Pith. sign in

REVIEW 4 major objections 9 minor 43 references

Efficient Ultrasound Image Segmentation with Token-Conditioned Neural Cellular Automata

T0 review · 4 major / 9 minor · reviewed 2026-07-31 · grok-4.5

Pith's one-line read Token-guided cellular automata refine ultrasound features enough to beat recent lightweight segmenters on fetal head and carotid images while staying compact enough for bedside use.

desk verdict Solid ultrasound segmentation architecture paper with real multi-site evidence; the accuracy story holds, the “lightweight for POCUS” framing does not. read the letter →

arxiv 2607.24529 v1 pith:OKDOO6CO submitted 2026-07-27 eess.IV

classification eess.IV
keywords NeuralCellularAutomataUltrasoundSegmentationLightweightModelPoint-of-CareTokenFiLMFetalheadDomainshift
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

Point-of-care ultrasound needs accurate organ outlines on devices that cannot run heavy neural nets. This paper argues that the missing piece is not a bigger backbone but a cheap iterative refiner: structure-aware tokens condition Neural Cellular Automata so local cell updates improve weak boundaries without large compute. The resulting network, LANCANet, matches or exceeds recent lightweight CNN and transformer segmenters on fetal-head, carotid, and pubic-symphysis datasets, and stays competitive on African fetal-head scans it never saw in training. A sympathetic reader cares because that combination—better contours plus a design aimed at low-resource clinics—is exactly what bedside biometric tools need.

What carries the argument

Token-conditioned NCA adapters: structure tokens summarize global anatomy, Token FiLM modulates the cellular state, and shared local NCA update rules iteratively refine features for T steps—preserving backbone semantics while sharpening boundaries at limited cost.

What would settle it

Run the same CPU latency and Dice comparison on a representative handheld ultrasound device or edge board: if LANCANet is slower or less accurate than the lighter baselines under that fixed power and memory budget, the efficiency-plus-accuracy claim fails.

Watch

Extended reading notes

Core claim

LANCANet shows that attaching token-conditioned NCA adapters to a lightweight encoder–decoder yields competitive or superior ultrasound segmentation versus recent efficient CNN- and transformer-based methods, with top Dice on HC18 (96.62%) and CCA (92.86%) and best pubic-symphysis Dice on PSFHS (79.00%), while remaining usable under domain shift when trained from scratch.

Load-bearing premise

The accuracy gains still leave the model efficient enough for real point-of-care and edge devices, even though measured size and CPU speed are worse than several lighter baselines.

Editorial extensions

If this is right

  • Bedside fetal-head and carotid tools can adopt token-NCA refinement instead of larger transformers when training data are limited.
  • Pubic-symphysis segmentation in labor ultrasound can improve without abandoning a compact architecture.
  • Models trained only on standard public sets can still transfer usefully to multi-country African fetal scans without ImageNet pretraining.
  • Ablation of iteration count becomes a practical knob: moderate T improves boundaries; too many steps waste latency.

Reading between the lines

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

  • If token FiLM is the main carrier of global context, similar adapters might lift other noisy local-update models (e.g., diffusion or recurrent refiners) in ultrasound without full attention stacks.
  • The external African results suggest token-NCA may reduce reliance on large-scale natural-image pretraining for cross-device fetal biometry.
  • A natural next measurement is on-device FPS and battery draw; the paper’s CPU numbers alone do not settle bedside feasibility.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 9 minor

Summary. The paper proposes LANCANet, an ultrasound segmentation network that inserts token-conditioned Neural Cellular Automata (NCA) adapters into the two deepest encoder levels of a lightweight ConvMixer/transformer U-shaped architecture. Structure tokens are extracted by adaptive pooling and attention, mean-pooled, and injected into the NCA state via FiLM-style (γ, β) modulation before each of T iterative NCA updates. The method is evaluated on HC18, CCA, and PSFHS against nine baselines, with external validation on two African fetal-head datasets (KEN-FH, AFR-FH) under domain shift, paired t-tests, module and T-step ablations, and a parameter/GFLOP/CPU-FPS efficiency comparison. The authors report the best DSC on HC18 (96.62%) and CCA (92.86%), the best pubic-symphysis DSC on PSFHS (79.00%), competitive external performance despite training from scratch, and claim a favorable accuracy–efficiency balance suitable for resource-constrained deployment.

