Pith. sign in

REVIEW 3 major objections 5 minor 53 references

ICGM-FRAX: Iterative Cross Graph Matching for Hip Fracture Risk Assessment using Dual-energy X-ray Absorptiometry Images

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read ICGM-FRAX converts each DXA femur scan into a graph whose nodes are anatomical regions of interest and whose edges encode spatial layout, then predicts hip fracture risk by iteratively matching the graph against templates with known…

desk verdict A legitimate graph-matching application to DXA fracture risk, but the near-perfect metrics rest on an unstated restriction to equal-size graph pairs, and the paper never reports how many test cases were unclassifiable. read the letter →

arxiv 2504.15384 v1 pith:VXI6MVSH submitted 2025-04-21 cs.CV

Chen Zhao , Anjum Shaik , Joyce H. Keyak , Nancy E. Lane , Jeffrey D. Deng , Kuan-Jui Su , Qiuying Sha , Hui Shen
show 2 more authors
Hong-Wen Deng Weihua Zhou
This is my paper · ORCID
classification cs.CV
keywords hipfractureriskpredictiondual-energyX-rayabsorptiometryDXAgraphmatchingradiomicsneuralnetworksemanticUKBiobank
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's claim is that hip fracture risk can be read from a standard dual-energy X-ray absorptiometry (DXA) scan by converting the femur into a small graph and comparing that graph against templates built from people known to have fractured. Each anatomical region of the femur — head, subcapital, inferior and superior neck, intertrochanteric region, greater trochanter, and shaft — becomes a node carrying radiomic, clinical, and bone-density features, and edges connect nearby regions by centroid distance. The proposed ICGM-FRAX matches a test graph against template graphs through three stacked mechanisms: intra-graph GCN embedding, Sinkhorn-based node affinities, and iterative cross-graph embedding, then classifies by majority vote over template matches above a threshold. On 547 UK Biobank subjects the method reports 99.7% accuracy, 98.7% sensitivity, and specificity of 1.0000, far above ResNet baselines, which matters because roughly half of hip-fracture patients are not flagged by bone-mineral-density T-scores alone.

What carries the argument

The load-bearing object is the DXA-derived semantic graph: seven manually annotated femur regions of interest as nodes, each node carrying a 130-dimensional feature vector of radiomics, clinical variables, and BMD/BMC measurements, with edges drawn by k-nearest-neighbors on the Euclidean distance between region centroids. Three phases do the matching: an intra-graph embedding that applies layered graph convolutions to each node; a Sinkhorn normalization that converts the learned node-to-node affinity matrix into a doubly-stochastic soft assignment; and M rounds of cross-graph embedding in which each node's representation is updated by affinity-weighted features from the other graph, so the two graphs co-evolve during comparison. Graph average pooling then yields one vector per graph, and the predicted similarity is their cosine similarity. The ablation establishes that the cross-graph stage is the mechanism doing the work: removing it lowers the best achievable sensitivity from 0.9869 to 0.9518 across all tested hyperparameter settings.

What would settle it

Re-run the published protocol on an independent DXA cohort with a different scanner or population and report the subject-level confusion matrix that includes every test subject, not only those whose region count matches a template. Concretely, count how many of the 98 test subjects had at least one same-size template graph — 3,036 matched pairs out of 5,390 possible test–template combinations suggests a substantial share of comparisons were excluded — and check whether sensitivity stays at 0.9869 when all subjects are included.

Watch

Extended reading notes

Core claim

The central discovery, stated on the paper's own terms, is that a similarity score between whole-femur graphs is enough to separate future hip-fracture cases from controls almost perfectly, so a DXA scan can be classified without training a disease-specific image classifier. On the UK Biobank test set, ICGM-FRAX attains ACC 0.9970 ± 0.0068, F1 0.9932 ± 0.0162, SN 0.9869 ± 0.0310, and SP 1.0000 ± 0.0000 across 10 repeated runs, surpassing transfer-learning baselines (ResNet18, ResNet50, ResNet152, with and without clinical features), whose best sensitivity is 0.3947. The paper also claims the decision is interpretable: a leave-one-out feature analysis ranks sex, GLCM and GLDM texture features, structural shape features, and femoral-neck BMD T-score as the most influential attributes, and shows that near-top performance is reached with roughly 20–30 of the 130 features.

