arxiv: 2604.13820 · v1 · submitted 2026-04-15 · 💻 cs.CE

Recognition: unknown

Stable Long-Horizon Neural ODE Reduced-Order Models via Learned Feedback for Biological Growth and Remodeling

Joel Laudo , Adrian Buganza Tepole

Authors on Pith no claims yet

Pith reviewed 2026-05-10 12:13 UTC · model grok-4.3

classification 💻 cs.CE

keywords reduced-order modelsneural ODEtissue growthtissue remodelingskin expansionproper orthogonal decompositionclosed-loop feedbackbiomechanical modeling

0 comments

The pith

A closed-loop Neural ODE reduced-order model with CNN growth feedback stabilizes long-horizon skin growth predictions.

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

This paper introduces a reduced-order modeling approach using Neural Ordinary Differential Equations to simulate the coupled mechanics and growth of skin during tissue expansion procedures. The method compresses the displacement field into a low-dimensional space using Proper Orthogonal Decomposition and learns the evolution dynamics with a NODE that incorporates patient-specific parameters. To counteract error buildup in extended simulations, the authors add a feedback mechanism that extracts features from the growth field and reintroduces them into the dynamics at each step, testing scalar, linear, and nonlinear CNN representations. The CNN version proves most effective, bringing 90.3 percent of test cases within clinical tolerance for final skin area gain, compared to only 43.7 percent without feedback, all while delivering speedups exceeding 20,000 times over detailed finite element models. This framework points toward practical tools for rapid, patient-specific biomechanical forecasting in clinical settings.

Core claim

The authors present a Neural ODE reduced-order model that learns latent dynamics of coupled mechanical deformation and tissue growth in skin during tissue expansion. Displacement fields are compressed via Proper Orthogonal Decomposition into a low-dimensional latent space, after which a NODE learns the resulting dynamics conditioned on patient-specific parameters. To address long-horizon error accumulation, a closed-loop architecture feeds back encoded features of the evolving growth field into the dynamics at each step; comparisons show that nonlinear CNN-based representations stabilize rollouts far better than scalar or linear POD-based alternatives. The resulting model captures 90.3% of 0

What carries the argument

The closed-loop architecture in which encoded features of the evolving growth field, extracted via CNN, are fed back into the Neural ODE governing POD latent space dynamics at every time step.

Load-bearing premise

The POD-compressed latent space together with the learned feedback features from the growth field capture all essential coupled mechanical and growth dynamics without losing information critical to long-term stability or clinical accuracy.

What would settle it

Running the model on a new set of patient cases and finding that the percentage of simulations within clinical tolerance on final skin area gain falls below the open-loop baseline level of 43.7 percent would falsify the claim of improved stability from the feedback mechanism.

read the original abstract

Reduced-order models (ROMs) are essential for rapid simulation of complex biomechanical systems and for bridging the gap between high fidelity models and clinical application. However, ROMs for tissue growth and remodeling (G&R) remain largely unexplored. Here, we present a Neural Ordinary Differential Equation (NODE) ROM framework that learns latent dynamics of coupled mechanical deformation and tissue growth, demonstrated in the context of skin growth during tissue expansion (TE). TE is a challenging problem involving nonlinear contact, history-dependent material behavior, and mechanobiology driven growth. The displacement field is compressed via Proper Orthogonal Decomposition (POD) into a low-dimensional latent space, and a NODE learns the resulting dynamics conditioned on patient-specific parameters. To address long-horizon error accumulation, a key challenge in autoregressive latent dynamical models, we propose a closed-loop architecture in which encoded features of the evolving growth field are fed back into the dynamics at each step. We compare feedback representations of increasing expressiveness: scalar, linear POD-based, and nonlinear CNN-based. The CNN-based growth feature feedback substantially stabilizes long-horizon rollouts. The best model captures 90.3% of validation cases within clinical tolerance based on the final skin area gain, compared to 43.7% for the open-loop baseline. Moreover, the NODE ROM achieves over 20000x the speed of full finite element simulations. More broadly, these results suggest that selectively retaining inexpensive physics of the state evolution and feeding features from these fields back into the latent dynamical system is a promising strategy for stable and accurate ROMs of G&R in biological tissues.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Feedback from the growth field stabilizes the Neural ODE ROM enough to reach 90% clinical tolerance on long rollouts, but the POD step may still drop localized mechanics that matter for growth.

