Pith. sign in

REVIEW 4 major objections 7 minor 2 cited by

MedConv: Convolutions Beat Transformers on Long-Tailed Bone Density Prediction

T0 review · 4 major / 7 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read MedConv, a 3D ResNet-50 with balanced cross-entropy loss and post-hoc logit adjustment, achieves the highest accuracy and ROC AUC on the AustinSpine CT dataset while using less compute than transformer baselines.

desk verdict The headline accuracy is not interpretable because Table II's sensitivity and specificity are internally inconsistent, and the test-set tuning makes the result a selected maximum rather than a prediction. read the letter →

arxiv 2502.00631 v2 pith:DB622UF2 submitted 2025-02-02 cs.CV

Xuyin Qi , Zeyu Zhang , Huazhan Zheng , Mingxi Chen , Numan Kutaiba , Ruth Lim , Cherie Chiang , Zi En Tham
show 14 more authors
Xuan Ren Wenxin Zhang Lei Zhang Hao Zhang Wenbing Lv Guangzhen Yao Renda Han Kangsheng Wang Mingyuan Li Hongtao Mao Yu Li Zhibin Liao Yang Zhao Minh-Son To
This is my paper · ORCID
classification cs.CV
keywords bonedensitypredictionT-scoreCTimaginglong-taileddistributionclassimbalance3Dconvolutionalneuralnetworkbalancedcross-entropypost-hoclogitadjustment
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that a straightforward 3D convolutional network, MedConv, can outperform transformer-based models for predicting bone density (T-score) from spinal CT scans. On the AustinSpine dataset it reports 65.38% accuracy and 79.34% ROC AUC, beating all transformer and CNN baselines in its comparison while using lower computational cost. The improvement is attributed to two mechanisms: balanced cross-entropy loss that up-weights minority classes, and post-hoc logit adjustment that amplifies minority-class logits at inference. If the claim holds, resource-efficient CNN models are sufficient for CT-based bone density screening, and long-tailed class imbalance can be handled without complex architectures.

What carries the argument

The load-bearing object is the 3D ResNet-50 backbone combined with two balancing mechanisms: the Bal-CE loss, which weights each training sample by the inverse frequency of its class, and a post-hoc logit adjustment that applies temperature scaling ($\tau_1=1$ for the majority class, $\tau_2=0.5$ for the minority class) to the final logits. The 3D convolutions capture volumetric context in CT scans, the loss weighting counters the long-tailed label distribution during training, and the logit adjustment reshapes the predicted probabilities at inference to favor minority classes. Together these parts let a standard CNN match or exceed transformer performance at a fraction of the compute.

What would settle it

Run the MedConv pipeline with $\tau_1$ and $\tau_2$ chosen only on a held-out validation fold, then evaluate once on the test set. If the resulting accuracy and ROC AUC do not exceed the transformer baselines in Table II, the central claim that convolutions beat transformers on this task is falsified.

Watch

Extended reading notes

Core claim

MedConv—a 3D ResNet-50 backbone fed with L1 vertebra segmentations and trained with inverse-frequency weighted cross-entropy plus a post-hoc logit adjustment ($\tau_1=1$, $\tau_2=0.5$)—reaches 65.38% accuracy and 79.34% ROC AUC on the AustinSpine test set, outscoring all reported transformer baselines as well as other CNN variants. The central discovery is that this conventional convolutional pipeline is enough to beat transformers on this long-tailed CT classification task, and that the class-imbalance techniques, rather than architectural novelty, account for most of the gain. The paper reports that the balanced loss raises ResNet-50 accuracy by 6.41 percentage points, and that lowering $\tau_2$ from 1.0 to 0.5 lifts accuracy from 64.10% to 65.38%, indicating that logit rebalancing is the final performance-critical step.