Significance. If the results hold, the work is a useful contribution: it is, to my knowledge, the first integration of token-conditioned NCA adapters (extending the AdaNCA idea of [24]) into ultrasound segmentation, and the evaluation is unusually complete for this literature — three in-domain datasets spanning fetal head, carotid, and multi-class intrapartum anatomy, two genuinely external African test sets under population/device shift, paired t-tests with non-significant results honestly disclosed in red, module and iteration-count ablations, and an anonymized code release. The accuracy numbers themselves (Tables I–III) are credible and the ablations support the claim that the NCA refinement is doing real work (Table VII: removing the Token Adapter or NCA TGB costs 1–2 DSC points; removing the Lite Transformer costs little). The weakness is not the accuracy evidence but the efficiency and "superiority" framing: the paper's own Table V shows LANCANet is the largest and slowest model in the comparison, and its headline advantage over SegFormer is statistically indistinguishable (Table VI). These are framing/positioning problems on top of sound empirical work, plus one table (VIII) with internal

major comments (4)
  1. [§V-E, Table V; Abstract; Conclusion] The efficiency claim is contradicted by the paper's own Table V. The abstract ('minimal computational cost'), §V-E ('remaining a compact and computationally efficient model'), and the Conclusion ('favorable balance ... for resource-constrained clinical deployment') frame LANCANet as lightweight and deployment-ready. Table V shows the opposite within the comparison set: LANCANet has the most parameters (6.70M, vs 3.71M for SegFormer, 1.47M for UNeXt, 0.05–0.32M for EGE-UNet/LB-UNet/MK-UNet), the second-highest GFLOPs (27.69, vs 10.42 for SegFormer), and the lowest throughput (3.49 FPS — slower even than UNet at 33.26 FPS and ~3.5× slower than SegFormer at 12.15 FPS). The statement in §V-E that LANCANet has 'a comparable model size' to SegFormer is not supported (1.8× the parameters, 2.7× the GFLOPs, 3.5× the latency). The paper does not need to win on efficiency to be valuable — its contr
  2. [§V-A/§V-B, Tables III, IV, VI] The headline 'superior' claim over the strongest baseline is not statistically supported, and the strongest baseline also wins the external evaluation. Table VI reports the average-DSC gain over SegFormer as non-significant (p=0.39 DSC, p=0.24 HD95), and Table IV shows non-significant differences on HC18 (p=0.46), PSFHS PS (p=0.51), and FH (p=0.33). Meanwhile Table III shows SegFormer is best on both external sets (KEN-FH 92.90 vs 89.63 DSC; AFR-FH 93.01 vs 90.64). The authors deserve credit for disclosing these p-values, and the from-scratch vs ImageNet-pretrained asymmetry is a legitimate mitigating point — but it cuts against, not for, the abstract's 'competitive or superior' and the Conclusion's framing. Since SegFormer is also cheaper on every efficiency axis (Comment 1), the paper's differentiator against this one baseline currently rests on within-noise in-domain differences. The
  3. [§V-F2, Table VIII] Table VIII is internally inconsistent in ways that suggest measurement or reporting errors in the efficiency numbers. (i) GFLOPs are reported as a constant 27.69 for T=1,2,4,8, but the NCA update (Eq. 4) is applied T times over the feature grid, so compute must grow with T unless the GFLOPs figure excludes the NCA iterations — if so, that must be stated, since the same 27.69 figure is used in Table V as the model's headline complexity. (ii) The FPS and latency columns are mutually inconsistent: 251.49 ms at T=1 implies ~3.98 FPS, not the reported 3.49; 286.34 ms at T=4 implies ~3.49 FPS, not the reported 3.64; and T=1 and T=2 report identical FPS (3.49) despite a 34 ms latency difference. (iii) The text cites 'latency rises from 274.70 ms at T=4' but Table VIII says 286.34 ms at T=4 (274.70 matches no row). These numbers feed directly into the efficiency claims of Comment 1, so they must
  4. [§IV-B vs §V-F2, Tables II and VIII] The default configuration and the ablation conclusion disagree. §IV-B sets T=2 as the default and Tables I–III report all main results at T=2, but §V-F2 concludes 'T=4 offers the best balance between segmentation performance and inference efficiency on the PSFHS dataset' — indeed at T=4 the FH DSC rises from 85.63 to 90.15 and average DSC from 82.32 to 84.17, which would change several Table II rankings (e.g., FH DSC 90.15 would be the best in the table rather than mid-pack). Either the main PSFHS results should be reported at the authors' own recommended operating point, or the discrepancy should be explained (e.g., T=2 generalizes better on HC18/CCA — if so, show it). As it stands, the reader cannot tell which configuration the paper is actually proposing.
minor comments (9)
  1. [§IV-A] §IV-A: 'we conducted experiments on four ultrasound datasets' is followed by a list of five (HC18, CCA, PSFHS, KEN-FH, AFR-FH).
  2. [Table I] Table I, HC18 row for LB-Unet: HD95 is reported as 2.94 (37.46); a mean of 2.94 with std 37.46 and that being far below all other methods' HD95 is implausible — likely a typo for 29.4 or 32.94. Please verify.
  3. [Table V / Fig. 9] Table V, EGE-UNet: 0.05M parameters with 39.71 GFLOPs is an unusual ratio (higher GFLOPs than LANCANet at 125× fewer parameters); please verify the GFLOPs figure, especially since Fig. 9's x-axis only spans 0–40 GFLOPs and this point sits at the edge.
  4. [§III-B2] §III-B2, Structure Token Extraction: N is used for both H×W ('flatten ... into N=H×W spatial cells') and the token count ('T ∈ R^{B×N×C}, where N=S²'). Please disambiguate the notation.
  5. [§III-B2] §III-B2: 'we extract a compact structural representation via adaptive pooling with output size S×S to get structure tokens T ∈ R^{B×N×C}' — the tensor named T collides with the refinement-step count T used throughout (Eq. 4, Table VIII). Rename one of them.
  6. [Eq. (2)] Eq. (2): the projections Q, K, V are applied but the dimensions of the attention output relative to the adaptive-pooled tokens (S² vs H×W queries) are not fully specified — a sentence clarifying whether tokens or cells are the queries would help reproduction.
  7. [Throughout] The running header reads 'AUTHOR et al.: TITLE' and affiliation/author placeholders appear unanonymized in the arXiv text but the code link is anonymized (anonymous.4open.science); please make the anonymization policy consistent with the venue's review model.
  8. [§V-E] Efficiency is reported only as CPU FPS on a server-class AMD EPYC 9654P. Even a single additional number (e.g., Jetson Orin or a mid-range mobile SoC latency) would substantially strengthen, or honestly bound, the POCUS-deployment motivation.
  9. [§IV-D / Tables I–III] The unified training protocol is a strength, but SegFormer's ImageNet initialization (§IV-D) breaks the 'controlled comparison' symmetry for exactly the baseline that matters most; a from-scratch SegFormer row in Tables I–III would make the comparison fully controlled.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: standard supervised segmentation with held-out metrics; architecture choices are not self-defining derivations.