read the letter

Here's the main thing: this work shows that feeding back encoded features of the growth field into a Neural ODE can keep reduced-order models stable over long times in biological growth and remodeling, lifting the fraction of cases within clinical tolerance from 43.7% to 90.3% on their skin expansion example. The new piece is the closed-loop architecture. They reduce the displacement field with POD, train a NODE on the latent dynamics, and then try scalar, POD, and CNN encodings of the growth field as feedback. The CNN one gives the biggest stability boost. This is a direct response to the error buildup that hits most autoregressive latent models, and they get a 20,000 times speedup over full finite element runs. The results are presented clearly with a concrete metric on final skin area gain. That makes the improvement easy to judge. The potential issue is whether the POD latent space plus the growth feedback captures everything needed. POD is a linear reduction, so it might miss sharp local strain or stress patterns that matter for the growth law. The paper does not appear to include a detailed look at reconstruction error in those critical areas or tests on how the model behaves when those features dominate. If the feedback cannot recover the lost information, the gains could shrink on more varied cases. This paper is for researchers in computational biomechanics who want faster models for patient-specific growth predictions. People working on Neural ODEs or ROMs for coupled physics will see value in the feedback strategy. It deserves a serious referee. The approach is grounded and the gains are quantified, even if some additional checks on the latent representation would help. I would send it for review.

Referee Report

2 major / 2 minor

Summary. The paper proposes a Neural ODE reduced-order model for coupled mechanical deformation and tissue growth in skin expansion, compressing displacements via POD into a latent space, learning dynamics with a NODE conditioned on patient parameters, and adding closed-loop feedback from growth-field features (scalar, POD, or CNN) to mitigate long-horizon error accumulation. It reports that CNN feedback raises the fraction of validation cases meeting clinical tolerance on final skin area gain from 43.7% (open-loop) to 90.3%, while delivering >20,000× speedup over full FEM simulations.

Significance. If the reported performance holds under scrutiny, the work demonstrates a practical route to stable, fast ROMs for history-dependent mechanobiological problems that have so far resisted reduced-order treatment. The systematic comparison of feedback representations and the emphasis on long-horizon stability address a recognized weakness of autoregressive latent models; the speedup figure is large enough to matter for clinical translation. The approach of selectively feeding inexpensive physics-derived features back into the latent dynamics is a reusable idea beyond the specific TE application.

major comments (2)

[§3.1 and §4.2] §3.1 (POD compression) and §4.2 (validation metrics): no reconstruction-error analysis is provided for the displacement field in high-growth regions or near expander boundaries. Because POD is a linear global basis, localized strain concentrations that directly enter the growth law may be lost; without quantitative evidence that these features are recoverable from the retained modes plus CNN growth feedback, the 90.3 % clinical-tolerance claim rests on an unverified assumption.
[§4.1] §4.1 (experimental setup): the manuscript gives insufficient detail on the number of training/validation simulations, the precise definition of “clinical tolerance” (e.g., allowable % error in area gain), the hyperparameter search procedure, and the train/validation split strategy. These omissions prevent independent verification of the 90.3 % vs 43.7 % comparison and of the claimed robustness outside the training distribution.

minor comments (2)

[§3.2] Notation for the growth-field encoder (CNN vs linear POD) is introduced without a compact equation or diagram; a single schematic would clarify the three feedback variants compared in the results.
[Abstract and §4.3] The abstract states “over 20000x the speed,” but the main text does not report wall-clock timings or hardware details for the full-order model; adding these would strengthen the practical claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed comments. We address each major comment point by point below and will revise the manuscript to incorporate the requested clarifications and analyses.

read point-by-point responses

Referee: [§3.1 and §4.2] §3.1 (POD compression) and §4.2 (validation metrics): no reconstruction-error analysis is provided for the displacement field in high-growth regions or near expander boundaries. Because POD is a linear global basis, localized strain concentrations that directly enter the growth law may be lost; without quantitative evidence that these features are recoverable from the retained modes plus CNN growth feedback, the 90.3 % clinical-tolerance claim rests on an unverified assumption.