Load-bearing premise

The reported test accuracy is treated as an unbiased estimate of model performance, which requires that the hyperparameters $\tau_1$ and $\tau_2$ were not selected using the test set; the paper selects them from test-set ablations, so the final numbers are a selected maximum rather than an independent prediction.

Editorial extensions

If this is right

  • If the claim is right, 3D CNNs with rebalanced losses become a strong default for CT-based T-score classification, reducing the need for transformer backbones in this task.
  • The reported gains imply that model selection in long-tailed medical imaging should first address loss rebalancing and logit calibration before changing architecture.
  • The lower computational cost makes deployment on portable or edge devices feasible for bone density screening.
  • The AustinSpine benchmark and the TotalSegmentator-based preprocessing pipeline can serve as a standard evaluation setup for future bone density prediction models.

Reading between the lines

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

  • Because $\tau_1$ and $\tau_2$ were chosen from test-set ablations, the reported 65.38% accuracy is likely an optimistic estimate of out-of-sample performance; a validation-based tuning loop would give a fairer number.
  • The 'convolutions beat transformers' conclusion may be sensitive to the specific transformer checkpoints and input preprocessing used, so it should be re-tested across more transformer variants and external datasets.
  • A direct extension would be to apply the same Bal-CE plus logit-adjustment recipe to other long-tailed 3D medical imaging tasks, such as fracture detection or liver steatosis grading, to see whether gains of this size reproduce.
  • Reporting confidence intervals or standard deviations across runs would make the benchmark more trustworthy, given the small dataset of 389 patients.
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 / 7 minor

Summary. The manuscript introduces MedConv, a 3D ResNet-50 with a balanced cross-entropy (Bal-CE) loss and a post-hoc logit adjustment, for classifying CT-based bone density T-scores into three WHO categories (Normal, Osteopenia, Osteoporosis). On the authors' private AustinSpine dataset (389 patients), it reports 65.38% accuracy and 79.34% ROC AUC for MedConv, claimed to exceed all 2D CNN and transformer baselines, and it presents ablations over two temperature hyperparameters (tau_1, tau_2) and over two segmentation tools (CTSpine1K, TotalSegmentator). The paper's thesis is that a convolutional model with rebalancing techniques is a lower-complexity alternative to transformers for this task.

Significance. The question of whether simple 3D CNNs can match or surpass transformer-based models on imbalanced volumetric medical imaging is practically relevant, and the paper offers a clinically curated dataset and a head-to-head comparison of segmentation backbones. If the reported numbers were trustworthy, MedConv would serve as a useful efficient baseline for T-score screening. However, the experimental evidence in Tables II, V, and VI is unreliable due to internal metric inconsistencies and test-set-based hyperparameter selection, so the central claims cannot currently be evaluated.

