Pith. sign in

REVIEW 4 major objections 5 minor 61 references

Symmetric Perception and Ordinal Regression for Detecting Scoliosis Natural Image

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper argues that scoliosis severity can be graded from ordinary back photographs using symmetry matching and ordinal regression, reporting 95.11% general and 81.46% fine-grained accuracy on a new dataset.

desk verdict Reasonable idea and sensible modules, but the headline accuracy is not yet trustworthy because the image-level split likely leaks patient identity and the human-comparison figures are duplicated. read the letter →

arxiv 2411.15799 v1 pith:XXNEMT2T submitted 2024-11-24 cs.CV

Xiaojia Zhu , Rui Chen , Xiaoqi Guo , Zhiwen Shao , Yuhu Dai , Ming Zhang , Chuandong Lang This is my paper · ORCID
classification cs.CV
keywords scoliosisdetectionnaturalimagesymmetricfeaturematchingordinalregressionseveritylevelestimationhumanbackCobbangledual-pathnetwork
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that scoliosis severity can be graded directly from ordinary photographs of a patient's back, without X-rays, by exploiting the fact that a healthy back is roughly symmetric and disease makes it asymmetric. The proposed network feeds both the photo and its horizontal mirror image through a shared backbone, uses a symmetric feature matching module to compare the two sides, and replaces flat multi-class classification with an ordinal regression head that converts the severity scale into a series of binary comparisons. On a collected dataset of 1,898 back images from 1,067 patients, the method reports 95.11% accuracy for the four-level general severity task and 81.46% for the ten-level fine-grained task, above the compared state-of-the-art models and above two spine surgeons asked to grade the same photographs. The practical goal is cheap, radiation-free, wide-ranging screening for adolescent scoliosis that could happen outside specialist clinics.

What carries the argument

The load-bearing mechanism is the symmetric feature matching module: original and horizontally flipped back images go through a weight-sharing backbone, their features are concatenated and convolved into a fused feature map, and self-attention uses that fused map as key and value while the two branch features act as queries, so the network can directly compare left and right sides of the back. The ordinal regression head then rewrites the $K$-class severity problem as $K-1$ binary classification subproblems, using the label ordering to sharpen class boundaries instead of treating severity levels as independent categories.

What would settle it

A patient-exclusive five-fold split of the collected dataset, keeping all images of each patient in a single fold, would settle the matter: if accuracy falls materially below 95.11%, the current split is leaking identity information.

Watch

Extended reading notes

Core claim

The central claim is that bilateral symmetry of the human back is a usable visual signal for scoliosis severity and that explicit symmetry matching plus ordinal regression turns that signal into accurate severity labels from natural images. The network takes the original back photo and its horizontal flip, extracts features with a shared visual attention backbone, fuses the two feature maps by concatenation-convolution followed by self-attention to model symmetric correspondence, and then sends the fused features through an ordinal regression head that learns $K-1$ binary classifiers, each asking whether the true severity rank exceeds $k$. The predicted level is obtained by summing the binary answers. With this design the paper reports state-of-the-art results on its collected dataset, including near-perfect recall for the severe class, and concludes that the approach is a viable economic screening tool.

Load-bearing premise

The evaluation's random five-way split of images assumes that the same patient's photographs never appear in both the training and test folds; if they do, the reported accuracy is inflated by the model recognizing familiar patients rather than generalizing to new backs.

Editorial extensions

If this is right

  • A radiation-free screening pipeline could be built from ordinary back photographs, lowering the barrier for mass or home-based checks of adolescent scoliosis.
  • The dual-path symmetry-aware design could transfer to other bilateral medical surfaces where asymmetry signals disease, such as facial palsy or limb swelling.
  • Because the ordinal head combines binary answers into a rank, the same network can report both coarse and fine severity levels through two parallel branches sharing the backbone.
  • At the reported 95.11% general accuracy, the model would misplace roughly 5% of images in a four-level task, with most remaining confusion between the minor and moderate classes rather than across distant levels.

