalization for cellular seg- mentation, bioRxiv (2025). URL: https://www.biorxiv.org/content/10.1101/2025.04.28.651001v1

    2025 doi

  40. [48]

    Ghaznavi, R

    A. Ghaznavi, R. Rycht´ arikov´ a, M. Saberioon, D. ˇStys, Cell segmentation from telecen- tric bright-field transmitted light microscopy images using a Residual Attention U-Net: A case study on HeLa line, Computers in Biology and Medicine 147 (2022) 105805. doi:10.1016/j.compb...

    2022

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.