major comments (4)
  1. [V-A, Table II] In every row of Table II, Sensitivity is numerically equal to Accuracy and Specificity equals 0.5*Accuracy + 0.5 (e.g., MedConv: 65.38/65.38/82.69; ViT: 33.54/33.54/66.77). For any standard binary confusion matrix with a nonempty negative class, Sensitivity = Accuracy forces Specificity = Accuracy, so the reported affine relation is impossible. The task is three-class, and the paper never defines how scalar sensitivity/specificity are aggregated over classes. Thus the headline numbers are not genuine, interpretable performance estimates, and the claimed improvements over prior SOTA are unfalsifiable as reported. Please provide the per-class confusion matrices and the exact aggregation formulas.
  2. [V-B, Tables V and VI] The final MedConv configuration tau_1=1, tau_2=0.5 is selected because it maximizes Accuracy in ablations computed on the test set (Table VI shows that tau_2=0.5 gives the maximum 65.38% accuracy; the text describes tau_1=1 as 'optimal' and tau_2=0.5 as giving the best accuracy). No validation set and no nested cross-validation are described anywhere. The 65.38%/79.34% result is therefore a selection over hyperparameters, not an independent test estimate, and the comparison with baselines is biased. Moreover, the test set size implied by 65.38% is 78 patients, so the 1.28 percentage-point gap over the best non-MedConv row is about one patient; no confidence intervals or significance tests are given.
  3. [III-C] The post-hoc logit adjustment formula yhat_i = e^{z_i/tau_1}/(e^{z_i/tau_1}+e^{z_j/tau_2}) is a two-class softmax, with z_j and j left undefined and no normalization over the three classes of the task. As written, it cannot be applied to the reported 3-class problem. Similarly, the Bal-CE loss in Section III-B is given in binary cross-entropy form for a single sample; the paper does not state the multiclass softmax cross-entropy version with class weights. The method section is therefore incomplete for the actual experimental setup.
  4. [V-A, Table II] The claim that 'convolutions beat transformers' is not supported by the baseline quality: ViT+pretrain reaches only 33.54% accuracy, essentially chance for a three-class problem, while JointViT variants also stay near chance; no evidence is provided that these baselines were trained to convergence or with adequate hyperparameters. In addition, no comparison of computational cost (parameters, FLOPs, latency) is reported in Table II or elsewhere, although the abstract and conclusion advance a lower-complexity argument. A fair head-to-head would require competitive transformer training and an actual efficiency table.
minor comments (7)
  1. [IV-B] The heading 'Evaluation Matrices' should read 'Evaluation Metrics'.
  2. [I] The Introduction contains the typo 'adress' for 'address'.
  3. [Tables II-IV] Several model names are corrupted timm identifiers (e.g., 'resnet18.a1in1k', 'mobilenetv2100.rain1k', 'efficientnetb0.rain1k', 'resnext5032x4d.a1hin1k'); these should be standard names such as 'resnet18.a1_in1k'.
  4. [Figure 2] The caption says the model ensures 'accurate predictions for the L1 vertebra segmentation task,' but the paper addresses T-score classification, not segmentation.
  5. [Table I] The boundary definitions overlap (Osteoporosis '-4 to -2.5' and Osteopenia '-2.5 to -1'), and the paper does not state how T-scores exactly at a boundary are assigned.
  6. [IV-A] The dataset description provides only the total patient count (389); class counts, the train/validation/test split sizes, and the test set composition are missing, making the long-tailed distribution in Figure 4 unquantified.
  7. [Abstract / Table II] The abstract's claim of 'up to 21% improvement in accuracy' is not tied to a specific baseline in Table II; the closest row (resnet50.a1in1k+pretrain+balce) differs from MedConv by only 1.28 percentage points, so the magnitude of the claimed improvement should be recomputed and stated with its reference baseline.

Circularity Check

1 steps flagged · score 6.0 of 10

Reported MedConv accuracy is a test-set selection maximum: τ1 and τ2 are tuned on the test set and the best row is then advertised as the prediction.

  1. fitted input called prediction [Section V-B, 'Effect of τ2 on Model Performance' (Tables V and VI), with Section IV-B evaluation definition]
    "For a fair comparison, we evaluated each method's overall classification performance on the test set using accuracy and ROC AUC scores. ... The temperature parameters τ1 = 1 and τ2 = 0.5 are empirically chosen to achieve an effective balance. ... A significant improvement is observed at τ2 = 0.6 and τ2 = 0.5, where accuracy rises to 65.38% and F1 score reaches 66.37%."