Reading between the lines

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

  • The headline accuracies may be optimistic because the split is random at the image level, not the patient level; since the dataset has 1,898 images from only 1,067 patients, the same patient can appear in both training and test folds, and identity memorization could inflate the numbers. A patient-exclusive split would be the natural stress test.
  • The symmetry mismatch between original and flipped feature maps could be turned into a continuous asymmetry score that regresses the Cobb angle directly, something the paper explicitly says its current method cannot do because it only predicts severity ranges.
  • Two spine surgeons were asked to grade photographs without the aid of corresponding X-rays; a real deployment would need to define how the tool is used and what follow-up it triggers, since the paper does not address referral workflows.
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 paper proposes a dual-path deep network for estimating scoliosis severity from natural (non-radiographic) back images. Two modules are introduced: a symmetric feature matching module (SFMM) that compares features of an image and its horizontal flip, and an ordinal regression head (ORH) that converts the multi-class severity estimation into K-1 binary classification sub-problems. The method is evaluated on a newly collected dataset of 1,898 natural back images from 1,067 patients, with five-fold cross-validation, and is reported to achieve 95.11% accuracy for general severity levels, 81.46% for fine-grained levels, and to outperform state-of-the-art methods and human experts. The paper includes ablations, statistical analyses, and visualizations supporting the contributions.

Significance. If the reported results hold, the work would be a useful step toward low-cost, wide-range scoliosis screening using ordinary photographs, avoiding radiation exposure. The SFMM gives a principled way to exploit the symmetric/asymmetric structure of the back, and the ORH appropriately respects the ordinal nature of severity levels. The paper is also commendable for collecting a natural-image dataset paired with X-ray Cobb-angle labels and for including human expert comparisons, ablation studies, and per-fold results. However, the evaluation protocol has a load-bearing weakness: the random split is at the image level, not the patient level, so identity leakage could substantially inflate the headline accuracies. The lack of confidence intervals and the duplicated human-comparison figures further weaken the central claims as currently presented.

major comments (4)
  1. [Sections 4.1.1 and 4.1.2] The dataset contains 1,898 images from only 1,067 patients, and Section 4.1.2 states that the data are "randomly divided into five folds" without any mention of patient-level stratification. With an average of about 1.78 images per patient, a random image-level split will with high probability place images from the same patient in both the training and test folds. This allows the model to memorize patient-specific appearance (clothing, background, skin texture) rather than generalizable asymmetry cues, inflating the reported 95.11% and 81.46% accuracies and the SOTA comparison in Table 2. The authors should re-run the experiments using a patient-level split, or at minimum report the fraction of patients whose images appear in both training and test folds and analyze the impact on the results.
  2. [Table 2 and Table 5] The reported accuracies are averages over five folds without confidence intervals or significance tests. The margin over TransNeXt-S is only 0.53 percentage points (95.11% vs 94.58%), while per-fold results in Table 5 vary from 93.16% to 97.88%. Without a paired per-fold statistical test (e.g., Wilcoxon signed-rank test or corrected paired t-test) or confidence intervals, the claim of state-of-the-art performance is not statistically supported. Please add the missing statistical measures or temper the claim accordingly.
  3. [Section 4.3, Figures 4 and 5] The confusion matrices shown in Figures 4 and 5 appear to be identical, despite the fact that the two figures are supposed to represent different tasks: general severity level estimation vs fine-grained severity level estimation. Since the reported accuracies for these two tasks differ substantially (95.11% vs 81.46%), the duplicated figures are internally inconsistent and cannot be correct as presented. This undermines the human-comparison experiment and the claim of outperforming human performance. The authors should provide the correct figures and describe the human evaluation protocol more rigorously, including inter-observer variability and results across more than one fold.
  4. [Section 4.2] The Spinecube and ScolioNets baselines are reimplemented by the authors because the original code is not released. The text states that "we utilize the original settings in their code or papers" and also that "for a fair comparison, the networks of these methods are trained up to the same 610 epochs as our method." These two statements are in tension: if the original methods used fewer training epochs, forcing 610 epochs may disadvantage them (e.g., through overfitting), or conversely give them an advantage. The reimplementation fidelity is not validated. Please clarify the exact training protocol, and if possible, compare against the original papers' reported metrics on comparable data, or release the reimplementation code to allow verification.
minor comments (5)
  1. [Table 4] The text states that "when lambda_general is higher, the model performs better in estimating the general severity level," but the 2:1 ratio yields 95.02% accuracy, which is lower than the 1:1 ratio's 95.11%. Please correct the narrative or explain why the 1:1 ratio is considered optimal despite this apparent contradiction.
  2. [Section 4.3] The human comparison is performed on only the fifth fold and with only two experts, which limits the generality of the "outperforms human performance" claim. Adding more experts and reporting inter-observer agreement would strengthen this comparison.
  3. [Equation (3)] The notation Y_k1 and Y_k2 is used but the second component Y_k2 is never formally introduced; please define both components of the label vector for clarity.
  4. [Section 4.2] The FLOPs and #Params reported for the proposed method include both the general and fine-grained branches, while the comparison methods are implemented as single-task models. For a fair complexity comparison, please also report the complexity of a single branch of the proposed network.
  5. [Section 4.6] There is a typo in the sentence "we use a popular Grad-CAM [38] technique to generate heatmaps for general severity level estimation branch of our method"; the phrase "for general severity level estimation branch" should read "for the general severity level estimation branch."