Authors: We agree that a quantitative reconstruction-error analysis for the displacement field, focused on high-growth regions and near expander boundaries, is a valuable addition to strengthen the validation of the POD step. Although the reported clinical tolerance is defined on the integrated final skin area gain (a global metric), we acknowledge that local strain recovery is relevant to the mechanobiological growth law. In the revised manuscript we will add relative L2-norm reconstruction errors for the displacement field in these critical subdomains, computed both with and without the CNN feedback, to demonstrate that the retained modes plus growth features suffice for the quantities entering the growth law. revision: yes
Referee: [§4.1] §4.1 (experimental setup): the manuscript gives insufficient detail on the number of training/validation simulations, the precise definition of “clinical tolerance” (e.g., allowable % error in area gain), the hyperparameter search procedure, and the train/validation split strategy. These omissions prevent independent verification of the 90.3 % vs 43.7 % comparison and of the claimed robustness outside the training distribution.

Authors: We thank the referee for highlighting these omissions. In the revised manuscript we will explicitly state the total number of training and validation simulations, the precise numerical definition of clinical tolerance (allowable percentage error on final skin area gain), the hyperparameter search procedure (including ranges and optimization method), and the train/validation split strategy used to assess generalization. These additions will allow independent reproduction and verification of the reported performance figures. revision: yes

Circularity Check

0 steps flagged

No circularity: performance metrics arise from held-out evaluation, not construction

full rationale

The paper trains a NODE on POD-compressed displacement trajectories from finite-element simulations of tissue expansion, then augments the latent dynamics with learned feedback encodings (scalar, POD, or CNN) of the growth field. The headline result—90.3 % of validation cases within clinical tolerance on final skin area gain versus 43.7 % for the open-loop baseline—is obtained by rolling out the trained model on held-out simulation cases whose ground-truth trajectories were never seen during training or hyper-parameter selection. No equation defines the reported accuracy in terms of the fitted parameters themselves, no self-citation supplies a uniqueness theorem that forces the architecture, and the feedback features are optimized against a standard trajectory loss rather than the final-area metric. Consequently the evaluation remains an independent test of generalization rather than a tautological restatement of the training objective.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The framework rests on standard assumptions about POD optimality for linear subspaces and the ability of NODEs to learn continuous dynamics from discrete simulation snapshots; no new physical entities are postulated.

free parameters (2)

neural network weights and biases
Learned during training to approximate latent dynamics and feedback mappings from simulation data.
POD basis truncation rank
Chosen to balance compression and reconstruction accuracy for the displacement field.

axioms (2)

domain assumption POD yields an optimal low-dimensional linear representation of the displacement snapshots
Invoked to compress the high-fidelity mechanical state into the latent space used by the NODE.
domain assumption Neural ODEs can accurately integrate learned vector fields over long horizons when conditioned on auxiliary features
Underlies the choice of NODE for the latent dynamics.

pith-pipeline@v0.9.0 · 5590 in / 1543 out tokens · 30271 ms · 2026-05-10T12:13:58.681919+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

47 extracted references · 20 canonical work pages

[1]

Plastic and Reconstructive Surgery19(2), 124–130 (1957)

Neumann, C.G.: The expansion of an area of skin by progressive distention of a subcutaneous balloon: Use of the method for securing skin for subtotal reconstruction of the ear. Plastic and Reconstructive Surgery19(2), 124–130 (1957)

1957
[2]

Journal of Theoretical Biology297, 166–175 (2012) https://doi.org/10.1016/j.jtbi.2011.12.022

Z¨ ollner, A.M., Buganza Tepole, A., Kuhl, E.: On the biomechanics and mechanobiology of growing skin. Journal of Theoretical Biology297, 166–175 (2012) https://doi.org/10.1016/j.jtbi.2011.12.022

work page doi:10.1016/j.jtbi.2011.12.022 2012
[3]

Clinics in Plastic Surgery32(1), 35–44 (2005)

Rivera, R., LoGiudice, J., Gosain, A.K.: Tissue expansion in pediatric patients. Clinics in Plastic Surgery32(1), 35–44 (2005). Clinics in Plastic Surgery

2005
[4]

Plastic and reconstructive surgery124(1), 1–13 (2009)

Arneja, J.S., Gosain, A.K.: Giant congenital melanocytic nevi. Plastic and reconstructive surgery124(1), 1–13 (2009)

2009
[5]

Plastic and reconstructive surgery69(2), 195–206 (1982)

Radovan, C.: Breast reconstruction after mastectomy using the temporary expander. Plastic and reconstructive surgery69(2), 195–206 (1982)