    The ablation section explicitly searches the τ1 and τ2 grids to 'identify the optimal value' and then reports the best configuration's test-set metrics as MedConv's result. Since the test set is both the selection criterion for τ2 (Table VI: τ2=0.5 gives 65.38% accuracy, 79.34 AUC) and the evaluation set on which the final 65.38%/79.34% is advertised (Section IV-B), the headline 'prediction' is the maximum of the test-set metric over the tuned hyperparameters, i.e., a fitted/selected number, not an independent out-of-sample estimate. With no separate validation split described, every reported gain for MedConv over baselines is conditional on having picked τ1 and τ2 from the same test labels.

full rationale

The method-level derivation—3D ResNet-50 backbone, inverse-frequency weighted cross-entropy, and post-hoc temperature adjustment—is self-contained and not circular: the formulas are standard, and the comparison against transformer baselines is a direct head-to-head experiment on the private AustinSpine set. Self-citations (e.g., JointViT [13], SegReg [36]) appear as baselines or segmentation tools, not as load-bearing justification for MedConv's performance. The central circularity is statistical: the test set is used both to choose τ1 and τ2 (Section V-B, Tables V and VI) and to report the final metrics (Section IV-B), so the headline 65.38% accuracy / 79.34% AUC is the maximum of the test-metric surface over the tuned hyperparameters, not an out-of-sample estimate. Separately, Table II's exact identity Sensitivity = Accuracy and affine Specificity = 50% + Accuracy/2 in every row cannot arise from any standard confusion matrix with a nonempty negative class; this is a correctness/integrity concern rather than a circularity step, but it further prevents interpreting the advertised numbers.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim depends on two empirically tuned hyperparameters (tau_1, tau_2) selected on the test set, and on assumptions about the dataset's labeling, the segmentation tool, and the train/test split. No new physical or conceptual entity is introduced.

free parameters (2)
  • tau_1 = 1
    Chosen by scanning values on the test set (Table V); the best accuracy 64.10% is achieved at tau_1=1.
  • tau_2 = 0.5
    Chosen by scanning values on the test set (Table VI); best accuracy 65.38% is achieved at tau_2=0.5.
assumptions (3)
  • domain assumption T-score categories follow WHO thresholds (Table I)
    Ground truth labels are defined by WHO criteria; this is a standard medical classification, not proven in the paper.
  • domain assumption Segmentation of L1 vertebra using TotalSegmentator provides sufficient and unbiased input for T-score prediction
    The paper relies on the chosen segmentation tool's output as the model input; the segmentation quality is not independently validated for this task.
  • domain assumption The train/test split is random and patient-independent, and the test set is representative of the population
    No split methodology is described; performance claims depend on this assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MedConv: Convolutions Beat Transformers on Long-Tailed Bone Density Prediction." pith.science (2026). https://pith.science/paper/DB622UF2

@misc{pith2026250200631,
  author       = {Pith},
  title        = {Pith review of: MedConv: Convolutions Beat Transformers on Long-Tailed Bone Density Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DB622UF2}},
  note         = {Machine review of arXiv:2502.00631}
}
read the original abstract

Bone density prediction via CT scans to estimate T-scores is crucial, providing a more precise assessment of bone health compared to traditional methods like X-ray bone density tests, which lack spatial resolution and the ability to detect localized changes. However, CT-based prediction faces two major challenges: the high computational complexity of transformer-based architectures, which limits their deployment in portable and clinical settings, and the imbalanced, long-tailed distribution of real-world hospital data that skews predictions. To address these issues, we introduce MedConv, a convolutional model for bone density prediction that outperforms transformer models with lower computational demands. We also adapt Bal-CE loss and post-hoc logit adjustment to improve class balance. Extensive experiments on our AustinSpine dataset shows that our approach achieves up to 21% improvement in accuracy and 20% in ROC AUC over previous state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2502.00631 by the authors.

Figure 1
Figure 1. Visualization of segmentation results on CT images. The first column [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison between 3D ResNet and 2D ResNet architectures [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Architecture of the proposed MedConv model, based on a 3D ResNet [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Long-tailed distribution of T-score classifications within the Austin [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Experiment pipeline for evaluating segmentation methods and [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Ablation study showing the impact of different [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. SSS: Semi-Supervised SAM-2 with Efficient Prompting for Medical Imaging Segmentation

    cs.CV 2025-06 reject novelty 4.0 of 10

    SSS applies SAM-2 with a Discriminative Feature Enhancement mechanism and a physical-constraint sliding-window prompt generator, reporting Dice scores of 53.15 on BHSD and 89.34 to 91.21 on ACDC.