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the reported accuracies are empirical results on held-out folds against external baselines, not derivations from the method's own definitions.

full rationale

The paper makes no theoretical derivation whose conclusion is equivalent to an input by construction. Its central claims are empirical: a dual-path network with a symmetric feature matching module and an ordinal regression head is trained and tested on a constructed dataset under five-fold cross-validation, with results compared against eleven external baselines and two human experts. The symmetry and ordinal-relationship observations in Sec. 1 are design motivations, not definitions that force the outcome. Ablations in Table 3 measure actual accuracy differences, and the ordinal regression formulation in Eqs. (3)-(5) is a standard label encoding that does not smuggle in the test labels. The self-citations (refs 41-44) appear only in a motivational sentence about deep learning and are not load-bearing. The loss-weight ratio in Table 4 is a small post-hoc tuning choice, and the chosen 1:1 ratio is not renamed as a prediction. A separate concern about image-level rather than patient-level splitting could threaten external validity, but that is an evaluation-leakage risk, not a circularity step: it does not make the reported test accuracy equal to a training input by definition. Therefore no circularity is present.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

All central results are empirical and depend on the private dataset and training setup. The main ledger items are manually chosen hyperparameters, plus domain assumptions about back asymmetry, Cobb-angle labels, image-level cross-validation, baseline reimplementation, and population representativeness. There are no invented physical entities; SFMM and ORH are architectural modules validated only by internal ablation.

free parameters (3)
  • Loss balance ratio lambda_general:lambda_fine = 0.5 : 0.5
    Selected from three ratios in Table 4; changes both Acc and MAE, e.g. 1:2 gives fine-grained 81.93% vs 81.46%, and 2:1 gives general 95.02% vs 95.11%.
  • Learning rate = 1e-4
    Set in Sec. 4.1.2 with cosine schedule and warmup; no sensitivity analysis reported.
  • Training epochs = 610
    All methods trained for the same 610 epochs; chosen to convergence but no early-stopping analysis.
assumptions (5)
  • domain assumption Image-level random five-fold split yields independent training and test samples.
    1,898 images come from 1,067 patients; Sec. 4.1.2 and Table 1 describe image-level folds without patient containment. If the same patient appears in both training and test, accuracy is inflated.
  • domain assumption Back asymmetry visible in natural images is a reliable ordinal signal for Cobb-angle severity levels.
    This is the motivation in Sec. 1 and Fig. 1. The paper provides no quantitative analysis of confounding from pose, clothing, lighting, body mass, or scapular asymmetry unrelated to scoliosis.
  • domain assumption Expert-measured Cobb angle from each paired X-ray is the correct ground truth for the corresponding natural image.
    Sec. 4.1.1 says each natural image has a corresponding X-ray and manually measured Cobb angle; assumes the photo and X-ray correspond in time and posture and that measurement error is small relative to severity bins.
  • domain assumption Reimplementations of Spinecube and ScolioNets from their papers are faithful.
    Sec. 4.2 states the code for these baselines is not released, so the authors implemented them 'based on their papers.' Any mismatch in architecture or training details can bias the SOTA comparison.
  • domain assumption The hospital-derived dataset represents a general screening population.
    The dataset has 370 severe cases among 1,898 samples and a mean Cobb angle of 28.88 degrees, which is not representative of an adolescent screening population where severe cases are rare. Accuracy on this dataset may not transfer to community screening.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Symmetric Perception and Ordinal Regression for Detecting Scoliosis Natural Image." pith.science (2026). https://pith.science/paper/XXNEMT2T