Load-bearing premise

The load-bearing premise is that every test scan can be paired with at least one template scan having the same number of femur regions: the network is trained only on pairs with equal region counts, and the feature-importance pass used only the 3,036 of 5,390 possible test–template pairs that had matching counts, so the paper does not establish how a test subject whose segmentation yields an unmatched region count would be classified.

Editorial extensions

If this is right

  • Screening could be upgraded without new imaging hardware, since the input is the DXA scan that is already the clinical standard for bone-fracture risk.
  • The decision is locally inspectable: zeroing individual features shows that sex, GLCM and GLDM texture features, and femoral-neck T-score carry most of the signal, so clinicians can see which anatomy and texture drive the risk label.
  • The model reaches near-top performance with roughly 20–30 of the 130 features, so a simplified clinical deployment could run with a much smaller feature set.
  • The ablation identifies cross-graph embedding as the essential component: without it, sensitivity falls to at most 0.9518 from 0.9869.
  • The same graph-matching construction transfers to other diagnostic imaging tasks where the spatial relationship between semantic regions matters, as the paper explicitly claims.

Reading between the lines

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

  • The paper never states how many of the 98 test subjects had at least one same-size template graph, yet 3,036 matched pairs out of 5,390 possible test–template combinations suggests many comparisons were impossible; a deployment version would need a defined fallback for unmatched subjects.
  • The zero-variance specificity (1.0000 ± 0.0000 over 10 runs) means the majority-vote threshold absorbs every ambiguity in the negative class; a threshold sweep would reveal the actual SN/SP operating curve of the method.
  • Because sex is the single most influential feature, a stratified analysis by sex and age would clarify whether the graph is separating bone structure or mostly demographic priors.
  • A direct head-to-head against femoral-neck T-score alone on the same 547 subjects would quantify the added value of the graph over the current gold standard, which the paper motivates but does not measure.
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

3 major / 5 minor

Summary. The paper proposes ICGM-FRAX, a graph-matching method for hip fracture risk assessment from DXA images. Each DXA image is converted into a graph whose nodes are femur RoIs with radiomic and clinical features; a network learns pairwise graph similarity via intra-graph and cross-graph embeddings, and a test subject is classified by comparing its graph to template graphs and applying majority voting. On a UK Biobank subset of 547 subjects (94 hip fractures), the method reports accuracy 0.9970, sensitivity 0.9869, and specificity 1.0000, substantially outperforming ResNet-based baselines. The authors also provide an ablation study, hyperparameter analysis, and feature-importance interpretation.

Significance. If the reported performance were robust, the method would be a major advance in hip fracture screening, offering near-perfect risk stratification from routine DXA images with an interpretable graph-based representation. The paper has some strengths: the idea of modeling DXA images as semantic graphs is interesting, the public code link is helpful, and the ablation and feature-importance analyses are useful additions. However, the empirical claims rest on a small, selected test set and on unstated and potentially exclusionary conditions in the matching procedure; these gaps prevent the results from supporting the paper's central claims as written.

