REVIEW 4 major objections 4 minor 24 references
Biologically Inspired Deep Learning Approaches for Fetal Ultrasound Image Classification
T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read A biologically inspired dual-branch ensemble classifies 16 second-trimester fetal ultrasound structures at 85% accuracy, and on the hardest class it matches the 54% agreement rate of expert clinicians.
desk verdict The 16-class dataset is a step forward, but the paper's reported numbers don't survive contact with its own tables, and the missing split protocol undermines every quantitative claim. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing mechanism is the two-branch ensemble: the same input image is resized to a low-resolution version for a shallow EfficientNet-B0 branch and to a high-resolution version for a detailed EfficientNet-B6 branch, and the global-average-pooled feature vectors of both branches are concatenated before classification. This coarse-to-fine stacking mirrors the hierarchical organization of biological vision, while the LDAM-Focal loss combines a label-distribution-aware margin with focal weighting to push performance on underrepresented classes. The ground truth itself is produced by reconciling three clinicians' annotations with the Dawid-Skene algorithm, and on-the-fly augmentation (gamma correction, random crops, flips, color jitter, blur, translation) is used to simulate real-world acquisition variability.
What would settle it
A reader who counts the per-class accuracies in Table 3 finds 7 of 16 classes above 85% and 14 above 75%, not the 75% and 90% the abstract claims; an independent held-out evaluation on data never used for model selection would settle whether the overall 85% accuracy holds.
Extended reading notes
Core claim
The paper's central claim is that a dual-branch ensemble which first captures coarse, low-resolution context and then refines it with high-resolution detail can simultaneously distinguish 16 second-trimester fetal structures in noisy, routinely acquired ultrasound images. Trained with LDAM-Focal loss, the EfficientNet-B0 plus EfficientNet-B6 ensemble achieves 85% overall accuracy and 0.86 F1-score, with per-class accuracy above 80% for 12 of 16 anatomical categories. On the hardest class, umbilical cord attachment at the anterior abdominal wall, the model reaches 55% accuracy, closely matching the 54% inter-expert agreement measured in a supplementary labeling experiment; the authors interpret this as reaching the practical accuracy limit set by annotation ambiguity rather than by model capacity.
Load-bearing premise
The reported accuracies are treated as reliable estimates of performance even though the paper does not describe the train/validation/test split, gives no confidence intervals, and selected the final configuration after roughly sixty experiments on the same dataset.
Editorial extensions
If this is right
- A lightweight two-branch model can cover 16 fetal anatomical classes simultaneously, not just a handful, at 85% overall accuracy and 0.86 F1-score.
- On the umbilical-cord anterior-abdominal-wall class, the model's 55% accuracy tracks the 54% inter-expert agreement, so further improvement on that class likely requires cleaner label definitions or additional imaging information, not larger models.
- The combination of LDAM and focal loss is the key to balanced performance on underrepresented classes, with per-class accuracy exceeding 80% for 12 of 16 structures.
- This level of multi-class coverage makes the approach a candidate for real-time, low-cost decision support in second-trimester screening, especially in settings where sonographer expertise is scarce.
Reading between the lines
- A consequence the authors leave implicit is that if the hardest class truly has only about 54% human agreement, then per-image accuracy on that class is not a well-defined target; the 55% result is perhaps better interpreted as agreement with noisy labels than as classification error.
- The dual-branch design suggests a natural extension: instead of only concatenating pooled features, one could add an anatomical-prior layer that models co-occurrence (for example, umbilical cord near the anterior abdominal wall) to resolve classes that often appear as secondary structures.
- Because the paper does not report the train/validation/test split, a straightforward next experiment is to freeze the final configuration and evaluate on an independent multi-site test set; the 85% figure would be materially stronger after such confirmation.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper presents a dual-branch ensemble (EfficientNet-B0 as a 'shallow' branch and EfficientNet-B6 as a 'detailed' branch, combined with LDAM-Focal loss) for classifying 16 fetal anatomical structures in second-trimester ultrasound images. The authors describe a 5,298-image dataset annotated by three clinicians and reconciled with the Dawid-Skene method, on-the-fly augmentations, and a comparison of five loss functions. The headline results are 85% overall accuracy, 0.86 F1-score, 55% accuracy on the hardest class (Umbilical Cord Anterior Abdominal Wall), and a claimed correspondence between this 55% and a measured 54% human expert agreement. They also claim in the abstract that 90% of organs achieve accuracy above 0.75 and 75% above 0.85.