@misc{pith2026241115799,
  author       = {Pith},
  title        = {Pith review of: Symmetric Perception and Ordinal Regression for Detecting Scoliosis Natural Image},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XXNEMT2T}},
  note         = {Machine review of arXiv:2411.15799}
}
read the original abstract

Scoliosis is one of the most common diseases in adolescents. Traditional screening methods for the scoliosis usually use radiographic examination, which requires certified experts with medical instruments and brings the radiation risk. Considering such requirement and inconvenience, we propose to use natural images of the human back for wide-range scoliosis screening, which is a challenging problem. In this paper, we notice that the human back has a certain degree of symmetry, and asymmetrical human backs are usually caused by spinal lesions. Besides, scoliosis severity levels have ordinal relationships. Taking inspiration from this, we propose a dual-path scoliosis detection network with two main modules: symmetric feature matching module (SFMM) and ordinal regression head (ORH). Specifically, we first adopt a backbone to extract features from both the input image and its horizontally flipped image. Then, we feed the two extracted features into the SFMM to capture symmetric relationships. Finally, we use the ORH to transform the ordinal regression problem into a series of binary classification sub-problems. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods as well as human performance, which provides a promising and economic solution to wide-range scoliosis screening. In particular, our method achieves accuracies of 95.11% and 81.46% in estimation of general severity level and fine-grained severity level of the scoliosis, respectively.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 60 canonical work pages

  1. [1]

    In: European Conference on Computer Vision, pp

    Atadjanov, I.R., Lee, S.: Reflection symmetry detection via appear- ance of structure descriptor. In: European Conference on Computer Vision, pp. 3–18. Springer (2016)

    work page 2016

  2. [2]

    Computational and Mathe- matical Methods in Medicine 2022(1), 3796,202 (2022)

    Chen, P., Zhou, Z., Yu, H., Chen, K., Yang, Y.: Computerized- assisted scoliosis diagnosis based on faster r-cnn and resnet for the classification of spine x-ray images. Computational and Mathe- matical Methods in Medicine 2022(1), 3796,202 (2022)

    work page 2022

  3. [3]

    IEEE Transactions on Medical Imaging 39(2), 387–399 (2019)

    Chen, Y., Gao, Y., Li, K., Zhao, L., Zhao, J.: Vertebrae identifi- cation and localization utilizing fully convolutional networks and a hidden markov model. IEEE Transactions on Medical Imaging 39(2), 387–399 (2019)

    work page 2019

  4. [4]

    Instructional Course Lecture (1948)

    Cobb, J.: Outline for the study of scoliosis. Instructional Course Lecture (1948)

    work page 1948

  5. [5]

    In: International Conference on Pattern Recognition, pp

    Cornelius, H., Loy, G.: Detecting rotational symmetry under affine projection. In: International Conference on Pattern Recognition, pp. 292–295. IEEE (2006)

    work page 2006

  6. [6]

    In: IEEE Conference on Com- puter Vision and Pattern Recognition, pp

    Dong, X., Bao, J., Chen, D., Zhang, W., Yu, N., Yuan, L., Chen, D., Guo, B.: Cswin transformer: A general vision transformer back- bone with cross-shaped windows. In: IEEE Conference on Com- puter Vision and Pattern Recognition, pp. 12,124–12,134. IEEE (2022)

    work page 2022

  7. [7]