major comments (3)
  1. [§4.3, Table 2; §3.1, Table 1] The claimed near-perfect performance is not compared against a baseline using only BMD T-scores and clinical features, even though node features 122–130 in Table 1 include BMD, BMC, and T-scores, which are established strong risk factors for hip fracture. Without a simple baseline such as logistic regression on the femoral neck T-score or a BMD-only classifier, the reader cannot determine whether the graph-matching architecture adds any predictive value beyond the well-known clinical predictors embedded in the node features. This comparison is essential to support the claim that ICGM-FRAX 'significantly outperforms existing methods.'
  2. [§3.3, Algorithm 2; §4.5] The method as defined cannot compute similarity for graph pairs with unequal numbers of RoI nodes: Eq. 2's Sinkhorn operator produces a doubly-stochastic matrix only for equal-size node sets, and the sums in Eqs. 3–4 are over n1 and n2 with no trained behavior for n1 ≠ n2. Training explicitly restricts to equal-size pairs in Section 3.3, yet Algorithm 2 compares every test graph against every template without specifying what happens when a test graph has no template with the same RoI count. Section 4.5 states that only 3036 of the 5390 possible pairs (98 test × 55 template) are generated 'when the testing graph and template graph contain the same number of RoIs,' meaning 2354 pairs (~44%) are never scored. The manuscript never reports how many test subjects, if any, have no same-size template, nor how such subjects would be classified. If any test subjects were excluded from the denominator, then the reported ACC 0.9970, SN 0.9869, and SP 1.0000 are conditional on a classifiable subset and do not support the image-level claim of near-perfect fracture risk prediction.
  3. [§4.2–§4.3] The evaluation uses only about 98 test subjects per split, with roughly 17 hip fracture cases expected given the cohort's fracture rate, yet the paper reports means and standard deviations over 10 runs without stating how many subjects were actually classified per run, the number of positive test cases, or any confidence intervals. With such a small positive class, the near-zero standard deviations (e.g., SP 1.0000±0.0000) are not informative, and no significance test is provided for the claim that ICGM-FRAX 'significantly outperforms' the baselines. The authors should report the exact test-set composition and per-run confusion matrices, and use appropriate statistical tests for comparing classifiers on small samples.
minor comments (5)
  1. [Table 2] The ResNet152 without clinical features row reports ACC 0.0771±0.0058, which is far below chance and inconsistent with the other baselines; this is likely a typographical or reporting error and should be corrected or explained.
  2. [§4.2 vs §4.5] The threshold θ for accepting graph-matching results is set to 0.5 in the main evaluation (Section 4.2) but to 0.8 in the feature-interpretation experiment (Section 4.5); the manuscript should justify this difference and state whether the reported performance is sensitive to θ.
  3. [§3.2, Eq. 6] The text states that the predicted similarity ŝ lies in [0,1], but the cosine similarity of arbitrary embedding vectors can be negative; the manuscript should either clamp the output, use a sigmoid-like transformation, or note that negative values are treated as zero.
  4. [Algorithm 2] Step 2 of Algorithm 2 uses the notation ŝ(G_i^test, G_j^temp) with an index i that is not defined in the algorithm; this appears to be a typo for G_test.
  5. [§3.2, module 2] The sentence 'The feature embedding module utilizes graph convolutional networks (GCNs) and multi-layer perceptron (MLP) to capture a comprehensive representation of each arterial segment' refers to 'arterial segment,' which is a leftover from a prior coronary-artery application and should be replaced with 'RoI' or 'femoral region.'

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation chain; reported metrics are empirical on a disjoint test set, though the equal-RoI-count evaluation gap is a separate validity concern.

full rationale

The derivation chain is not circular. The ground-truth similarity in Eq. 7 is defined as label equality (1 if both graphs are from the same fracture-status group, 0 otherwise), and the network is trained to predict that similarity on randomly sampled pairs from the training split; at test time, Algorithm 2 applies the learned similarity to compare held-out test graphs with the separate template set and uses majority voting over template labels. Test labels are never used to fit the model or to choose templates, so the reported ACC 0.9970/SN 0.9869 are empirical results on a disjoint test set, not consequences of the definitions. The self-citations ([10], [14], [16], [17]) are methodological references for cohort exclusion, radiomics features, and graph-matching modules; the architecture is fully specified in Eqs. 1-7 and evaluated on UK Biobank data, so these citations are not load-bearing for the central claim. The paper does have an omitted specification: Section 3.3 trains only on equal-RoI-count pairs, and Section 4.5 reports only 3036 of the 5390 possible test-template pairs because it restricts to same-size graphs, leaving unstated how a test graph with no matching template size is classified; this is a correctness/generalization concern, not a circular reduction of the prediction to its inputs.