  2. A Novel Convolutional-Free Method for 3D Medical Imaging Segmentation

    eess.IV 2025-02 reject novelty 3.0 of 10

    The paper proposes a convolution-free transformer pipeline and a thick-to-thin joint loss but reports no experiments and no performance numbers.

Reference graph

Works this paper leans on

37 extracted references · 24 canonical work pages · cited by 2 Pith papers

  1. [1]

    Bone health and osteoporosis,

    Beatrice C Lupsa and Karl Insogna, “Bone health and osteoporosis,” Endocrinology and Metabolism Clinics , vol. 44, no. 3, pp. 517–530, 2015

    work page 2015

  2. [2]

    Deep learning for bone mineral density and t-score prediction from chest x-rays: A multicenter study,

    Yoichi Sato, Norio Yamamoto, Naoya Inagaki, Yusuke Iesaki, Takamune Asamoto, Tomohiro Suzuki, and Shunsuke Takahara, “Deep learning for bone mineral density and t-score prediction from chest x-rays: A multicenter study,” Biomedicines, vol. 10, no. 9, pp. 2323, 2022

    work page 2022

  3. [3]

    Advancing osteoporosis diagnosis and fracture prediction: role of quantitative ct and future research directions,

    Yu Mori and Naoko Mori, “Advancing osteoporosis diagnosis and fracture prediction: role of quantitative ct and future research directions,” Archives of Orthopaedic and Trauma Surgery, vol. 144, no. 8, pp. 3895– 3896, 2024

    work page 2024

  4. [4]

    Sine activated low-rank matrices for parameter efficient learning,

    Yiping Ji, Hemanth Saratchandran, Cameron Gordon, Zeyu Zhang, and Simon Lucey, “Sine activated low-rank matrices for parameter efficient learning,” arXiv preprint arXiv:2403.19243 , 2024

    arXiv 2024

  5. [5]

    Thin-thick adapter: Segmenting thin scans using thick annotations,

    Zeyu Zhang, Bowen Zhang, Abhiram Hiwase, Christen Barras, Feng Chen, Biao Wu, Adam James Wells, Daniel Y Ellis, Benjamin Reddi, Andrew William Burgan, Minh-Son To, Ian Reid, and Richard Hartley, “Thin-thick adapter: Segmenting thin scans using thick annotations,” OpenReview, 2023

    work page 2023

  6. [6]

    Bhsd: A 3d multi-class brain hemorrhage segmentation dataset,

    Biao Wu, Yutong Xie, Zeyu Zhang, Jinchao Ge, Kaspar Yaxley, Suzan Bahadir, Qi Wu, Yifan Liu, and Minh-Son To, “Bhsd: A 3d multi-class brain hemorrhage segmentation dataset,” in International Workshop on Machine Learning in Medical Imaging . Springer, 2023, pp. 147–156

    work page 2023

  7. [7]

    Segstitch: Multidimensional transformer for robust and efficient medical imaging segmentation,

    Shengbo Tan, Zeyu Zhang, Ying Cai, Daji Ergu, Lin Wu, Binbin Hu, Pengzhang Yu, and Yang Zhao, “Segstitch: Multidimensional transformer for robust and efficient medical imaging segmentation,” arXiv preprint arXiv:2408.00496 , 2024

    arXiv 2024

  8. [8]

    Esa: Annotation-efficient active learning for semantic segmentation,

    Jinchao Ge, Zeyu Zhang, Minh Hieu Phan, Bowen Zhang, Akide Liu, and Yang Zhao, “Esa: Annotation-efficient active learning for semantic segmentation,” arXiv preprint arXiv:2408.13491 , 2024

    arXiv 2024

Show all 37 references
  1. [9]

    Meddet: Generative adversarial distillation for efficient cervical disc herniation detection,

    Zeyu Zhang, Nengmin Yi, Shengbo Tan, Ying Cai, Yi Yang, Lei Xu, Qingtai Li, Zhang Yi, Daji Ergu, and Yang Zhao, “Meddet: Generative adversarial distillation for efficient cervical disc herniation detection,” in 2024 IEEE International Conference on Bioinformatics and Biomedici...