    In: ACM International Conference on Multimedia, pp

    Foteinopoulou, N.M., Patras, I.: Learning from label relationships in human affect. In: ACM International Conference on Multimedia, pp. 80–89. ACM (2022)

    work page 2022

  8. [8]

    Plos One 17(5), e0267,851 (2022)

    Fraiwan, M., Audat, Z., Fraiwan, L., Manasreh, T.: Using deep transfer learning to detect scoliosis and spondylolisthesis from x- ray images. Plos One 17(5), e0267,851 (2022)

    work page 2022

Show all 61 references
  1. [9]

    In: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Fu, H., Gong, M., Wang, C., Batmanghelich, K., Tao, D.: Deep ordinal regression network for monocular depth estimation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 2002–2011. IEEE (2018)

    2018

  2. [10]

    In: IEEE International Conference on Computer Vision, pp

    Funk, C., Liu, Y.: Beyond planar symmetry: Modeling human perception of reflection and rotation symmetries in the wild. In: IEEE International Conference on Computer Vision, pp. 793–803 (2017)

    2017

  3. [11]

    European Spine Journal 28, 951–960 (2019)

    Galbusera, F., Niemeyer, F., Wilke, H.J., Bassani, T., Casaroli, G., Anania, C., Costa, F., Brayda-Bruno, M., Sconfienza, L.M.: Fully automated radiological analysis of spinal disorders and de- formities: a deep learning approach. European Spine Journal 28, 951–960 (2019)

    2019

  4. [12]

    Computational Visual Media 9(4), 733–752 (2023)

    Guo, M.H., Lu, C.Z., Liu, Z.N., Cheng, M.M., Hu, S.M.: Visual attention network. Computational Visual Media 9(4), 733–752 (2023)

    2023

  5. [13]

    Medical Physics 48(4), 1571–1583 (2021)

    He, Z., Wang, Y., Qin, X., Yin, R., Qiu, Y., He, K., Zhu, Z.: Classi- fication of neurofibromatosis-related dystrophic or nondystrophic scoliosis based on image features using bilateral cnn. Medical Physics 48(4), 1571–1583 (2021)

    2021

  6. [14]

    IEEE Transactions on Medical Imaging 41(7), 1610–1624 (2022)

    Huang, Z., Zhao, R., Leung, F.H., Banerjee, S., Lee, T.T.Y., Yang, D., Lun, D.P., Lam, K.M., Zheng, Y.P., Ling, S.H.: Joint spine segmentation and noise removal from ultrasound volume projec- tion images with selective feature sharing. IEEE Transactions on Medical Imaging 41(7...

    2022

  7. [15]

    In: International Conference on Machine Learning, pp

    Ioffe, S., Szegedy, C.: Batch normalization: Accelerating deep net- work training by reducing internal covariate shift. In: International Conference on Machine Learning, pp. 448–456 (2015)

    2015

  8. [16]

    IEEE Transactions on Image Processing 15(8), 2198–2207 (2006)

    Keller, Y., Shkolnisky, Y.: A signal processing approach to sym- metry detection. IEEE Transactions on Image Processing 15(8), 2198–2207 (2006)

    2006

  9. [17]

    The Spine Journal 21(6), 980–987 (2021)

    Kokabu, T., Kanai, S., Kawakami, N., Uno, K., Kotani, T., Suzuki, T., Tachi, H., Abe, Y., Iwasaki, N., Sudo, H.: An algorithm for using deep learning convolutional neural networks with three di- mensional depth sensor imaging in scoliosis detection. The Spine Journal 21(6), 98...

    2021

  10. [18]

    Journal of Children’s Orthopaedics 7(1), 3–9 (2013)

    Konieczny, M.R., Senyurt, H., Krauspe, R.: Epidemiology of ado- lescent idiopathic scoliosis. Journal of Children’s Orthopaedics 7(1), 3–9 (2013)

    2013

  11. [19]

    Polish Orthopedics and Trauma- tology 79, 118–122 (2014)

    Korbel, K., Kozinoga, M., Stoli ´nski, L., Kotwicki, T.: Scoliosis re- search society (srs) criteria and society of scoliosis orthopaedic and rehabilitation treatment (sosort) 2008 guidelines in non-operative treatment of idiopathic scoliosis. Polish Orthopedics and Trauma- to...