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

The paper introduces no new physical entities. Its central claim depends on a fitted threshold (theta=0.5), multiple tuned hyperparameters, and strong domain assumptions about feature sufficiency and annotation consistency.

free parameters (5)
  • theta (threshold) = 0.5
    Empirically set in Section 4.2; used to accept graph matches for majority voting.
  • d_intra and d_cross (hidden unit dimensions) = 256
    Tuned in Section 4.4; set equal for both embedding modules.
  • L (number of GCN layers) = 5
    Tuned in Section 4.4.
  • M (number of cross-graph embedding layers) = 3
    Tuned in Section 4.4.
  • k (number of nearest neighbors in kNN graph construction)
    The specific value of k is not stated; it is adjusted to ensure graph connectivity (Section 3.1), making it an unspecified free parameter.
assumptions (3)
  • domain assumption The 130 hand-crafted features, including BMD T-scores and clinical variables, are sufficient to predict hip fracture risk.
    These features are used as node attributes throughout; if BMD T-score alone drives the result, the graph-matching contribution is not isolated.
  • domain assumption The manual RoI annotations are consistent across subjects and the kNN graph with connectivity adjustment captures the spatial anatomy of the femur.
    Graph construction in Section 3.1 relies on manual LabelMe annotations and heuristic connectivity, with no inter-observer variability reported.
  • domain assumption The training and template sets are representative of the target population.
    The cohort is a selected subset of UK Biobank with exclusion criteria in Section 4.1, which may bias the results and limit generalizability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ICGM-FRAX: Iterative Cross Graph Matching for Hip Fracture Risk Assessment using Dual-energy X-ray Absorptiometry Images." pith.science (2026). https://pith.science/paper/VXI6MVSH

@misc{pith2026250415384,
  author       = {Pith},
  title        = {Pith review of: ICGM-FRAX: Iterative Cross Graph Matching for Hip Fracture Risk Assessment using Dual-energy X-ray Absorptiometry Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VXI6MVSH}},
  note         = {Machine review of arXiv:2504.15384}
}
read the original abstract

Hip fractures represent a major health concern, particularly among the elderly, often leading decreased mobility and increased mortality. Early and accurate detection of at risk individuals is crucial for effective intervention. In this study, we propose Iterative Cross Graph Matching for Hip Fracture Risk Assessment (ICGM-FRAX), a novel approach for predicting hip fractures using Dual-energy X-ray Absorptiometry (DXA) images. ICGM-FRAX involves iteratively comparing a test (subject) graph with multiple template graphs representing the characteristics of hip fracture subjects to assess the similarity and accurately to predict hip fracture risk. These graphs are obtained as follows. The DXA images are separated into multiple regions of interest (RoIs), such as the femoral head, shaft, and lesser trochanter. Radiomic features are then calculated for each RoI, with the central coordinates used as nodes in a graph. The connectivity between nodes is established according to the Euclidean distance between these coordinates. This process transforms each DXA image into a graph, where each node represents a RoI, and edges derived by the centroids of RoIs capture the spatial relationships between them. If the test graph closely matches a set of template graphs representing subjects with incident hip fractures, it is classified as indicating high hip fracture risk. We evaluated our method using 547 subjects from the UK Biobank dataset, and experimental results show that ICGM-FRAX achieved a sensitivity of 0.9869, demonstrating high accuracy in predicting hip fractures.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 50 canonical work pages

  1. [1]

    Department of Computer Science, Kennesaw State University, 680 Arntson Dr, Marietta, GA 30060

    work page

  2. [2]

    Department of Applied Computing, Michigan Technological University, 1400 Townsend Dr, Houghton, MI, 49931

    work page

  3. [3]

    Department of Radiological Sciences, Department of Biomedical Engineering, and Department of Mechanical and Aerospace Engineering, University of California, Irvine, CA 92697

    work page

  4. [4]

    Department of Internal Medicine and Division of Rheumatology, UC Davis Health, Sacramento, CA 95817

    work page

  5. [5]