    2024

  2. [10]

    Msdet: Receptive field enhanced multiscale detection for tiny pulmonary nodule,

    Guohui Cai, Ying Cai, Zeyu Zhang, Daji Ergu, Yuanzhouhan Cao, Binbin Hu, Zhibin Liao, and Yang Zhao, “Msdet: Receptive field enhanced multiscale detection for tiny pulmonary nodule,” arXiv preprint arXiv:2409.14028, 2024

    2024 arXiv

  3. [11]

    Segkan: High-resolution medical image segmentation with long-distance depen- dencies,

    Shengbo Tan, Rundong Xue, Shipeng Luo, Zeyu Zhang, Xinran Wang, Lei Zhang, Daji Ergu, Zhang Yi, Yang Zhao, and Ying Cai, “Segkan: High-resolution medical image segmentation with long-distance depen- dencies,” arXiv preprint arXiv:2412.19990 , 2024

    2024 arXiv

  4. [12]

    Gamed-snake: Gradient-aware adaptive momentum evolution deep snake model for multi-organ segmentation,

    Ruicheng Zhang, Haowei Guo, Zeyu Zhang, Puxin Yan, and Shen Zhao, “Gamed-snake: Gradient-aware adaptive momentum evolution deep snake model for multi-organ segmentation,” arXiv preprint arXiv:2501.12844, 2025

    2025 arXiv

  5. [13]

    Jointvit: Modeling oxygen saturation levels with joint supervision on long-tailed octa,

    Zeyu Zhang, Xuyin Qi, Mingxi Chen, Guangxi Li, Ryan Pham, Ayub Qassim, Ella Berry, Zhibin Liao, Owen Siggs, Robert Mclaughlin, et al., “Jointvit: Modeling oxygen saturation levels with joint supervision on long-tailed octa,” in Annual Conference on Medical Image Understand- in...

    2024

  6. [14]

    Xlip: Cross-modal attention masked modelling for medical language-image pre-training,

    Biao Wu, Yutong Xie, Zeyu Zhang, Minh Hieu Phan, Qi Chen, Ling Chen, and Qi Wu, “Xlip: Cross-modal attention masked modelling for medical language-image pre-training,” arXiv preprint arXiv:2407.19546, 2024

    2024 arXiv

  7. [15]

    Can rotational thromboelastometry rapidly identify theragnostic targets in isolated traumatic brain injury?,

    Abhiram D Hiwase, Christopher D Ovenden, Lola M Kaukas, Mark Finnis, Zeyu Zhang, Stephanie O’Connor, Ngee Foo, Benjamin Reddi, Adam J Wells, and Daniel Y Ellis, “Can rotational thromboelastometry rapidly identify theragnostic targets in isolated traumatic brain injury?,” Emerg...

    2025

  8. [16]

    A landmark-based approach for instability predic- tion in distal radius fractures,

    Yang Zhao, Zhibin Liao, Yunxiang Liu, Koen Oude Nijhuis, Britt Barvelink, Jasper Prijs, Joost Colaris, Mathieu Wijffels, Max Reijman, Zeyu Zhang, et al., “A landmark-based approach for instability predic- tion in distal radius fractures,” in 2024 IEEE International Symposium o...