Significance. If supported by a rigorous evaluation, this would be a useful contribution to fetal ultrasound classification: the 16-class scope is broader than many prior studies, the use of routinely acquired noisy clinical images is realistic, and the comparison of imbalance-aware losses is practically relevant. The dataset and Dawid-Skene annotation procedure are assets. However, the current manuscript contains major reporting inconsistencies and an underspecified evaluation protocol that, as it stands, prevent the headline quantitative claims from being accepted at face value.
major comments (4)
- [§2.1, §3] The paper never describes how the 5,298 images are divided into training, validation, and test sets. Section 3 states that 'approximately sixty experiments' were run and the best configuration was selected, with Table 2 listing the top ten results. If the reported metrics were computed on data that also influenced the selection of the architecture, loss, and hyperparameters, then the 85% accuracy, 0.86 F1, and 55% class accuracy are optimistically biased estimates of generalization. A precise description of the split (including whether it is image-level or patient-level) and confirmation that the test set was used only once are required for the central claim.
- [Abstract, §4] The abstract and discussion claim that 90% of organs have accuracy above 0.75 and 75% have accuracy above 0.85, but Table 3 does not support these numbers. Counting from Table 3, 14 of 16 classes (87.5%) exceed 0.75 (exceptions: 55% and 70%), and only 8 of 16 classes (50%) exceed 0.85. The authors should either recompute the aggregate percentages or correct the text.
- [§4] The expert-agreement experiment is internally impossible as reported: Expert A identified 880 images as positive, Expert B marked 455 images, yet the paper states that 492 annotations overlapped. Since an intersection cannot exceed the smaller set (455), the reported overlap is arithmetically inconsistent. The 54% agreement figure needs to be recomputed from the correct overlap or the description of the annotations must be revised.
- [§2.4, Table 2] No confidence intervals, standard deviations, or repeated-run variability are reported for any metric. Given the class imbalance and the small size of the hardest class (149 images for Umbilical Cord Anterior Abdominal Wall), the 55% accuracy is a single point estimate whose uncertainty should be quantified. Reporting bootstrap or repeated-seed intervals would substantially strengthen the evaluation.
minor comments (4)
- [Table 3] Class labels are inconsistent with Table 1: 'Shoulder bone' appears in Table 3 but not in Table 1, which lists 'Humerus'; 'Nasal triangle' appears instead of 'Nasolabial triangle'; and 'Placenta (Umbilical cord)' differs from 'Umbilical cord (placenta)' in Table 1.
- [§2.2] In the first bullet of Section 2.2, the sentence 'In some cases, scaling factors are applied rather than fixed dimensional rescaling.' is incomplete and should be removed or expanded.
- [References] The Focal Loss reference is attributed to 'Ross and Dollár, 2017'; the correct author list is Lin, Goyal, Girshick, He, and Dollár. This should be corrected.
- [§3] The paper reports 'approximately sixty experiments' but does not specify hyperparameter ranges, augmentation parameter ranges, or random seeds, making the results difficult to reproduce independently.
Circularity Check
No circularity: the paper reports empirical measurements of a trained ensemble, and no load-bearing claim reduces to its own inputs by construction.
full rationale
The paper's central claims are empirical evaluations of a trained ensemble (EfficientNet-B0 + EfficientNet-B6 with LDAM-Focal loss) on a clinical ultrasound dataset. No equation in Sections 2-3 defines a predicted quantity in terms of the reported accuracy, F1-score, or per-class accuracy; the architecture, loss functions, and metrics are standard and externally cited. The 'biologically inspired' framing is an analogy, not a derivation, so the architecture is not mathematically obtained from the results. The intro self-citations (Aliev et al., Alsahanova et al., Bernstein et al., Pavlov et al.) are background references and none is load-bearing for the central performance claim. The internal numerical inconsistencies (Table 3 vs. the abstract's 90%/75% statements and the impossible 492-overlap expert count) and the absence of a described train/validation/test split are serious validity and reporting concerns, but they are not circular reductions: they do not show that an output equals an input by construction or that a fitted parameter was renamed as a prediction. Accordingly, no circular step can be quoted, and the score is 0.
Assumptions & free parameters
free parameters (4)
- EfficientNet-B0 and EfficientNet-B6 network weights =
Not shipped
- LDAM-Focal loss hyperparameters (alpha, beta, gamma) =
Not reported
- Input resolution pair (224x224 and 528x528) =
Approximately 224 and 528
- Augmentation parameter ranges =
Not fully specified
assumptions (3)
- domain assumption Dawid-Skene consensus labels reflect reliable ground truth for all 16 classes.
- domain assumption The single-institution GE ultrasound dataset is representative of clinical second-trimester fetal ultrasound.
- ad hoc to paper The reported test performance is measured on data not used for model selection.