    2014

  12. [20]

    arXiv preprint arXiv:1211.5355 (2012)

    Kundu, R., Chakrabarti, A., Lenka, P.K.: Cobb angle mea- surement of scoliosis with reduced variability. arXiv preprint arXiv:1211.5355 (2012)

    2012 arXiv

  13. [21]

    In: International Conference on Learning Representations (2017)

    Larsson, G., Maire, M., Shakhnarovich, G.: Fractalnet: Ultra-deep neural networks without residuals. In: International Conference on Learning Representations (2017)

    2017

  14. [22]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 32(9), 1659–1672 (2009) Symmetric Perception and Ordinal Regression for Detecting Scoliosis Natural Image 11

    Lee, S., Liu, Y.: Skewed rotation symmetry group detection. IEEE Transactions on Pattern Analysis and Machine Intelligence 32(9), 1659–1672 (2009) Symmetric Perception and Ordinal Regression for Detecting Scoliosis Natural Image 11

    2009

  15. [23]

    In: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Li, C., Liu, Q., Liu, J., Lu, H.: Learning ordinal discriminative features for age estimation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 2570–2577. IEEE (2012)

    2012

  16. [24]

    In: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Li, W., Huang, X., Lu, J., Feng, J., Zhou, J.: Learning proba- bilistic ordinal embeddings for uncertainty-aware regression. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 13,896–13,905. IEEE (2021)

    2021

  17. [25]

    Pattern Recognition Letters 18(5), 433–443 (1997)

    Lin, H.C., Wang, L.L., Yang, S.N.: Extracting periodicity of a regu- lar texture based on autocorrelation functions. Pattern Recognition Letters 18(5), 433–443 (1997)

    1997

  18. [26]

    In: IEEE International Conference on Computer Vision, pp

    Lin, W., Wu, Z., Chen, J., Huang, J., Jin, L.: Scale-aware mod- ulation meet transformer. In: IEEE International Conference on Computer Vision, pp. 6015–6026. IEEE (2023)

    2023

  19. [27]

    In: Inter- national Conference on Medical Image Computing and Computer- Assisted Intervention, pp

    Lin, Y., Liu, L., Ma, K., Zheng, Y.: Seg4reg+: Consistency learning between spine segmentation and cobb angle regression. In: Inter- national Conference on Medical Image Computing and Computer- Assisted Intervention, pp. 490–499. Springer (2021)

    2021

  20. [28]

    In: International Workshop and Challenge on Computational Methods and Clinical Applications for Spine Imaging, pp

    Lin, Y., Zhou, H.Y., Ma, K., Yang, X., Zheng, Y.: Seg4reg net- works for automated spinal curvature estimation. In: International Workshop and Challenge on Computational Methods and Clinical Applications for Spine Imaging, pp. 69–74. Springer (2020)

    2020

  21. [29]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 26(3), 354–371 (2004)

    Liu, Y., Collins, R.T., Tsin, Y.: A computational model for periodic pattern perception based on frieze and wallpaper groups. IEEE Transactions on Pattern Analysis and Machine Intelligence 26(3), 354–371 (2004)

    2004

  22. [30]

    In: IEEE International Conference on Computer Vision, pp

    Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., Guo, B.: Swin transformer: Hierarchical vision transformer using shifted windows. In: IEEE International Conference on Computer Vision, pp. 10,012–10,022. IEEE (2021)

    2021

  23. [31]

    In: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Liu, Z., Mao, H., Wu, C.Y., Feichtenhofer, C., Darrell, T., Xie, S.: A convnet for the 2020s. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 11,976–11,986. IEEE (2022)

    2022

  24. [32]

    In: International Conference on Learning Repre- sentations (2017)

    Loshchilov, I., Hutter, F.: Sgdr: Stochastic gradient descent with warm restarts. In: International Conference on Learning Repre- sentations (2017)

