Pith. sign in

REVIEW 4 major objections 8 minor 42 references

A Hybrid Quantum Classical Pipeline for X Ray Based Fracture Diagnosis

T0 review · 4 major / 8 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A 16-dimensional fused PCA-plus-quantum feature vector lets a plain SVM match deep transfer learning at 99% accuracy on bone X-rays while cutting feature-extraction time by 82%.

desk verdict A novel application with a plausible-sounding result, but the quantum advantage claim is unsubstantiated because the comparison is confounded and the circuit is never defined. read the letter →

arxiv 2505.14716 v1 pith:5XJFNZFR submitted 2025-05-19 eess.IV cs.CVcs.ETcs.LG

Sahil Tomar , Rajeshwar Tripathi , Sandeep Kumar This is my paper · ORCID
classification eess.IVcs.CVcs.ETcs.LG
keywords BonefracturedetectionX-raydiagnosisPrincipalComponentAnalysisQuantumfeatureextractionAmplitudeencodingHybridquantum-classicalpipelineSupportvectormachineMedicalimageclassification
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

Automated fracture screening usually means heavy deep networks trained on large annotated image sets. This paper proposes a far lighter route: compress each X-ray with PCA to eight numbers, feed those numbers through a simulated four-qubit quantum amplitude-encoding circuit, and fuse the eight classical values with eight quantum-derived values into a 16-feature vector for an ordinary classifier such as an SVM. On a public multi-region bone X-ray dataset, this pipeline is reported to reach 99% accuracy, matching the transfer-learned baseline it is compared with, while cutting mean feature-extraction time by 82%. The paper's thesis is that the quantum half of the vector supplies nonlinear structure that pure PCA lacks, letting a 16-feature SVM outperform a classical PCA-only SVM (99% versus 83% accuracy). If the result holds, accurate fracture screening no longer requires a deep-learning stack or large annotated corpora.

What carries the argument

The load-bearing object is the 4-qubit quantum amplitude-encoding circuit used for feature enrichment. Eight PCA-reduced features are normalized and mapped onto the amplitudes of a four-qubit state, $|\psi\rangle = \sum_{i=0}^{N-1} x_i |i\rangle$, then passed through rotation gates (RX, RY) and entanglement operations before being measured to yield eight quantum-enhanced features. These eight values are concatenated with the eight PCA values to form the 16-dimensional dual-source feature vector that feeds the classical classifiers. The circuit is what the paper credits for the accuracy jump: removing the quantum module is reported to drop accuracy by 10–15 percent, and the SVM's gain from 83% to 99% is attributed to the structure these measured features add on top of PCA.

What would settle it

Re-run the pipeline with the split performed before any augmentation, filtering, or edge detection, and verify by image hashing that no training image or any transformed copy of it appears in the test set; then recompute SVM accuracy and Cohen's kappa. As a second check, fit a classical SVM on the eight PCA features alone under the same leakage-free split: if its accuracy already approaches 99%, the claimed contribution of the quantum module collapses.

Watch

Extended reading notes

Core claim

The paper's central claim is that a dual-source feature vector—eight principal components of an X-ray image plus eight features measured from a 4-qubit amplitude-encoding circuit—is enough to classify bone fractures at 99% accuracy on the Bone Fracture Multi-Region X-Ray Dataset. With this representation, an SVM achieves 99% accuracy, an F1-score of 0.99, and a Cohen's kappa of 0.98, while a classical SVM on PCA features alone reaches 83% accuracy and 0.65 kappa; the paper describes the hybrid SVM as beating the classical SVM "by a clear margin in every possible parameter." The same fused pipeline matches the 99% accuracy reported by the MobLG-Net transfer-learning baseline while reducing mean feature-extraction time from 67.5 seconds to 12.1 seconds. The intended conclusion is that quantum amplitude encoding contributes genuine, non-classically-redundant information to the feature set, and that this enrichment is obtainable within a 4-qubit simulator constraint.

Load-bearing premise

The 99% figure stands or falls on the test images being truly unseen: the paper preprocesses, augments, and edge-detects the images before splitting the data, and it reports the split inconsistently (80/10/10 in the methodology, 80/20 in the experiments), so any transformed copy of a training image that lands in the test set would inflate the reported accuracy.