    Geisel School of Medicine, Dartmouth College, Hanover, NH 03755

    work page

  6. [6]

    Division of Biomedical Informatics and Genomics, Tulane Center of Biomedical Informatics and Genomics, Deming Department of Medicine, Tulane University, New Orleans, LA 70112

    work page

  7. [7]

    Department of Mathematical Sciences, Michigan Technological University, Houghton, MI 49931

    work page

  8. [8]

    Center for Biocomputing and Digital Health, Institute of Computing and Cybersystems, and Health Research Institute, Michigan Technological University, Houghton, MI 49931 * Corresponding authors: Weihua Zhou, Ph.D. Department of Applied Computing, Michigan Technological University, 1400 Townsend Dr, Houghton, MI, 49931, USA Tel: 906-487-2666 E-Mail: whzhou...

    work page

Show all 53 references
  1. [9]

    The consequences of hip fracture are severe, leading to increased mortality, long-term disability, and substantial healthcare costs [2]

    Introduction Hip fractures represent a major public health challenge, particularly for the aging population [1]. The consequences of hip fracture are severe, leading to increased mortality, long-term disability, and substantial healthcare costs [2]. Most hip fractures occur in...

  2. [10]

    We propose a novel method that defines the DXA-derived semantic image as a graph and applies a graph matching algorithm to predict hip fractures by comparing the similarity between graphs derived from different subjects

  3. [11]

    The proposed ICGM -FRAX method demonstrates not only high performance but also high interpretability

  4. [12]

    The proposed method can also be applied to other diagnostic tasks using medical images where the connectivity and relationships between semantic RoIs are crucial

  5. [13]

    However, its ability to predict hip fractures specifically, while valuable, is limited when used in isolation

    Related Work 2.1 Hip Fracture Risk Assessment using DXA DXA has long been a standard imaging technique for measuring BMD, a key indicator of osteoporosis and fracture risk [6]. However, its ability to predict hip fractures specifically, while valuable, is limited when used in ...

  6. [14]

    The proposed ICGM-FRAX framework transforms hip fracture risk assessment into a problem of evaluating the similarity between a test DXA image and multiple template DXA images

    Methodology This study introduces an image -level classification approach for assessing hip fracture risk using DXA images. The proposed ICGM-FRAX framework transforms hip fracture risk assessment into a problem of evaluating the similarity between a test DXA image and multipl...

  7. [15]

    To construct individual graphs, a processing algorithm is applied to generate a graph representation for each DXA image

    Extracting features for nodes in individual graphs. To construct individual graphs, a processing algorithm is applied to generate a graph representation for each DXA image. In this representation, each node corresponds to a RoI within the femur. For each node, we extract pixel...

  8. [16]

    The feature embedding module utilizes graph convolutional networks (GCNs) and multi -layer perceptron (MLP) to capture a comprehensive representation of each arterial segment

    Intra-graph Feature Embedding. The feature embedding module utilizes graph convolutional networks (GCNs) and multi -layer perceptron (MLP) to capture a comprehensive representation of each arterial segment. In the context of a DXA derived graph 𝐺 , the intra -graph feature emb...

  9. [17]

    Cross-graph feature embedding is a critical process for improving the reliability of node correspondences between graphs

    Cross-graph Feature Embedding in Graph Pairs. Cross-graph feature embedding is a critical process for improving the reliability of node correspondences between graphs. Without interactive aggregation across graphs, direct node-to-node matching often lacks robustness [20]. This...

  10. [18]

    After the feature embedding process, graph pooling is applied to generate the average representation for each graph

    Global feature extraction. After the feature embedding process, graph pooling is applied to generate the average representation for each graph. Formally, the features for graph 𝐺𝑔 after applying the graph average pooling operation are calculated in Eq. 5. 𝑧𝑝𝑜𝑜𝑙 𝑔 = 1 𝑛 ∑(𝑧𝑖 𝑔)...