Cite this review
Pith. "Pith review of Biologically Inspired Deep Learning Approaches for Fetal Ultrasound Image Classification." pith.science (2026). https://pith.science/paper/JRKFXZEV
@misc{pith2026250608623,
author = {Pith},
title = {Pith review of: Biologically Inspired Deep Learning Approaches for Fetal Ultrasound Image Classification},
year = {2026},
howpublished = {\url{https://pith.science/paper/JRKFXZEV}},
note = {Machine review of arXiv:2506.08623}
}
read the original abstract
Accurate classification of second-trimester fetal ultrasound images remains challenging due to low image quality, high intra-class variability, and significant class imbalance. In this work, we introduce a simple yet powerful, biologically inspired deep learning ensemble framework that-unlike prior studies focused on only a handful of anatomical targets-simultaneously distinguishes 16 fetal structures. Drawing on the hierarchical, modular organization of biological vision systems, our model stacks two complementary branches (a "shallow" path for coarse, low-resolution cues and a "detailed" path for fine, high-resolution features), concatenating their outputs for final prediction. To our knowledge, no existing method has addressed such a large number of classes with a comparably lightweight architecture. We trained and evaluated on 5,298 routinely acquired clinical images (annotated by three experts and reconciled via Dawid-Skene), reflecting real-world noise and variability rather than a "cleaned" dataset. Despite this complexity, our ensemble (EfficientNet-B0 + EfficientNet-B6 with LDAM-Focal loss) identifies 90% of organs with accuracy > 0.75 and 75% of organs with accuracy > 0.85-performance competitive with more elaborate models applied to far fewer categories. These results demonstrate that biologically inspired modular stacking can yield robust, scalable fetal anatomy recognition in challenging clinical settings.
Figures
Reference graph
Works this paper leans on
-
[1]
write newline
" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...
-
[2]
author Aliev, R. , author Kondrateva, E. , author Sharaev, M. , author Bronov, O. , author Marinets, A. , author Subbotin, S. , author Bernstein, A. , author Burnaev, E. , year 2020 . title Convolutional neural networks for automatic detection of focal cortical dysplasia , in: booktitle Advances in Cognitive Research, Artificial Intelligence and Neuroinfo...
work page 2020
-
[3]
author Alsahanova, N. , author Yarkin, V. , author Spodarev, E. , author Bronov, O. , author Bychenko, V. , author Marinets, A. , author Syrkashev, E. , author Karpov, O. , author Burnaev, E. , author Bernstein, A. , author Alferova, V. , author Sharaev, M. , year 2025 . title Knowledge-informed randomized machine learning and data fusion for anomaly area...
-
[4]
author Baumgartner, C.F. , author Kamnitsas, K. , author Matthew, J. , author Fletcher, T.P. , author Smith, S. , author Koch, L.M. , author Kainz, B. , author Rueckert, D. , year 2017 . title Sononet: real-time detection and localisation of fetal standard scan planes in freehand ultrasound . journal IEEE transactions on medical imaging volume 36 , pages ...
work page 2017
-
[5]
author Bernstein, A.V. , author Burnaev, E.V. , author Sharaev, M.G. , author Kondrateva, E.A. , author Kachan, O.V. , year 2020 . title Topological data analysis in computer vision , in: editor Osten, W. , editor Nikolaev, D.P. (Eds.), booktitle Twelfth International Conference on Machine Vision ( ICMV 2019) , publisher SPIE . p. pages 140 . :10.1117/12.2562501
-
[6]
author Boumeridja, H. , author Ammar, M. , author Alzubaidi, M. , author Mahmoudi, S. , author Benamer, L.N. , author Agus, M. , author Househ, M. , author Lekadir, K. , author El Habib Daho, M. , year 2025 . title Enhancing fetal ultrasound image quality and anatomical plane recognition in low-resource settings using super-resolution models . journal Sci...
work page 2025
-
[7]
, author Coronado-Guti \'e rrez, D
author Burgos-Artizzu, X.P. , author Coronado-Guti \'e rrez, D. , author Valenzuela-Alcaraz, B. , author Bonet-Carne, E. , author Eixarch, E. , author Crispi, F. , author Gratac \'o s, E. , year 2020 . title Evaluation of deep convolutional neural networks for automatic classification of common maternal fetal ultrasound planes . journal Scientific Reports...
work page 2020
-
[8]
author Cao, K. , author Wei, C. , author Gaidon, A. , author Arechiga, N. , author Ma, T. , year 2019 . title Learning imbalanced datasets with label-distribution-aware margin loss . journal Advances in neural information processing systems volume 32
work page 2019
Show all 24 references
-
[9]
, author Ni, D
author Chen, H. , author Ni, D. , author Qin, J. , author Li, S. , author Yang, X. , author Wang, T. , author Heng, P.A. , year 2015 . title Standard plane localization in fetal ultrasound via domain transferred deep neural networks . journal IEEE journal of biomedical and hea...