full rationale

LANCANet is an empirical deep-learning segmentation paper. The load-bearing claims are comparative Dice/Jaccard/HD95/ASD on held-out splits of HC18, CCA, and PSFHS, plus zero-shot external evaluation on KEN-FH and AFR-FH. Models are trained from labeled masks under a joint Dice+CE loss; reported numbers are measurements on data not used to define the architecture. Token FiLM, structure-token extraction, and NCA update steps (Eqs. 2–4) are design choices inspired by external citations (Med-NCA, FiLM, AdaNCA) whose authors do not overlap with this paper; none of those citations is used as a uniqueness theorem that forces the result. Ablations (Tables VII–VIII) and statistical tests (Tables IV, VI) are independent checks, not fitted inputs renamed as predictions. There is no self-definitional identity between inputs and claimed outputs, no parameter fit that statistically forces a closely related “prediction,” and no renaming of a known law. Framing tension about efficiency vs. SegFormer is a positioning issue, not circularity. Circularity score is therefore 0.

Assumptions & free parameters 5 free parameters · 5 assumptions · 3 invented entities

Load-bearing content is architectural and empirical, not axiomatic physics. Claims rest on standard segmentation-learning assumptions, borrowed NCA/FiLM mechanisms, hand-chosen depth/T/token settings, and the premise that adapter-style NCA preserves semantics while refining boundaries cheaply. Free parameters are training and adapter hyperparameters; invented entities are the composite modules, not new physical objects.