    2017

  25. [33]

    In: International Conference on Learning Representations (2019)

    Loshchilov, I., Hutter, F.: Decoupled weight decay regularization. In: International Conference on Learning Representations (2019)

    2019

  26. [34]

    In: European Conference on Computer Vision, pp

    Loy, G., Eklundh, J.O.: Detecting symmetry and symmetric con- stellations of features. In: European Conference on Computer Vision, pp. 508–521. Springer (2006)

    2006

  27. [35]

    Biochemia Medica 22(3), 276–282 (2012)

    McHugh, M.L.: Interrater reliability: the kappa statistic. Biochemia Medica 22(3), 276–282 (2012)

    2012

  28. [36]

    In: Ad- vances in Neural Information Processing Systems, pp

    Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., Desmaison, A., Kopf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chil- amkurthy, S., Steiner, B., Fang, L., Bai, J., Chintala, S.: Pytorch: An imperat...

    2019

  29. [37]

    In: IEEE International Conference on Computer Vision, pp

    Prasad, V.S.N., Davis, L.S.: Detecting rotational symmetries. In: IEEE International Conference on Computer Vision, pp. 954–961. IEEE (2005)

    2005

  30. [38]

    In: IEEE International Conference on Computer Vision, pp

    Selvaraju, R.R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., Batra, D.: Grad-cam: Visual explanations from deep networks via gradient-based localization. In: IEEE International Conference on Computer Vision, pp. 618–626. IEEE (2017)

    2017

  31. [39]

    In: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Seo, A., Kim, B., Kwak, S., Cho, M.: Reflection and rotation sym- metry detection via equivariant learning. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 9539–9548 (2022)

    2022

  32. [40]

    In: IEEE International Conference on Computer Vision, pp

    Seo, A., Shim, W., Cho, M.: Learning to discover reflection sym- metry via polar matching convolution. In: IEEE International Conference on Computer Vision, pp. 1285–1294 (2021)

    2021

  33. [41]

    International Journal of Computer Vision 129(2), 321–340 (2021)

    Shao, Z., Liu, Z., Cai, J., Ma, L.: J ˆaa-net: Joint facial action unit detection and face alignment via adaptive attention. International Journal of Computer Vision 129(2), 321–340 (2021)

    2021

  34. [42]

    IEEE Transactions on Image Processing 32, 3354–3366 (2023)

    Shao, Z., Zhou, Y., Cai, J., Zhu, H., Yao, R.: Facial action unit detection via adaptive attention and relation. IEEE Transactions on Image Processing 32, 3354–3366 (2023)

    2023

  35. [43]

    IEEE Transactions on Image Processing 30, 4610–4621 (2021)

    Shao, Z., Zhu, H., Tang, J., Lu, X., Ma, L.: Explicit facial expression transfer via fine-grained representations. IEEE Transactions on Image Processing 30, 4610–4621 (2021)

    2021

  36. [44]

    International Journal of Computer Vision (2024)

    Shao, Z., Zhu, H., Zhou, Y., Xiang, X., Liu, B., Yao, R., Ma, L.: Facial action unit detection by adaptively constraining self- attention and causally deconfounding sample. International Journal of Computer Vision (2024)

    2024

  37. [45]

    In: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Shi, D.: Transnext: Robust foveal visual perception for vision trans- formers. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 17,773–17,783. IEEE (2024)

    2024

  38. [46]

    In: International Conference on Information Processing in Medical Imaging, pp

    Sun, H., Zhen, X., Bailey, C., Rasoulinejad, P., Yin, Y., Li, S.: Direct estimation of spinal cobb angles by structured multi-output regression. In: International Conference on Information Processing in Medical Imaging, pp. 529–540. Springer (2017)

    2017

  39. [47]

    In: International Conference on Machine Learning, pp

    Tan, M., Le, Q.: Efficientnet: Rethinking model scaling for convo- lutional neural networks. In: International Conference on Machine Learning, pp. 6105–6114. PMLR (2019)

    2019

  40. [48]