  11. [19]

    Specifically, the pooled feature embeddings for graph 𝐺1 and 𝐺2 are used to compute the similarity between the two graphs

    Graph similarity measurement. Specifically, the pooled feature embeddings for graph 𝐺1 and 𝐺2 are used to compute the similarity between the two graphs. In particular, the cosine similarity is calculated between the averaged embeddings of both graphs, serving as a measure of t...

  12. [20]

    Randomly select two individual graphs 𝐺𝑖 𝑡𝑟𝑎𝑖𝑛 and 𝐺𝑗 𝑡𝑟𝑎𝑖𝑛

  13. [21]

    Extract features for each RoI in 𝐺𝑖 𝑡𝑟𝑎𝑖𝑛 and 𝐺𝑗 𝑡𝑟𝑎𝑖𝑛

  14. [22]

    Perform intra-graph feature embedding for 𝐺𝑖 𝑡𝑟𝑎𝑖𝑛 and 𝐺𝑗 𝑡𝑟𝑎𝑖𝑛 using Eqs. 1 and 2

  15. [23]

    Perform cross-graph feature embedding between 𝐺𝑖 𝑡𝑟𝑎𝑖𝑛 and 𝐺𝑗 𝑡𝑟𝑎𝑖𝑛 using Eqs. 3 and 4

  16. [24]

    Calculate the graph similarity between 𝐺𝑖 𝑡𝑟𝑎𝑖𝑛 and 𝐺𝑗 𝑡𝑟𝑎𝑖𝑛 using Eqs. 5 and 6

  17. [25]

    hip - fracture

    Optimize ICGM using the objective function defined in Eq. 7. Testing: During testing, each graph from the testing set is compared against all graphs in the template set to compute the similarity score. If the similarity between a test graph and any template graph exceeds a pre...

  18. [26]

    Calculate the similarity between 𝐺𝑡𝑒𝑠𝑡 and 𝐺𝑗 𝑡𝑒𝑚𝑝 using trained ICGM, as 𝑠̂(𝐺𝑡𝑒𝑠𝑡, 𝐺𝑗 𝑡𝑒𝑚𝑝)

  19. [27]

    If 𝑠̂(𝐺𝑖 𝑡𝑒𝑠𝑡, 𝐺𝑗 𝑡𝑒𝑚𝑝) > 𝜃 then accept the graph matching results for majority voting

  20. [28]

    F1 Score provides a balance between precision and recall, especially useful in imbalanced datasets, defined as 𝐹1 = 2 × 𝑇𝑃 2×𝑇𝑃+𝐹𝑃+𝐹𝑁

    Assign labels for 𝐺𝑡𝑒𝑠𝑡 according to major voting among 𝐺𝑗 𝑡𝑒𝑚𝑝, 𝑗 ∈ {1, ⋯ , 𝑛𝑡𝑝} the total number of actual negatives , as 𝑆𝑃 = 𝑇𝑁 𝑇𝑁+𝐹𝑃. F1 Score provides a balance between precision and recall, especially useful in imbalanced datasets, defined as 𝐹1 = 2 × 𝑇𝑃 2×𝑇𝑃+𝐹𝑃+𝐹𝑁. 3.5...

  21. [29]

    Experimental Results and Analysis 4.1 Materials and Enrolled Subjects The study cohort comprised 547 subjects, with 94 individuals who experienced hip fractures from UKBiobank dataset . Similarity to our previous publication [10], subjects who did not undergo DXA scanning were...

  22. [31]

    Delaunay triangulation method creates the structured graph by connecting points to form triangles, ensuring that no point lies inside the circumcircle of any triangle in the mesh. This method guarantees that the resulting graph captures the spatial distribution of points in a ...

  23. [32]

    The Euclidean distance is calculated, and an edge is formed between the points if their distance is below a predefined threshold

    Distance-based threshold method involves generating a graph by connecting nodes based on the centroids of each RoIs. The Euclidean distance is calculated, and an edge is formed between the points if their distance is below a predefined threshold. This approach is effective in ...

  24. [33]

    Conclusion In this study, we propose a novel method, ICGM-FRAX, which defines the DXA-derived semantic image as a graph and applies a graph matching algorithm to predict hip fractures. This approach enables the model to compare the similarity between graphs derived f rom diffe...