    2024

  9. [17]

    Medical ai for early detection of lung cancer: A survey,

    Guohui Cai, Ying Cai, Zeyu Zhang, Yuanzhouhan Cao, Lin Wu, Daji Ergu, Zhinbin Liao, and Yang Zhao, “Medical ai for early detection of lung cancer: A survey,” arXiv preprint arXiv:2410.14769 , 2024

    2024 arXiv

  10. [18]

    Projectedex: Enhancing generation in explainable ai for prostate cancer,

    Xuyin Qi, Zeyu Zhang, Aaron Berliano Handoko, Huazhan Zheng, Mingxi Chen, Ta Duc Huy, Vu Minh Hieu Phan, Lei Zhang, Linqi Cheng, Shiyu Jiang, et al., “Projectedex: Enhancing generation in explainable ai for prostate cancer,” arXiv preprint arXiv:2501.01392 , 2025

    2025 arXiv

  11. [19]

    Ctspine1k: a large-scale dataset for spinal vertebrae segmentation in computed tomography,

    Yang Deng, Ce Wang, Yuan Hui, Qian Li, Jun Li, Shiwei Luo, Mengke Sun, Quan Quan, Shuxin Yang, You Hao, et al., “Ctspine1k: a large-scale dataset for spinal vertebrae segmentation in computed tomography,” arXiv preprint arXiv:2105.14711 , 2021

    2021 arXiv

  12. [20]

    Totalsegmentator: robust segmentation of 104 anatomic structures in ct images,

    Jakob Wasserthal, Hanns-Christian Breit, Manfred T Meyer, Maurice Pradella, Daniel Hinck, Alexander W Sauter, Tobias Heye, Daniel T Boll, Joshy Cyriac, Shan Yang, et al., “Totalsegmentator: robust segmentation of 104 anatomic structures in ct images,” Radiology: Artificial Int...

    2023

  13. [21]

    Survey on deep learning with class imbalance,

    Justin M Johnson and Taghi M Khoshgoftaar, “Survey on deep learning with class imbalance,” Journal of big data, vol. 6, no. 1, pp. 1–54, 2019

    2019

  14. [22]

    A deep learning approach to diabetes diagnosis,

    Zeyu Zhang, Khandaker Asif Ahmed, Md Rakibul Hasan, Tom Gedeon, and Md Zakir Hossain, “A deep learning approach to diabetes diagnosis,” in Asian Conference on Intelligent Information and Database Systems . Springer, 2024, pp. 87–99

    2024

  15. [23]

    Automated bone mineral density prediction and fracture risk assessment using plain radiographs via deep learning,

    Chen-I Hsieh, Kang Zheng, Chihung Lin, Ling Mei, Le Lu, Weijian Li, Fang-Ping Chen, Yirui Wang, Xiaoyun Zhou, Fakai Wang, et al., “Automated bone mineral density prediction and fracture risk assessment using plain radiographs via deep learning,” Nature communications, vol. 12,...

    2021

  16. [24]

    Prediction of bone mineral density from computed tomography: application of deep learning with a convolutional neural network,

    Koichiro Yasaka, Hiroyuki Akai, Akira Kunimatsu, Shigeru Kiryu, and Osamu Abe, “Prediction of bone mineral density from computed tomography: application of deep learning with a convolutional neural network,” European radiology, vol. 30, pp. 3549–3557, 2020

    2020

  17. [25]

    Automated opportunistic osteoporotic fracture risk assessment using computed tomography scans to aid in frax underutilization,

    Noa Dagan, Eldad Elnekave, Noam Barda, Orna Bregman-Amitai, Amir Bar, Mila Orlovsky, Eitan Bachmat, and Ran D Balicer, “Automated opportunistic osteoporotic fracture risk assessment using computed tomography scans to aid in frax underutilization,” Nature medicine , vol. 26, no...