Editorial extensions

If this is right

  • A clinic with only a CPU workstation could run fracture screening: the reported SVM training time is 0.07 seconds and evaluation time is 0.04 seconds.
  • If a 16-feature fused vector really matches deep transfer learning at 99%, the need for large annotated datasets and GPU training is removed for this task, lowering barriers in resource-limited settings.
  • The reported 82% reduction in feature-extraction time and roughly 15% reduction in inference time suggest the hybrid representation could serve real-time decision support where transfer-learned CNNs are too slow.
  • The ablation finding—accuracy falls 10–15% without the quantum module—means the quantum features, not the PCA features, are the active ingredient to optimize in later work.

Reading between the lines

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

  • An inference the authors leave implicit: since the quantum circuit is simulated on a classical workstation, the 83-to-99 point gain is a claim about feature representation, not about a hardware speedup; quantum hardware is not needed to reproduce the result.
  • A testable extension they do not run: replace the quantum module with a classical nonlinear expansion of the same eight PCA features (for example, polynomial or random Fourier features) at equal dimension; if that also reaches 99%, the measured gain is not specific to quantum encoding.
  • If the 99% figure survives a strictly leakage-free split, the same PCA-plus-amplitude-encoding-plus-fused-SVM recipe is a natural candidate for other scarce-data medical imaging tasks, such as chest or pediatric X-ray screening, where annotation costs are the main bottleneck.
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 / 8 minor

Summary. The paper proposes a hybrid quantum-classical pipeline for X-ray fracture detection. It reduces each image to 8 PCA features, encodes them into a 4-qubit amplitude-encoding circuit to produce 8 quantum features, fuses the two sets into a 16-dimensional vector, and trains classical ML classifiers (SVM, Random Forest, KNN, Gradient Boosting, Decision Tree). On the publicly available 9,463-image Bone Fracture Multi-Region X-Ray Dataset, it reports 99% accuracy with SVM, an 82% reduction in feature-extraction time relative to a MobileNet-LGBM transfer-learning baseline, and claims that hybrid quantum SVM beats classical SVM by a clear margin (99% vs 83% accuracy, Cohen's kappa 0.98 vs 0.65). An ablation bullet states that removing the quantum module lowers accuracy by approximately 10-15%.

Significance. If the claims were substantiated, the paper would be of interest to both the quantum-machine-learning and medical-imaging communities, since a small 4-qubit feature-enrichment module could improve classical classifiers while reducing computational cost versus deep transfer learning. The paper has notable strengths: it uses a public dataset, reports concrete metrics for several classifiers (Table 1), includes a classical-versus-hybrid comparison (Table 2), and provides wall-clock timing comparisons (Table 3). However, the central claim of quantum advantage is not currently established because the hybrid/classical comparison changes both feature dimensionality and feature construction, the quantum circuit is not specified, the data-split description is inconsistent, and the ablation study is not reported with enough detail to isolate the quantum contribution. These issues are load-bearing and prevent verification of the headline results.