  25. [34]

    Tian C, Shi L, Wang J, Zhou J, Rui C, Yin Y , Du W, Chang S, Rui Y . Global, regional, and national burdens of hip fractures in elderly individuals from 1990 to 2021 and predictions up to 2050: A systematic analysis of the Global Burden of Disease Study 2021 . Arch Gerontol Ge...

    1990

  26. [35]

    Quality of life in elderly people after a hip fracture: a prospective study

    Amarilla-Donoso FJ, López -Espuela F, Roncero -Martín R, Leal -Hernandez O, Puerto -Parejo LM, Aliaga-Vera I, Toribio-Felipe R, Lavado-García JM. Quality of life in elderly people after a hip fracture: a prospective study. Health Qual Life Outcomes. 2020 Mar 14;18:71. PMCID: P...

    2020

  27. [36]

    DXA -based statistical models of shape and intensity outperform aBMD hip fracture prediction: A retrospective study

    Aldieri A, Paggiosi M, Eastell R, Bignardi C, Audenino AL, Bhattacharya P, Terzini M. DXA -based statistical models of shape and intensity outperform aBMD hip fracture prediction: A retrospective study. Bone. 2024 May 1;182:117051

    2024

  28. [37]

    Clinical significance of trabecular bone score of DXA in hip fracture patients -comparative study between trochanteric fractures and neck fractures

    Kwon S, Yoo J, Yoon Y , Lee M, Hwang J. Clinical significance of trabecular bone score of DXA in hip fracture patients -comparative study between trochanteric fractures and neck fractures. BMC Musculoskelet Disord. 2024 Nov 13;25(1):908

    2024

  29. [38]

    Artificial Intelligence for Hip Fracture Detection and Outcome Prediction

    Lex JR, Di Michele J, Koucheki R, Pincus D, Whyne C, Ravi B. Artificial Intelligence for Hip Fracture Detection and Outcome Prediction. JAMA Netw Open. 2023 Mar 17;6(3):e233391. PMCID: PMC10024206

    2023

  30. [39]

    Dual -Energy X-Ray Absorptiometry

    Krugh M, Langaker MD. Dual -Energy X-Ray Absorptiometry. StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2025 [cited 2025 Mar 28]. Available from: http://www.ncbi.nlm.nih.gov/books/NBK519042/ PMID: 30085584

    2025

  31. [40]

    [cited 2025 Mar 28]

    Classification of Fracture Risk in Fallers Using Dual‐Energy X‐Ray Absorptiometry (DXA) Images and Deep Learning‐Based Feature Extraction - PMC [Internet]. [cited 2025 Mar 28]. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC10731096/

    2025

  32. [41]

    Improving the Hip Fracture Risk Prediction Through 2D Finite Element Models From DXA Images: Validation Against 3D Models

    Terzini M, Aldieri A, Rinaudo L, Osella G, Audenino AL, Bignardi C. Improving the Hip Fracture Risk Prediction Through 2D Finite Element Models From DXA Images: Validation Against 3D Models. Front Bioeng Biotechnol. 2019;7:220. PMCID: PMC6746936

    2019

  33. [42]

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

    Hsieh C-I, Zheng K, Lin C, Mei L, Lu L, Li W, Chen F -P, Wang Y , Zhou X, Wang F, Xie G, Xiao J, Miao S, Kuo C-F. Automated bone mineral density prediction and fracture risk assessment using plain radiographs via deep learning. Nat Commun. 2021 Sep 16;12:5472. PMCID: PMC8446034

    2021

  34. [43]

    A staged approach using machine learning and uncertainty quantification to predict the risk of hip fracture

    Shaik A, Larsen K, Lane NE, Zhao C, Su K-J, Keyak JH, Tian Q, Sha Q, Shen H, Deng H-W, Zhou W. A staged approach using machine learning and uncertainty quantification to predict the risk of hip fracture. Bone Rep. 2024 Sep;22:101805. [11]Yan J, Yin X-C, Lin W, Deng C, Zha H, Y...