    2020

  18. [26]

    Deep learning for biomarker regression: application to osteoporosis and emphysema on chest ct scans,

    Germ ´an Gonz ´alez, George R Washko, and Ra ´ul San Jos ´e Est ´epar, “Deep learning for biomarker regression: application to osteoporosis and emphysema on chest ct scans,” in Proceedings of SPIE–the International Society for Optical Engineering . NIH Public Access, 2018, vol. 10574

    2018

  19. [27]

    Opportunistic osteoporosis screening in multi-detector ct images using deep convolutional neural networks,

    Yijie Fang, Wei Li, Xiaojun Chen, Keming Chen, Han Kang, Pengxin Yu, Rongguo Zhang, Jianwei Liao, Guobin Hong, and Shaolin Li, “Opportunistic osteoporosis screening in multi-detector ct images using deep convolutional neural networks,” European Radiology, vol. 31, pp. 1831–1842, 2021

    2021

  20. [28]

    Remix: Calibrated resampling for class imbalance in deep learning,

    Colin Bellinger, Roberto Corizzo, and Nathalie Japkowicz, “Remix: Calibrated resampling for class imbalance in deep learning,” arXiv preprint arXiv:2012.02312, 2020

    2012 arXiv

  21. [29]

    Class-balanced loss based on effective number of samples,

    Yin Cui, Menglin Jia, Tsung-Yi Lin, Yang Song, and Serge Belongie, “Class-balanced loss based on effective number of samples,” in Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 9268–9277

    2019

  22. [30]

    Long-tailed recognition by learning from latent categories,

    Weide Liu, Zhonghua Wu, Yiming Wang, Henghui Ding, Fayao Liu, Jie Lin, and Guosheng Lin, “Long-tailed recognition by learning from latent categories,” arXiv preprint arXiv:2206.01010 , 2022

    2022 arXiv

  23. [31]

    Sharpness-aware minimization for efficiently improving generaliza- tion,

    Pierre Foret, Ariel Kleiner, Hossein Mobahi, and Behnam Neyshabur, “Sharpness-aware minimization for efficiently improving generaliza- tion,” arXiv preprint arXiv:2010.01412 , 2020

    2010 arXiv

  24. [32]

    Learning multiple layers of features from tiny images,

    Alex Krizhevsky, “Learning multiple layers of features from tiny images,” Technical report, University of Toronto, 2009

    2009

  25. [33]

    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,” in 2009 IEEE conference on computer vision and pattern recognition . Ieee, 2009, pp. 248–255

    2009

  26. [34]

    The road less scheduled,

    Aaron Defazio, Xingyu Alice Yang, Harsh Mehta, Konstantin Mishchenko, Ahmed Khaled, and Ashok Cutkosky, “The road less scheduled,” arXiv preprint arXiv:2405.15682 , 2024

    2024 arXiv

  27. [35]

    World Health Organization, 1994

    WHO Study Group on Assessment of Fracture Risk and its Application to Screening for Postmenopausal Osteoporosis, Assessment of fracture risk and its application to screening for postmenopausal osteoporosis: report of a WHO study group , Number 843-849. World Health Organization, 1994

    1994

  28. [36]

    Segreg: Segmenting oars by registering mr images and ct annotations,

    Zeyu Zhang, Xuyin Qi, Bowen Zhang, Biao Wu, Hien Le, Bora Jeong, Zhibin Liao, Yunxiang Liu, Johan Verjans, Minh-Son To, et al., “Segreg: Segmenting oars by registering mr images and ct annotations,” in 2024 IEEE International Symposium on Biomedical Imaging (ISBI) . IEEE, 2024...

    2024

  29. [37]

    Cer- vicothoracic volumetric bone mineral density assessed by opportunistic qct may be a reliable marker for osteoporosis in adults,

    Constanze Ramsch ¨utz, Nico Sollmann, Malek El Husseini, Karina Kupfer, Karolin J Paprottka, Maximilian T L ¨offler, Moritz R Hernandez Petzsche, Julian Schwarting, Jannis Bodden, Thomas Baum, et al., “Cer- vicothoracic volumetric bone mineral density assessed by opportunistic...

    2024

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.