major comments (4)
  1. [Section III-E and Section IV, Table 2] The classical SVM baseline is trained on 'only PCA-derived features' (8 dimensions), while the hybrid SVM uses the fused 16-dimensional vector (8 PCA + 8 quantum). The comparison therefore changes two variables at once: feature dimensionality (8 vs 16) and feature construction (linear PCA vs PCA plus a nonlinear quantum map). The reported jump from 83% to 99% accuracy and from 0.65 to 0.98 kappa could plausibly result from the additional 8 informative dimensions alone, with no quantum contribution. To support the paper's central claim, the authors must include a matched-dimensionality classical baseline, e.g., an SVM on 16 PCA components, or on 8 PCA features concatenated with 8 classical nonlinear features. Without such a control, the statement that the hybrid approach 'is beating the classical SVM approach by a clear margin in every possible parameter' is not supported by the reported experiment.
  2. [Section III-D] The quantum feature extraction circuit is not defined. The text states that amplitude encoding is used with a 4-qubit scheme and that 'the data undergoes various quantum operations such as rotation gates (RX, RY) and entanglement mechanisms', but it does not provide the circuit diagram, the gate sequence, the entanglement pattern, whether the rotation angles are fixed or variational, or which observables are measured. Because the quantum circuit is the central novel component of the pipeline, this omission makes the method impossible to reproduce or audit. A precise specification of the circuit, including the mapping from the 8 PCA features to the 4-qubit state, is required.
  3. [Section III-A and Section IV (Experiment Setup)] The data split is described inconsistently across the paper. Section III-A states that the dataset is split 80/10/10 into training, validation, and testing (7,570/946/947 images), while Section IV says an 80/20 train-test split is used. Section III-B further states that preprocessing, including data augmentation and Canny edge detection, is completed before the split. If augmented or filtered versions of the same original image appear in both the training and test partitions, the reported 99% accuracy could be inflated. The authors must clarify the actual split, state how the validation set was used, and confirm that the train/test split was performed before any augmentation or filtering. This is essential to the credibility of the headline accuracy.
  4. [Section IV (Ablation Study)] The ablation study does not provide enough information to assess the quantum contribution. The bullet 'when the quantum feature extraction module was removed, accuracy dropped by approximately 10-15%' gives no exact accuracy values, no standard deviations, and no indication of the feature dimensionality of the ablated model. It also repeats the confound described above: removing the quantum module reduces the feature vector from 16 to 8 dimensions, so the observed drop could be due to dimensionality alone. The authors should report the ablation with matched feature dimensionality (e.g., an 8-PCA-feature model versus a 16-feature model that replaces quantum features with classical features) and provide exact metrics, ideally with repeated runs.
minor comments (8)
  1. [Abstract] The final sentence is an incomplete grammatical construction: '...into a 16-dimensional vector and then classifying with different machine learning models achieving 99% accuracy...'. Please revise to form a complete sentence.
  2. [Section III-D, Eq. (1)] The text says the basis states are of 'log2 n qubits', but the correct expression is log2 N qubits, where N is the number of features. Please correct the notation.
  3. [Figures 3 and 4] The confusion matrices are not described with axis labels or class names, making it difficult to interpret the counts. Please add labels and, if possible, report the underlying counts in the text.
  4. [Table 1] The MCVS column reports mean cross-validation values but no standard deviations. Since the text defines MCVs as repeated random splits, report the variability to allow assessment of stability.
  5. [Section V (Conclusion) versus Table 2 and Table 3] The conclusion states training times are reduced 'by up to 30%', but Table 2 shows 0.11 sec versus 0.07 sec (about 36% reduction) and Table 3 reports an 82% reduction in feature-extraction time. Please reconcile these numbers and state clearly which time comparison is intended.
  6. [Reference [28]] Please include the dataset version and access date for the Kaggle link. In addition, the claim of being 'on par with state-of-the-art transfer learning models' is supported only by comparison with a single model [16]; consider phrasing this as 'the transfer-learning baseline reported in [16]'.
  7. [Section III-F] The statement that the dataset size is 'relatively limited' for DL methods is questionable given 9,463 images. Please justify this claim or remove it.
  8. [Section IV (Experiment Setup)] The text says 'Python 3.0'; this should be specified as a concrete version or 'Python 3.x' to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the claims are empirical benchmark results against a public dataset, not a derivation that reduces to its own assumptions.

full rationale

The paper makes no formal derivation that is equivalent to its inputs by construction. Its central claim is an empirical accuracy figure obtained by training classical ML classifiers on a fused PCA-plus-quantum feature vector and evaluating on a public X-ray dataset, with comparisons to a transfer-learning baseline from a separate publication. No parameter is fitted to a subset and then renamed as a prediction; no uniqueness theorem is imported from the authors' prior work; the authors do not rely on their own prior results as the load-bearing justification. The described confounding between the classical baseline (8 PCA features) and the hybrid model (16 fused features) is a fairness and soundness concern about whether the reported quantum advantage is established, not a circularity concern. Likewise, the inconsistency between the 80/10/10 and 80/20 splits and the preprocessing-before-split description raise data-leakage questions, but they do not make any claimed result equivalent to its own inputs. The ablation statement reporting a 10–15% drop when the quantum module is removed lacks exact numbers and matched-dimensionality control, but this is an evidentiary weakness rather than a constructional circularity. Under the requested standard, the honest finding is no significant circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on several design choices and unstated assumptions: the number of PCA and quantum features, the undisclosed quantum circuit parameters, a consistent data split, and the faithfulness of the quantum simulator. No new physical entities are introduced. The free parameters and axioms above are the main items the reader supplies without independent evidence.