free parameters (5)
  • NCA refinement steps T = default 2 (ablation best 4 on PSFHS)
    Default T=2; ablation selects T=4 as best on PSFHS. Directly controls iterative refinement and latency; chosen experimentally, not derived.
  • Structure token grid size S×S
    Adaptive pooling output size that sets number of structure tokens; architectural hyperparameter guiding global conditioning.
  • Optimizer and schedule (SGD lr=0.001, momentum=0.9, wd=1e-4, 200 epochs, batch 5) = lr=0.001, 200 epochs, batch=5
    Unified training protocol fitted for stable convergence across methods; affects all reported rankings.
  • Input resolution 448×448 and augmentation probabilities = 448×448; rot/brightness p=0.5, blur/noise p=0.3
    Fixed preprocessing/augmentation choices that influence absolute metrics and domain-shift behavior.
  • Loss = Dice + Cross-Entropy (unweighted sum) = L_Dice + L_CE
    Objective weighting is a design choice that shapes boundary vs region trade-offs.
assumptions (5)
  • domain assumption Shared local NCA update rules with residual neighborhood perception can refine medical segmentation features usefully when stacked for few iterations.
    Imported from Med-NCA / NCA literature (Methodology Preliminary and Token Adapter); not re-proved here.
  • ad hoc to paper Global structure tokens mean-pooled into FiLM (γ, β) supply enough anatomical context to condition cell updates without full heavy attention everywhere.
    Core design hypothesis of Token FiLM (Eqs. 2–4); supported only by ablations in this work.
  • domain assumption A single unified optimization protocol yields a fair ranking of heterogeneous lightweight architectures.
    Stated in Implementation Details; authors note it may leave per-model gains on the table.
  • domain assumption Standard supervised Dice/CE on expert masks is an adequate proxy for clinical segmentation utility.
    Implicit throughout Experiments/Results; no reader study or downstream biometry error analysis.
  • domain assumption Depthwise/pointwise ConvMixer blocks and group-convolution fusion preserve enough representation for ultrasound at low cost.
    Encoder/decoder design following CMU-Net/CMUNeXt-style priors.
invented entities (3)
  • LANCANet (LiteAdaNCA-Net) full architecture
    purpose: End-to-end lightweight ultrasound segmentation network combining stem, Lite ConvMixer, token-guided NCA blocks, lite transformer, and fusion decoder.
    Named composite system; performance claims attach to this whole, not a single prior module.
  • Token Adapter (Structure Token Extraction + Token FiLM + NCA)
    purpose: Iteratively condition and refine encoder features with global tokens and local cellular updates.
    Paper’s main methodological object; ablations show removing it drops HC18/CCA metrics.
  • Token FiLM conditioning of NCA state
    purpose: Map pooled structure tokens to per-channel γ, β modulating NCA state before each update.
    Specific token→cell interface beyond generic FiLM or plain NCA; validated only inside this paper’s experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Ultrasound Image Segmentation with Token-Conditioned Neural Cellular Automata." pith.science (2026). https://pith.science/paper/OKDOO6CO

@misc{pith2026260724529,
  author       = {Pith},
  title        = {Pith review of: Efficient Ultrasound Image Segmentation with Token-Conditioned Neural Cellular Automata},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OKDOO6CO}},
  note         = {Machine review of arXiv:2607.24529}
}
read the original abstract

Point-of-Care Ultrasound (POCUS) plays an important role in bedside diagnosis and clinical decision-making, particularly in resource-constrained settings. Recent deep learning methods have substantially improved ultrasound image segmentation, enabling accurate diagnosis and biometric estimation. However, their computational cost limits deployment on portable and low-resource devices. To address this challenge, we propose LiteAdaNCA-Net (LANCANet), a lightweight ultrasound segmentation framework that incorporates token-conditioned Neural Cellular Automata (NCA) adapters for iterative feature refinement. Specifically, structure-aware tokens guide local NCA refinement via Token FiLM, enabling boundary-aware feature refinement with minimal computational cost. We evaluate LANCANet on HC18, CCA, and PSFHS, and assess robustness on two independent African fetal head datasets collected from multiple clinical centers. Experimental results demonstrate that LANCANet achieves competitive or superior performance to recent lightweight CNN- and transformer-based methods. On HC18 and CCA, LANCANet achieves the highest Dice Similarity Coefficient (DSC) of 96.62\% and 92.86\%, respectively. On PSFHS, it achieves the best performance on the challenging pubic symphysis structure while maintaining competitive fetal head segmentation accuracy. Furthermore, despite being trained from scratch, LANCANet maintains competitive performance on the external KEN-FH and AFR-FH datasets under substantial domain shifts. These results show that token-conditioned NCA refinement improves segmentation accuracy and boundary localization while maintaining computational efficiency for resource-constrained clinical deployment. Our code is on \href{https://anonymous.4open.science/r/LANCAN-21A0/README.md}{GitHub}.