1982
[6]

Annals of Medicine and Surgery21, 34–44 (2017) 23

Bertozzi, N., Pesce, M., Santi, P., Raposio, E.: Tissue expansion for breast recon- struction: Methods and techniques. Annals of Medicine and Surgery21, 34–44 (2017) 23

2017
[7]

JAMA Surgery150(1), 9–16 (2015)

Kummerow, K.L., Du, L., Penson, D.F., Shyr, Y., Hooks, M.A.: Nationwide trends in mastectomy for early-stage breast cancer. JAMA Surgery150(1), 9–16 (2015)

2015
[8]

Journal of Craniofacial Surgery14(6), 866–872 (2003)

LoGiudice, J., Gosain, A.K.: Pediatric tissue expansion: indications and compli- cations. Journal of Craniofacial Surgery14(6), 866–872 (2003)

2003
[9]

British journal of plastic surgery41(3), 239–250 (1988)

Antonyshyn, O., Gruss, J., Mackinnon, S., Zuker, R.: Complications of soft tissue expansion. British journal of plastic surgery41(3), 239–250 (1988)

1988
[10]

Journal of the Mechanics and Physics of Solids59(10), 2177–2190 (2011)

Tepole, A.B., Ploch, C.J., Wong, J., Gosain, A.K., Kuhl, E.: Growing skin: a computational model for skin expansion in reconstructive surgery. Journal of the Mechanics and Physics of Solids59(10), 2177–2190 (2011)

2011
[11]

Journal of the Mechanical Behavior of Biomedical Materials152, 106453 (2024) https: //doi.org/10.1016/j.jmbbm.2024.106453

Mathur, M., Malinowski, M., Jazwiec, T., Timek, T.A., Rausch, M.K.: Leaflet remodeling reduces tricuspid valve function in a computational model. Journal of the Mechanical Behavior of Biomedical Materials152, 106453 (2024) https: //doi.org/10.1016/j.jmbbm.2024.106453

work page doi:10.1016/j.jmbbm.2024.106453 2024
[12]

Journal of the Royal Society Interface16(157) (2019)

Ambrosi, D., Ben Amar, M., Cyron, C.J., DeSimone, A., Goriely, A., Humphrey, J.D., Kuhl, E.: Growth and remodelling of living tissues: perspectives, challenges and opportunities. Journal of the Royal Society Interface16(157) (2019)

2019
[13]

Acta Biomaterialia198, 267–280 (2025)

Laudo, J., Han, T., Figueroa, A.E., Ledwon, J., Gosain, A.K., Lee, T., Buganza Tepole, A.: Development and calibration of digital twins for human skin growth in tissue expansion. Acta Biomaterialia198, 267–280 (2025)

2025
[14]

Journal of Biomechanical Engineering147(7), 071002 (2025)

Laudo, J., Han, T., Ledwon, J., Figueroa, A.E., Gosain, A.K., Lee, T., Tepole, A.B.: Predictive modeling of human skin deformation and growth during tissue expansion in postmastectomy breast reconstruction. Journal of Biomechanical Engineering147(7), 071002 (2025)

2025
[15]

Computer Methods and Programs in Biomedicine105(1), 1–12 (2012)

Niroomandi, S., Alfaro, I., Cueto, E., Chinesta, F.: Accounting for large defor- mations in real-time simulations of soft tissues based on reduced-order models. Computer Methods and Programs in Biomedicine105(1), 1–12 (2012)

2012
[16]

Journal of Biomechanics54, 33–43 (2017)

Chang, G.H., Schirmer, C.M., Modarres-Sadeghi, Y.: A reduced-order model for wall shear stress in abdominal aortic aneurysms by proper orthogonal decomposition. Journal of Biomechanics54, 33–43 (2017)

2017
[17]

Annals of Biomedical Engineering52, 208–225 (2024) https://doi.org/10.1007/s10439-023-03360-5

Shah, I., Samaee, M., Razavi, A., Esmailie, F., Ballarin, F., Dasi, L.P., Veneziani, A.: Reduced order modeling for real-time stent deformation simulations of tran- scatheter aortic valve prostheses. Annals of Biomedical Engineering52, 208–225 (2024) https://doi.org/10.1007/s10439-023-03360-5

work page doi:10.1007/s10439-023-03360-5 2024
[18]