free parameters (3)
  • Number of PCA components = 8
    Chosen by hand to fit the 4-qubit encoding constraint; no explained variance criterion or sweep is reported.
  • Number of quantum features = 8
    Set to make the fused vector 16-dimensional; no optimization or ablation for this count is reported.
  • Quantum circuit rotation and entanglement parameters = not reported
    The circuit is described only as using RX, RY and "entanglement mechanisms"; any angles or entangling gate choices that produced the eight measured features are not disclosed.
assumptions (4)
  • domain assumption The 4-qubit amplitude encoding of eight PCA features produces a valid quantum state with unit norm (Equations 1 and 2).
    The paper uses the encoding as the basis for quantum feature extraction, but gives no circuit construction or verification of the amplitudes.
  • ad hoc to paper The data split described in Section IV is valid even though Section III-A states a different split and preprocessing happens before splitting.
    The paper does not reconcile the 80/10/10 and 80/20 descriptions, and it does not state that augmentation was applied only inside the training partition.
  • domain assumption The simulated PennyLane circuit faithfully represents a quantum device for the purpose of clinical ML advantage.
    All results come from a classical simulator, yet the paper frames the quantum enhancement as a genuine advantage; no error model or hardware validation is provided.
  • domain assumption The public dataset labels and train/test folder structure are reliable and match the described partition.
    Section III-A relies on the dataset organization; no label verification or class-balance analysis is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Hybrid Quantum Classical Pipeline for X Ray Based Fracture Diagnosis." pith.science (2026). https://pith.science/paper/5XJFNZFR

@misc{pith2026250514716,
  author       = {Pith},
  title        = {Pith review of: A Hybrid Quantum Classical Pipeline for X Ray Based Fracture Diagnosis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5XJFNZFR}},
  note         = {Machine review of arXiv:2505.14716}
}
read the original abstract

Bone fractures are a leading cause of morbidity and disability worldwide, imposing significant clinical and economic burdens on healthcare systems. Traditional X ray interpretation is time consuming and error prone, while existing machine learning and deep learning solutions often demand extensive feature engineering, large, annotated datasets, and high computational resources. To address these challenges, a distributed hybrid quantum classical pipeline is proposed that first applies Principal Component Analysis (PCA) for dimensionality reduction and then leverages a 4 qubit quantum amplitude encoding circuit for feature enrichment. By fusing eight PCA derived features with eight quantum enhanced features into a 16 dimensional vector and then classifying with different machine learning models achieving 99% accuracy using a public multi region X ray dataset on par with state of the art transfer learning models while reducing feature extraction time by 82%.

Figures

Figures reproduced from arXiv: 2505.14716 by the authors.

Figure 2
Figure 2. The x-ray images examination using the target label (a), (b) fractured and (c), (d) not fractured To improve model generalization, data augmentation is carried out by incorporating operations such as rotation, flipping, scaling, and adjustments to contrast. Finally, edge detection algorithms, such as Canny edge detection, are used to highlight fracture lines, aiding the model in identifying subtle fractures that mig… view at source ↗
Figure 3
Figure 3. Confusion Matrix of Decision tree, Gradient Boosting, KNN, SVM, and Random Forest respectively via Hybrid [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Confusion Matrix of (a) Classical SVM, and (b) Hybrid [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Precision, Recall, F1-score and accuracy comparison graph of Classical SVM and Hybrid￾Quantum SVM [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 37 canonical work pages

  1. [16]

    Novel transfer learning based bone fracture detection using radiographic images,

    A. Alam, A. S. Al-Shamayleh, N. Thalji, et al., “Novel transfer learning based bone fracture detection using radiographic images,” BMC Med. Imaging, vol. 25, p. 5, 2025

    work page 2025

  2. [1]