    2024

  35. [44]

    Structure-aware siamese graph neural networks for encounter -level patient similarity learning

    Gu Y , Yang X, Tian L, Yang H, Lv J, Yang C, Wang J, Xi J, Kong G, Zhang W. Structure-aware siamese graph neural networks for encounter -level patient similarity learning. J Biomed Inform. 2022 Mar;127:104027

    2022

  36. [45]

    HAGMN-UQ: Hyper Association Graph Matching Network with Uncertainty Quantification for Coronary Artery Semantic Labeling

    Zhao C, Esposito M, Xu Z, Zhou W. HAGMN-UQ: Hyper Association Graph Matching Network with Uncertainty Quantification for Coronary Artery Semantic Labeling. Med Image Anal. 2024 Oct;103374

    2024

  37. [46]

    LabelMe: a database and web-based tool for image annotation

    Russell BC, Torralba A, Murphy KP, Freeman WT. LabelMe: a database and web-based tool for image annotation. Int J Comput Vis. Springer; 2008;77:157–173

    2008

  38. [47]

    AGMN: Association graph-based graph matching network for coronary artery semantic labeling on invasive coronary angiograms

    Zhao C, Xu Z, Jiang J, Esposito M, Pienta D, Hung G -U, Zhou W. AGMN: Association graph-based graph matching network for coronary artery semantic labeling on invasive coronary angiograms. Pattern Recognit. 2023 Nov;143:109789

    2023

  39. [48]

    Lung segmentation and automatic detection of COVID -19 using radiomic features from chest CT images

    Zhao C, Xu Y , He Z, Tang J, Zhang Y , Han J, Shi Y , Zhou W. Lung segmentation and automatic detection of COVID -19 using radiomic features from chest CT images. Pattern Recognit. Elsevier; 2021;119:108071

    2021

  40. [49]

    Ranking via Sinkhorn Propagation [Internet]

    Adams RP, Zemel RS. Ranking via Sinkhorn Propagation [Internet]. arXiv; 2011 [cited 2024 Jul 12]. Available from: http://arxiv.org/abs/1106.1925

    2011 arXiv

  41. [50]

    Visual Permutation Learning

    Cruz RS, Fernando B, Cherian A, Gould S. Visual Permutation Learning. IEEE Trans Pattern Anal Mach Intell. 2019 Dec 1;41(12):3100–3114

    2019

  42. [51]

    Neural Graph Matching for Pre-training Graph Neural Networks

    Hou Y , Hu B, Zhao WX, Zhang Z, Zhou J, Wen J-R, editors. Neural Graph Matching for Pre-training Graph Neural Networks. Proc 2022 SIAM Int Conf Data Min SDM [Internet]. Philadelphia, PA: Society for Industrial and Applied Mathematics; 2022 [cited 2024 Jul 11]. Available from: ...

    2022 doi

  43. [52]

    Combinatorial Learning of Robust Deep Graph Matching: an Embedding based Approach

    Wang R, Yan J, Yang X. Combinatorial Learning of Robust Deep Graph Matching: an Embedding based Approach. IEEE Trans Pattern Anal Mach Intell. 2020;1–1

    2020

  44. [53]

    Large-Scale Machine Learning with Stochastic Gradient Descent

    Bottou L. Large-Scale Machine Learning with Stochastic Gradient Descent. In: Lechevallier Y , Saporta G, editors. Proc COMPSTAT2010 [Internet]. Heidelberg: Physica-Verlag HD; 2010 [cited 2025 Mar 16]. p. 177–186. Available from: http://link.springer.com/10.1007/978-3-7908-2604...

    2010 doi

  45. [256]

    The number of GCN layers, i.e. 𝐿 in Eq. 1, was set as 5. And the number of cross -graph feature embedding, i.e. 𝑀 in Eq. 4, was set as 3. To compare performance, we set several baseline models and compared them with our previous staged model with ensemble [10]. For the baselin...

    1947

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.