Figures

Figures reproduced from arXiv: 2607.24529 by the authors.

Figure 1
Figure 1. Overview of the proposed LANCANet architecture. Token-conditioned NCA adapters iteratively refine feature representations using global structure-aware guidance and local cellular interactions, enabling robust and computationally efficient ultrasound image segmentation. H W Adaptive Pooling Linear Projection Position Encoding Layer Norm Adaptive Pooling Flatten N C [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The flow of Structure Token Extraction (Cell → Token). iteratively refine spatial features through residual, stochastic updates in a parameter-efficient manner. In short, the Token Adapter performs iterative refinement for T iterations. In each iteration, Token FiLM first injects global structural context into the cellular state S, followed by an NCA update step fNCA that refines local spatial representations [18]. … view at source ↗
Figure 4
Figure 4. Visual comparison of methods on HC18 dataset. The figure shows the Ground Truth (GT) in red and the predicted results in green [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: Visual comparison of methods on CCA dataset. The figure shows the GT in red and the predicted results in green [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Visual comparison of methods on PSFHS dataset. The figure shows the GT in red and the predicted results in green. with fewer irregular regions and fewer isolated false predic￾tions. These visual improvements are particularly evident in challenging cases with weak bound…
Figure 7
Figure 7. Figure 7: Visual comparison of methods on KEN-FH dataset. The figure shows the GT in red and the predicted results in green. EGY GHA MWI LANCAN PMFSNet MK-Unet CMUNeXt LB-Unet EGE-UNet UNeXt Segformer MobileNet UNet GHA DZA UGA MWI EGY DZA UGA [PITH_FULL_IMAGE:figures/full_fig_…
Figure 8
Figure 8. Figure 8: Examples are shown from five African countries, GHA (Ghana), MWI (Malawi), EGY (Egypt), DZA (Algeria), and UGA (Uganda), illustrating the robustness of different methods under substantial cross-country and cross-device domain shifts. The GT are in red and the predicted…
Figure 9
Figure 9. Figure 9: Accuracy–efficiency comparison of different segmentation methods. Bubble size represents the number of model parameters. TABLE VI THE SIGNIFICANCE p OF PAIRED T-TEST BETWEEN LANCANET AND OTHER METHODS ON THE OVERALL AVERAGE DSC ACROSS HC18, CCA, PSFHS, KEN-FH, AND AFR-…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 1 linked inside Pith

  1. [24]

    Adanca: neural cellular automata as adaptors for more robust vision transformer,

    Yitao Xu, Tong Zhang, and Sabine S ¨usstrunk, “Adanca: neural cellular automata as adaptors for more robust vision transformer,” inProceedings of the 38th International Conference on Neural Information Processing Systems. 2024, Curran Associates Inc

  2. [1]

    Practice guidelines for performance of the routine mid-trimester fetal ultrasound scan,

    L. J. Salomon, Z. Alfirevic, V . Berghella, C. Bilardo, E. Hernandez- Andrade, S. L. Johnsen, K. Kalache, K.-Y . Leung, G. Malinger, H. Munoz, F. Prefumo, A. Toi, W. Lee, and on behalf of the ISUOG Clinical Standards Committee, “Practice guidelines for performance of the routine mid-trimester fetal ultrasound scan,”Ultrasound in Obstetrics & Gynecology, v...

  3. [2]

    Applause: Automatic prediction of placental health via u-net segmentation and statistical evaluation,

    Maximilian Pietsch, Alison Ho, Alessia Bardanzellu, Aya Mutaz Ahmad Zeidan, and et al., “Applause: Automatic prediction of placental health via u-net segmentation and statistical evaluation,”Medical Image Analysis, vol. 72, pp. 102145, 2021

  4. [3]

    Point-of-Care ultrasound,

    Linda Lee and Jeanne M DeCara, “Point-of-Care ultrasound,”Current Cardiology Reports, vol. 22, no. 11, pp. 149, Sept. 2020

  5. [4]

    Alice Self, Qingchao Chen, Bapu Koundinya Desiraju, Sumeet Dhari- wal, Alexander D Gleed, Divyanshu Mishra, and et al., “Developing clinical artificial intelligence for obstetric ultrasound to improve access in underserved regions: Protocol for a computer-assisted low-cost point- of-care ultrasound (calopus) study,”JMIR Res Protoc, vol. 11, no. 9, pp. e37...