    Williamson , F

    S. Williamson , F. Landeiro , T. McConnell , L. Fulford -Smith, M. K. Javaid, A. Judge and J. Leal, Costs of fragility hip fractures globally: a systematic review and meta -regression analysis. Osteoporos Int. ,; vol. 28, pp. 2791–800, 2017

    work page 2017

  3. [2]

    The global burden of fragility fractures – what are the differences, and where are the gaps,

    P. J. Mitchell, D.-C. Chan, J.-K. Lee, I. Tabu, and B. B. Alpuerto, “The global burden of fragility fractures – what are the differences, and where are the gaps,” Best Pract. Res. Clin. Rheumatol. , vol. 36, no . 3, p. 101777, 2022

    work page 2022

  4. [3]

    S. L. Ayesa, A. G. Katelaris , P. C. Brennan , S. M. Grieve. Medical imaging education opportunities for junior doctors and non -radiologist clinicians: A review. J Med Imaging Radiat Oncol. , vol. 65, no. 6, pp. 710–718, 2021

    work page 2021

  5. [4]

    Artificial intelligence -based applications for bone fracture detection using medical images: A systematic review,

    M. Kutbi, “Artificial intelligence -based applications for bone fracture detection using medical images: A systematic review,” Diagnostics (Basel), vol. 14, no. 17, p. 1879, Aug. 27, 2024

    work page 2024

  6. [5]

    Advancements a nd initial experiences in AI -assisted X -ray based fracture diagnosis: A narrative review,

    S. S. Hosseini -Begtary, A. Gurabi, P. Hegedus, and N. Marton, “Advancements a nd initial experiences in AI -assisted X -ray based fracture diagnosis: A narrative review,” Imaging, published online ahead of print, 2025

    work page 2025

  7. [6]

    Deep learning in fracture detection: a narrative review,

    P. H. S. Kalmet, S. Sanduleanu, S. Primakov, G. Wu, A. Jochems, T. Refaee, A. Ibrahim, L. v. Hulst, P. Lambin, and M. Poeze, “Deep learning in fracture detection: a narrative review,” Acta Orthop., vol. 91, no. 2, pp. 215–220, 2020

    work page 2020

  8. [7]

    Bone Fracture Detection and Classification using Deep Learning Appr oach,

    D. P. Yadav and S. Rathor, "Bone Fracture Detection and Classification using Deep Learning Appr oach," 2020 International Conference on Power Electronics & IoT Applications in Renewable Energy and its Control (PARC), Mathura, India, 2020

    work page 2020

Show all 42 references
  1. [8]

    Bone fracture detection using deep supervised learning from radiological images: A paradigm shift,

    T. Meena and S. Roy, “Bone fracture detection using deep supervised learning from radiological images: A paradigm shift,” Diagnostics, vol. 12, p. 2420, 2022

    2022

  2. [9]

    Artificial intelligence in fracture detection: Transfer learning from deep convolutional neural networks,

    D. H. Kim and T. MacKinnon, “Artificial intelligence in fracture detection: Transfer learning from deep convolutional neural networks,” Clin. Radiol., vol. 73, no. 5, pp. 439–445, 2018. 8

    2018

  3. [10]

    Skeletal fracture detection with deep learning: A comprehensive review,

    Z. Su, A. Adam, M. F. Nasrudin, M. Ayob, and G. Punganan, “Skeletal fracture detection with deep learning: A comprehensive review,” Diagnostics, vol. 13, no. 20, p. 3245, 2023

    2023

  4. [11]

    Deep neural network improves fracture detection by clinicians,

    R. Lindsey, A. Daluiski, S. Chopra, A. Lachapelle, M. Mozer, S. Sicular, D. Hanel, M. Gardner, A. Gu pta, R. Hotchkiss, and H. Potter, “Deep neural network improves fracture detection by clinicians,” Proc. Natl. Acad. Sci. U.S.A., vol. 115, no. 45, pp. 11591–11596, Nov. 2018

    2018

  5. [12]

    Deep Convolutional Neural Networks with Tran sfer Learning for Bone Fracture Recognition using Small Exemplar Image Datasets,

    K. N. K. Reddy and V. Cutsuridis, "Deep Convolutional Neural Networks with Tran sfer Learning for Bone Fracture Recognition using Small Exemplar Image Datasets," 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW) , Rhodes Island,...