    In: International Conference on Machine Learning, pp

    Touvron, H., Cord, M., Douze, M., Massa, F., Sablayrolles, A., J´egou, H.: Training data-efficient image transformers & distilla- tion through attention. In: International Conference on Machine Learning, pp. 10,347–10,357. PMLR (2021)

    2021

  41. [49]

    In: Advances in Neural Information Processing Systems, pp

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, L., Polosukhin, I.: Attention is all you need. In: Advances in Neural Information Processing Systems, pp. 5998–

  42. [50]

    arXiv preprint arXiv:2307.09004 (2023)

    Wang, J., Cheng, Y., Chen, J., Chen, T., Chen, D., Wu, J.: Ord2seq: Regard ordinal regression as label sequence prediction. arXiv preprint arXiv:2307.09004 (2023)

    2023 arXiv

  43. [51]

    In: IEEE International Conference on Computer Vision, pp

    Wang, W., Xie, E., Li, X., Fan, D.P., Song, K., Liang, D., Lu, T., Luo, P., Shao, L.: Pyramid vision transformer: A versatile backbone for dense prediction without convolutions. In: IEEE International Conference on Computer Vision, pp. 568–578. IEEE (2021)

    2021

  44. [52]

    Pattern Recognition 47(4), 1764–1776 (2014)

    Wang, Z., Fu, L., Li, Y.: Unified detection of skewed rotation, reflection and translation symmetries from affine invariant contour features. Pattern Recognition 47(4), 1764–1776 (2014)

    2014

  45. [53]

    The Lancet 371(9623), 1527–1537 (2008)

    Weinstein, S.L., Dolan, L.A., Cheng, J.C., Danielsson, A., Mor- cuende, J.A.: Adolescent idiopathic scoliosis. The Lancet 371(9623), 1527–1537 (2008)

    2008

  46. [54]

    New England Journal of Medicine 369(16), 1512–1521 (2013)

    Weinstein, S.L., Dolan, L.A., Wright, J.G., Dobbs, M.B.: Effects of bracing in adolescents with idiopathic scoliosis. New England Journal of Medicine 369(16), 1512–1521 (2013)

    2013

  47. [55]

    In: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Xie, S., Girshick, R., Doll ´ar, P., Tu, Z., He, K.: Aggregated residual transformations for deep neural networks. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1492–1500. IEEE (2017)

    2017

  48. [56]

    Communications Biology 2(1), 390 (2019)

    Yang, J., Zhang, K., Fan, H., Huang, Z., Xiang, Y., Yang, J., He, L., Zhang, L., Yang, Y., Li, R., et al.: Development and validation of deep learning algorithms for scoliosis screening using back images. Communications Biology 2(1), 390 (2019)

    2019

  49. [57]

    Beijing China: People’s Health Publishing House (2015)

    Zhang, H., Sucato, D., Richards, B.: Principles of Surgical Plan for Adolescent Idiopathic Scoliosis. Beijing China: People’s Health Publishing House (2015)

    2015

  50. [58]

    International Journal of Biomedical Imaging 2017 (2017)

    Zhang, J., Li, H., Lv, L., Zhang, Y., et al.: Computer-aided cobb measurement based on automatic detection of vertebral slopes us- ing deep neural network. International Journal of Biomedical Imaging 2017 (2017)

    2017

  51. [59]

    JAMA Network Open 6(8), e2330,617– e2330,617 (2023)

    Zhang, T., Zhu, C., Zhao, Y., Zhao, M., Wang, Z., Song, R., Meng, N., Sial, A., Diwan, A., Liu, J., et al.: Deep learning model to clas- sify and monitor idiopathic scoliosis in adolescents using a single smartphone photograph. JAMA Network Open 6(8), e2330,617– e2330,617 (2023)

    2023

  52. [60]

    In: IEEE Conference on Computer Vision and Pattern Recognition, pp

    Zhao, P., Quan, L.: Translation symmetry detection in a fronto- parallel view. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1009–1016. IEEE (2011)

    2011

  53. [6008]

    Curran Associates, Inc. (2017)

    2017

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.