  6. [5]

    Ultrasound image segmentation: a survey,

    J.A. Noble and D. Boukerroui, “Ultrasound image segmentation: a survey,”IEEE Transactions on Medical Imaging, vol. 25, no. 8, pp. 987–1010, 2006

  7. [6]

    Image quality improvement of hand-held ultrasound devices with a two- stage generative adversarial network,

    Zixia Zhou, Yuanyuan Wang, Yi Guo, Yanxing Qi, and Jinhua Yu, “Image quality improvement of hand-held ultrasound devices with a two- stage generative adversarial network,”IEEE Transactions on Biomedical Engineering, vol. 67, no. 1, pp. 298–311, 2020

  8. [7]

    Enhancing fetal ultrasound image quality and anatomical plane recognition in low-resource settings using super-resolution models,

    Hafida Boumeridja, Mohammed Ammar, Mahmood Alzubaidi, Sa ¨ıd Mahmoudi, Lamya Nawal Benamer, Marco Agus, Mowafa Househ, Karim Lekadir, and Mostafa El Habib Daho, “Enhancing fetal ultrasound image quality and anatomical plane recognition in low-resource settings using super-resolution models,”Scientific Reports, vol. 15, no. 1, pp. 8376, Mar. 2025

Show all 43 references
  1. [8]

    Sas: Segment anything small for ultrasound–a non-generative data augmentation technique for robust deep learning in ultrasound imaging,

    Danielle L Ferreira, Ahana Gangopadhyay, Hsi-Ming Chang, Ravi Soni, and Gopal Avinash, “Sas: Segment anything small for ultrasound–a non-generative data augmentation technique for robust deep learning in ultrasound imaging,” 2025, arXiv:2503.05916

  2. [9]

    U-Net: Convo- lutional networks for biomedical image segmentation,

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox, “U-Net: Convo- lutional networks for biomedical image segmentation,” inInternational Conference on Medical Image Computing and Computer-Assisted Inter- vention. 2015, vol. 9351, pp. 234–241, Springer

  3. [10]

    A survey on u-shaped networks in medical image segmentations,

    Liangliang Liu, Jianhong Cheng, Quan Quan, Fang-Xiang Wu, Yu-Ping Wang, and Jianxin Wang, “A survey on u-shaped networks in medical image segmentations,”Neurocomputing, vol. 409, pp. 244–258, 2020

  4. [11]

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

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weis- senborn, Xiaohua Zhai, Thomas Unterthiner, and et al., “An image is worth 16x16 words: Transformers for image recognition at scale,”ICLR, 2021

  5. [12]

    Threef-net: Fine-grained feature fusion network for breast ultrasound image segmentation,

    Xuesheng Bian, Jia Liu, Sen Xu, Weiquan Liu, Leyi Mei, Chaoshen Xiao, and Fan Yang, “Threef-net: Fine-grained feature fusion network for breast ultrasound image segmentation,”Computers in Biology and Medicine, vol. 194, pp. 110527, 2025

  6. [13]

    UNeXt: MLP-Based rapid medical image segmentation network,

    Jeya Maria Jose Valanarasu and Vishal M Patel, “UNeXt: MLP-Based rapid medical image segmentation network,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. 2022, pp. 23–33, Springer Nature Switzerland

  7. [14]

    Laednet: A lightweight attention encoder–decoder network for ultrasound medical image segmentation,

    Quan Zhou, Qianwen Wang, Yunchao Bao, Lingjun Kong, Xin Jin, and Weihua Ou, “Laednet: A lightweight attention encoder–decoder network for ultrasound medical image segmentation,”Computers and Electrical Engineering, vol. 99, pp. 107777, 2022

  8. [15]

    Cmunext: An efficient medical image segmentation network based on large kernel and skip fusion,

    Fenghe Tang, Jianrui Ding, Quan Quan, Lingtao Wang, Chunping Ning, and S. Kevin Zhou, “Cmunext: An efficient medical image segmentation network based on large kernel and skip fusion,” in2024 IEEE International Symposium on Biomedical Imaging (ISBI), 2024, pp. 1–5

  9. [16]