    2023

  6. [13]

    Convolutional neural networks for automated fracture detection and localization on wrist radiographs,

    Y. L. Thian, Y. Li, P. Jagmohan, D. Sia, V. E. Y. Chan, and R. T. Tan, “Convolutional neural networks for automated fracture detection and localization on wrist radiographs,” Radiol. Artif. Intell., vol. 1, no. 1, p. e180001, Jan. 30, 2019

    2019

  7. [14]

    Rib fracture d etection system based on deep learning,

    L. Yao, X. Guan, X. Song, et al., “Rib fracture d etection system based on deep learning,” Sci. Rep., vol. 11, p. 23513, 2021

    2021

  8. [15]

    Gray -level co -occurrence matrix bone fracture detection,

    H. Y. Chai, L. K. Wee, T. T. Swee, S. Salleh, A. Ariff, and K. Kamarulafizam, “Gray -level co -occurrence matrix bone fracture detection,” Am. J. Appl. Sci., vol. 8, no. 1, pp. 26–32, 2011

    2011

  9. [17]

    Comparative analysis of deep learning models for fracture detection and classification in X -ray images,

    A. K. Hamzat, M. S. Murad, M. Kanan, and R. Asmatulu, “Comparative analysis of deep learning models for fracture detection and classification in X -ray images,” Adv. Eng. Tec. Appl. , vol. 13, no. 1, pp. 109 –115, 2024

    2024

  10. [18]

    A comparative analysis on image processing for bone fracture detection system usi ng machine learning techniques,

    V. A. Shah and H. P. Chauhan, “A comparative analysis on image processing for bone fracture detection system usi ng machine learning techniques,” YMER, vol. 23, no. 10, Oct. 2024

    2024

  11. [19]

    Fracture detection in wrist X - ray images using deep learning-based object detection models,

    F. Hardalaç, F. Uysal, O. Peker, M. Çiçeklidağ, T. Tolunay, N. Tokgöz, U. Kutbay, B. Demirciler, and F. Mert, “Fracture detection in wrist X - ray images using deep learning-based object detection models,” Sensors, vol. 22, no. 3, p. 1285, 2022

    2022

  12. [20]

    MobileNets: Efficient convolutional neural networks for mobile vision applications,

    A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto, and H. Adam, “MobileNets: Efficient convolutional neural networks for mobile vision applications,” arXiv preprint arX iv:1704.04861, 2017. [Online]. Available: https://arxiv.org/abs/1704.04861

    2017 arXiv

  13. [21]

    Principal component analysis: a review and recent developments,

    I. T. Jolliffe and J. Cadima, “Principal component analysis: a review and recent developments,” Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. , vol. 374, no. 2065, p. 20150202, 2016

    2016

  14. [22]

    Quantum machine learning,

    J. Biamonte, P. Wittek, N. Pancotti, et al., “Quantum machine learning,” Nature, vol. 549, pp. 195 –202, 2017. [Online]. Available: https://doi.org/10.1038/nature23474

    2017 doi

  15. [23]

    Schuld and F

    M. Schuld and F. Petruccione, Supervised Learning with Quantum Computers. Cham: Springer, 2018

    2018

  16. [24]

    Supervised learning with quantum-enhanced feature spaces,

    V. Havlíček, A. D. Córcoles, K. Temme, et al., “Supervised learning with quantum-enhanced feature spaces,” Nature, vol. 567, pp. 209 –212,

  17. [25]

    Quantum machine learning in me dical image analysis: A survey,

    W. Lin, H. Liu, J. Xu, L. Shi, Z. Shan, B. Zhao, and Y. Gao, “Quantum machine learning in me dical image analysis: A survey,” Neurocomputing, vol. 525, pp. 42–53, 2023

    2023

  18. [26]

    Quantum generative learning for high-resolution medical image generation,

    A. Khatun, K. Y. Aydeniz, Y. S. Weinstein, and M. Usman, “Quantum generative learning for high-resolution medical image generation,” arXiv preprint arXiv:2406.13196 , 2025. [Online ]. Available: https://arxiv.org/abs/2406.13196

    2025 arXiv

  19. [27]

    Artificial intelligence versus radiologist in the accuracy of fracture detection based on computed tomography images: a multi -dimensional, multi-region analysis,

    Y. Liu, W. Liu, H. Chen, S. Xie, C. Wang, T. Liang, Y. Yu, and X. Liu, “Artificial intelligence versus radiologist in the accuracy of fracture detection based on computed tomography images: a multi -dimensional, multi-region analysis,” Quant. Imaging Med. Surg. , vol. 13, no. ...