2015
-
[10]
, author McIntire, D.D
author Dashe, J.S. , author McIntire, D.D. , author Twickler, D.M. , year 2009 . title Effect of maternal obesity on the ultrasound detection of anomalous fetuses . journal Obstetrics & Gynecology volume 113 , pages 1001--1007
2009
-
[11]
, author Skene, A.M
author Dawid, A.P. , author Skene, A.M. , year 1979 . title Maximum likelihood estimation of observer error-rates using the em algorithm . journal Journal of the Royal Statistical Society: Series C (Applied Statistics) volume 28 , pages 20--28
1979
-
[12]
, author Villani, F.P
author Fiorentino, M.C. , author Villani, F.P. , author Di Cosmo, M. , author Frontoni, E. , author Moccia, S. , year 2023 . title A review on deep-learning algorithms for fetal ultrasound-image analysis . journal Medical image analysis volume 83 , pages 102629
2023
-
[13]
, author Lee, W
author Gon c alves, L.F. , author Lee, W. , author Espinoza, J. , author Romero, R. , year 2005 . title Three-and 4-dimensional ultrasound in obstetric practice: does it help? journal Journal of Ultrasound in Medicine volume 24 , pages 1599--1624
2005
-
[14]
, author Zhang, X
author He, K. , author Zhang, X. , author Ren, S. , author Sun, J. , year 2016 . title Deep residual learning for image recognition , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 770--778
2016
-
[15]
, year 2014
author Kingma, D.P. , year 2014 . title Adam: A method for stochastic optimization . journal arXiv preprint arXiv:1412.6980
2014 arXiv
-
[16]
, author Jimenez-Cruz, J
author Matschl, J. , author Jimenez-Cruz, J. , author Sch \"a fer, V.S. , author Wittek, A. , author Berg, C. , author Geipel, A. , author Gembruch, U. , author Strizek, B. , author Recker, F. , year 2024 . title Current status of ultrasound training in obstetrics and gynecolo...
2024
-
[17]
, author Gulino, F.A
author Padula, F. , author Gulino, F.A. , author Capriglione, S. , author Giorlandino, M. , author Cignini, P. , author Mastrandrea, M.L. , author D'Emidio, L. , author Giorlandino, C. , year 2015 . title What is the rate of incomplete fetal anatomic surveys during a second-tr...
2015
-
[18]
, author Tafheem, M
author Patil, V. , author Tafheem, M. , author Melkundi, S. , author Masimade, S. , year 2022 . title A comparative study of fetal mri and antenatal ultrasound in prenatal diagnosis with histopathological correlation . journal Journal of Fetal Medicine volume 9 , pages 83--90
2022
-
[19]
, author Artemov, A
author Pavlov, S. , author Artemov, A. , author Sharaev, M. , author Bernstein, A. , author Burnaev, E. , year 2019 . title Weakly supervised fine tuning approach for brain tumor segmentation problem , in: booktitle 2019 18th IEEE International Conference On Machine Learning A...
2019
-
[20]
, author Doll \'a r, G
author Ross, T.Y. , author Doll \'a r, G. , year 2017 . title Focal loss for dense object detection , in: booktitle proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 2980--2988
2017
-
[21]
, author Chowdhury, L
author Sadi, A.A. , author Chowdhury, L. , author Jahan, N. , author Rafi, M.N.S. , author Chowdhury, R. , author Khan, F.A. , author Mohammed, N. , year 2022 . title Lmfloss: A hybrid loss for imbalanced medical image classification . journal arXiv preprint arXiv:2212.12741
2022 arXiv
-
[22]
, author Ioannou, C
author Sarris, I. , author Ioannou, C. , author Chamberlain, P. , author Ohuma, E. , author Roseman, F. , author Hoch, L. , author Altman, D. , author Papageorghiou, A. , author Fetal, I. , author for the 21st Century (INTERGROWTH-21st), N.G.C. , year 2012 . title Intra-and in...
2012
-
[23]
, author Le, Q
author Tan, M. , author Le, Q. , year 2019 . title Efficientnet: Rethinking model scaling for convolutional neural networks , in: booktitle International conference on machine learning , organization PMLR . pp. pages 6105--6114
2019
-
[24]
, author Zhang, J
author Xiao, S. , author Zhang, J. , author Zhu, Y. , author Zhang, Z. , author Cao, H. , author Xie, M. , author Zhang, L. , year 2023 . title Application and progress of artificial intelligence in fetal ultrasound . journal Journal of clinical medicine volume 12 , pages 3298
2023
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.