    Cellular automata as convolutional neural networks,

    William Gilpin, “Cellular automata as convolutional neural networks,” Physical Review E, vol. 100, no. 3, pp. 032402, 2019

  10. [17]

    Growing neural cellular automata,

    Alexander Mordvintsev, Ettore Randazzo, Eyvind Niklasson, and Michael Levin, “Growing neural cellular automata,”Distill, 2020

  11. [18]

    Med- nca: Robust and lightweight segmentation with neural cellular automata,

    John Kalkhof, Camila Gonz ´alez, and Anirban Mukhopadhyay, “Med- nca: Robust and lightweight segmentation with neural cellular automata,” inInternational Conference on Information Processing in Medical Imaging. Springer, 2023, pp. 705–716

  12. [19]

    M3d-nca: Robust 3d segmentation with built-in quality control,

    John Kalkhof and Anirban Mukhopadhyay, “M3d-nca: Robust 3d segmentation with built-in quality control,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 2023, pp. 169–178

  13. [20]

    Medsegdiffnca: Diffusion models with neural cellular automata for skin lesion segmentation,

    Avni Mittal, John Kalkhof, Anirban Mukhopadhyay, and Arnav Bhavsar, “Medsegdiffnca: Diffusion models with neural cellular automata for skin lesion segmentation,” inIEEE 38th International Symposium on Computer-Based Medical Systems. IEEE, 2025, pp. 35–40

  14. [21]

    eNCApsulate: neural cellular automata for precision diagnosis on capsule endoscopes,

    Henry John Krumb and Anirban Mukhopadhyay, “eNCApsulate: neural cellular automata for precision diagnosis on capsule endoscopes,”In- ternational Journal of Computer Assisted Radiology and Surgery, July 2025

  15. [22]

    Attention pooling enhances nca-based classification of microscopy images,

    Chen Yang, Michael Deutges, Jingsong Liu, Han Li, Nassir Navab, Carsten Marr, and Ario Sadafi, “Attention pooling enhances nca-based classification of microscopy images,” inMachine Learning in Medical Imaging. 2026, pp. 583–593, Springer Nature Switzerland

  16. [23]

    Neural cellular automata for weakly supervised segmentation of white blood cells,

    Michael Deutges, Chen Yang, Raheleh Salehi, Nassir Navab, Carsten Marr, and Ario Sadafi, “Neural cellular automata for weakly supervised segmentation of white blood cells,” inEfficient Medical Artificial Intelligence. 2026, pp. 289–298, Springer Nature Switzerland

  17. [25]

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

    Fabian Isensee, Paul F Jaeger, Simon A A Kohl, Jens Petersen, and Klaus 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, Feb. 2021

  18. [26]

    Mobilenetv2: Inverted residuals and linear bottlenecks,

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, and Liang-Chieh Chen, “Mobilenetv2: Inverted residuals and linear bottlenecks,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018

  19. [27]

    EGE-UNet: An efficient group enhanced UNet for skin lesion segmen- tation,

    Jiacheng Ruan, Mingye Xie, Jingsheng Gao, Ting Liu, and Yuzhuo Fu, “EGE-UNet: An efficient group enhanced UNet for skin lesion segmen- tation,” inInternational Conference on Medical Image Computing and Computer-Assisted Intervention. 2023, pp. 481–490, Springer

  20. [28]

    Segformer: Simple and efficient design for semantic segmentation with transformers,

    Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M Alvarez, and Ping Luo, “Segformer: Simple and efficient design for semantic segmentation with transformers,” inNeural Information Processing Systems (NeurIPS), 2021

  21. [29]

    LB-UNet: A lightweight Boundary- Assisted UNet for skin lesion segmentation,

    Jiahao Xu and Lyuyang Tong, “LB-UNet: A lightweight Boundary- Assisted UNet for skin lesion segmentation,” inInternational Confer- ence on Medical Image Computing and Computer-Assisted Intervention. 2024, pp. 361–371, Springer Nature Switzerland

  22. [30]

    Mk-unet: Multi-kernel lightweight cnn for medical image segmentation,

    Md Mostafijur Rahman and Radu Marculescu, “Mk-unet: Multi-kernel lightweight cnn for medical image segmentation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 1042–1051

  23. [31]

    Pmfsnet: Polarized multi-scale feature self- attention network for lightweight medical image segmentation,

    Jiahui Zhong, Wenhong Tian, Yuanlun Xie, Zhijia Liu, Jie Ou, Taoran Tian, and Lei Zhang, “Pmfsnet: Polarized multi-scale feature self- attention network for lightweight medical image segmentation,”Com- AUTHORet al.: TITLE 13 puter Methods and Programs in Biomedicine, vol. 261,...