    2023

  20. [28]

    Papers With Code

    Bone Fracture Multi -Region X -Ray Datase t. Papers With Code. Available online: https://www.kaggle.com/datasets/ bmadushanirodrigo/fracture-multi-region-x-ray-data/data

  21. [29]

    J. S. Lim, Two-Dimensional Signal and Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1990

    1990

  22. [30]

    R. C. Gonzalez, and R. E. Woods, Digital Image Processing. 4th Edition, Pearson Education, New York, 2018

    2018

  23. [31]

    Artificial Intelligence -Based Applications for Bone Fracture Detection Using Medical Images: A Systematic Review,

    M. Kutbi, "Artificial Intelligence -Based Applications for Bone Fracture Detection Using Medical Images: A Systematic Review," Diagnostics, vol. 14, no. 17, p. 1879, Aug. 2024

    2024

  24. [32]

    Bone fracture detection through X -ray using edge detection algorithms,

    T. Panda, H. S. P. Peddada, A. Gupta, and G. Kanimozhi, "Bone fracture detection through X -ray using edge detection algorithms," Indonesian Journal of Electrical Engineering and Informatics (IJEEI) , vol. 10, no. 3, pp. 508–521, Sep. 2022

    2022

  25. [33]

    Preparing medical imaging data for machine learning,

    M. J. Willemink, W. A. Koszek, C. Hardell, J. Wu, D. Fleischmann, H. Harvey, L. R. Folio, R. M. Summers, D. L. Rubin, and M. P. Lungren, "Preparing medical imaging data for machine learning," Radiology, vol. 295, no. 1, pp. 4–15, Apr. 2020

    2020

  26. [34]

    I. T. Jolliffe, Principal Component Analysis , Springer Series in Statistics, Springer New York, NY, 2002

    2002

  27. [35]

    A Quantum Annealer for Subset Feature Selection and the Classification of Hyperspectral Imag es,

    S. Otgonbaatar and M. Datcu, "A Quantum Annealer for Subset Feature Selection and the Classification of Hyperspectral Imag es," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 14, pp. 7057-7065, 2021

    2021

  28. [36]

    Brain tumor diagnosis using quantum convolutional neural networks,

    M. A.-Z. Khan, N. Innan, A. A. O. Galib, and M. Bennai, "Brain tumor diagnosis using quantum convolutional neural networks," arXiv, Jan

  29. [37]

    Artificial intelligence for diagnosis in otologic patients: Is it ready to be your doctor?

    C. Marshall, J. Forbes, M. D. Seidman, L. Roldan, and J. Atkins, "Artificial intelligence for diagnosis in otologic patients: Is it ready to be your doctor?" Otolaryngology–Head and Neck Surgery , vol. 45, no. 8, pp. 863–869, Sep. 2024

    2024

  30. [38]

    C. M. Bishop, Pattern Recognition and Machine Learning , Information Science and Statistics, Springer New York, NY, 2006

    2006

  31. [39]

    A systematic analysis of performance measures for classification tasks,

    M. Sokolova and G. Lapalme, "A systematic analysis of performance measures for classification tasks," Information Processing & Management, vol. 45, no. 4, pp. 427–437, 2009

    2009

  32. [40]

    Prediction error estimation: a comparison of resampling methods,

    A. M. Molinaro, R. Simon, and R. M. Pfeiffer, "Prediction error estimation: a comparison of resampling methods," Bioinformatics, vol. 21, no. 15, pp. 3301–3307, Aug. 2005

    2005

  33. [2019]

    Available: https://doi.org/10.1038/s41586-019-0980-2

    [Online]. Available: https://doi.org/10.1038/s41586-019-0980-2

    doi

  34. [2024]

    Available: https://arxiv.org/abs/2401.15804

    [Online]. Available: https://arxiv.org/abs/2401.15804

    arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.