  24. [32]

    Learning cellular automaton dynamics with neural networks,

    N. Wulff and J A Hertz, “Learning cellular automaton dynamics with neural networks,” inAdvances in Neural Information Processing Systems, S. Hanson, J. Cowan, and C. Giles, Eds. 1992, vol. 5, Morgan- Kaufmann

  25. [33]

    Patches are all you need?,

    Asher Trockman and J Zico Kolter, “Patches are all you need?,” Transactions on Machine Learning Research, 2023

  26. [34]

    Cmu-net: a strong convmixer-based medical ultrasound image segmentation network,

    Fenghe Tang, Lingtao Wang, Chunping Ning, Min Xian, and Jianrui Ding, “Cmu-net: a strong convmixer-based medical ultrasound image segmentation network,” inIEEE 20th international symposium on biomedical imaging (ISBI), 2023, pp. 1–5

  27. [35]

    Film: visual reasoning with a general conditioning layer,

    Ethan Perez, Florian Strub, Harm de Vries, Vincent Dumoulin, and Aaron Courville, “Film: visual reasoning with a general conditioning layer,” inProceedings of the 32nd AAAI Conference on Artificial Intelligence, 2018, AAAI-18

  28. [36]

    Attention is all you need,

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

  29. [37]

    Automated measurement of fetal head circumference using 2d ultrasound images,

    Thomas L. A. van den Heuvel, Dagmar de Bruijn, Chris L. de Korte, and Bram van Ginneken, “Automated measurement of fetal head circumference using 2d ultrasound images,”PLOS ONE, vol. 13, no. 8, pp. 1–20, Aug. 2018

  30. [38]

    Common carotid artery ultrasound images,

    Agata Momot, “Common carotid artery ultrasound images,”Mendeley Data, 2022

  31. [39]

    The jnu-ifm dataset for segmenting pubic symphysis-fetal head,

    Yaosheng Lu, Mengqiang Zhou, Dengjiang Zhi, Minghong Zhou, Xi- aosong Jiang, Ruiyu Qiu, Zhanhong Ou, Huijin Wang, Di Qiu, Mei Zhong, Xiaoxing Lu, Gaowen Chen, and Jieyun Bai, “The jnu-ifm dataset for segmenting pubic symphysis-fetal head,”Data in Brief, vol. 41, pp. 107904, 2022

  32. [40]

    Psfhs challenge report: Pubic symphysis and fetal head segmentation from intrapartum ultrasound images,

    Jieyun Bai, Zihao Zhou, Zhanhong Ou, Gregor Koehler, Raphael Stock, Klaus Maier-Hein, Marawan Elbatel, and et al., “Psfhs challenge report: Pubic symphysis and fetal head segmentation from intrapartum ultrasound images,”Medical Image Analysis, vol. 99, pp. 103353, 2025

  33. [41]

    Maternal fetal ultrasound planes from low-resource imaging settings in five african countries,

    Carla Sendra-Balcells, V ´ıctor M. Campello, Jordina Torrents-Barrena, Yahya Ali Ahmed, Mustafa Elattar, Benard Ohene-Botwe, Pempho Nyangulu, William Stones, Mohammed Ammar, Lamya Nawal Be- namer, Harriet Nalubega Kisembo, Senai Goitom Sereke, Sikolia Z. Wanyonyi, Marleen Temm...

  34. [42]

    Generalisability of fetal ultrasound deep learning models to low-resource imaging settings in five African countries,

    Carla Sendra-Balcells, V ´ıctor M. Campello, Jordina Torrents-Barrena, Yahya Ali Ahmed, Mustafa Elattar, Benard Ohene-Botwe, Pempho Nyangulu, William Stones, Mohammed Ammar, Lamya Nawal Be- namer, Harriet Nalubega Kisembo, Senai Goitom Sereke, Sikolia Z. Wanyonyi, Marleen Temm...

  35. [43]

    Imagenet: A large-scale hierarchical image database,

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei- Fei, “Imagenet: A large-scale hierarchical image database,” in2009 IEEE Conference on Computer Vision and Pattern Recognition, 2009, pp. 248–255

Pith tools

Reviewed July 31, 2026 · model on record in the stance chip above.