Journal of Computational Physics532, 113957 (2025) https://doi.org/10.1016/j.jcp.2025.113957 24

Puri, V., Prakash, A., Kara, L.B., Zhang, Y.J.: Snf-rom: Projection-based nonlin- ear reduced order modeling with smooth neural fields. Journal of Computational Physics532, 113957 (2025) https://doi.org/10.1016/j.jcp.2025.113957 24

work page doi:10.1016/j.jcp.2025.113957 2025
[19]

https://arxiv.org/abs/2102

Rojas, C.J.G., Dengel, A., Ribeiro, M.D.: Reduced-order Model for Fluid Flows via Neural Ordinary Differential Equations (2021). https://arxiv.org/abs/2102. 02248

2021
[20]

npj Computational Materials8(1), 190 (2022) https://doi.org/10

Oommen, V., Shukla, K., Goswami, S., Dingreville, R., Karniadakis, G.E.: Learn- ing two-phase microstructure evolution using neural operators and autoencoder architectures. npj Computational Materials8(1), 190 (2022) https://doi.org/10. 1038/s41524-022-00869-3

2022
[21]

Nature Communications15(1), 1834 (2024) https://doi.org/10.1038/ s41467-024-46177-7

Regazzoni, F., Pagani, S., Salvador, M., Dede’, L., Quarteroni, A.: Learning the intrinsic dynamics of spatio-temporal processes through latent dynamics networks. Nature Communications15(1), 1834 (2024) https://doi.org/10.1038/ s41467-024-46177-7

2024
[22]

Frontiers in Ar- tificial Intelligence6(2023)

Lu, H., Tartakovsky, D.M.: Drips: A framework for dimension reduction and interpolation in parameter space. Journal of Computational Physics493, 112455 (2023) https://doi.org/10.1016/j.jcp.2023.112455

work page doi:10.1016/j.jcp.2023.112455 2023
[23]

International Journal of Solids and Structures301, 112966 (2024) https://doi.org/10.1016/j.ijsolstr.2024.112966

Jones, R.E., Hamel, C.M., Bolintineanu, D., Johnson, K., Buarque de Macedo, R., Fuhg, J., Bouklas, N., Kramer, S.: Multiscale simulation of spatially correlated microstructure via a latent space representation. International Journal of Solids and Structures301, 112966 (2024) https://doi.org/10.1016/j.ijsolstr.2024.112966

work page doi:10.1016/j.ijsolstr.2024.112966 2024
[24]

https://arxiv.org/abs/2503

Charles, P., Ray, D., Yu, Y., Prins, J., Melchers, H., Abdelmalik, M.R.A., Cochran, J., Oberai, A.A., Hughes, T.J.R., Larson, M.G.: An optimal Petrov- Galerkin framework for operator networks (2025). https://arxiv.org/abs/2503. 04024

2025
[25]

URL https://www.sciencedirect.com/ science/article/pii/S0021999122009019

Linot, A.J., Burby, J.W., Tang, Q., Balaprakash, P., Graham, M.D., Maulik, R.: Stabilized neural ordinary differential equations for long-time forecasting of dynamical systems. Journal of Computational Physics474, 111838 (2023) https: //doi.org/10.1016/j.jcp.2022.111838

work page doi:10.1016/j.jcp.2022.111838 2023
[26]

Computer Methods in Applied Mechanics and Engineering449, 118526 (2026) https://doi.org/10.1016/j.cma.2025.118526

Hamdi, E., Lejeune, E.: Towards robust surrogate models: Benchmarking machine learning approaches to expediting phase field simulations of brittle fracture. Computer Methods in Applied Mechanics and Engineering449, 118526 (2026) https://doi.org/10.1016/j.cma.2025.118526

work page doi:10.1016/j.cma.2025.118526 2026
[27]

Garikipati, K.: Reduced-Order Models: Numerical Homogenization for the Elastic Response of Material Microstructures, pp. 49–79. Springer, Cham (2024). https:// doi.org/10.1007/978-3-031-62029-4-5 .https://doi.org/10.1007/978-3-031-62029- 4-5

work page doi:10.1007/978-3-031-62029-4-5 2024
[28]

Journal of Elasticity61(1-3), 1–48 (2006) 25

Gasser, T.C., Ogden, R.W., Holzapfel, G.A.: A new constitutive framework for arterial wall mechanics and a comparative study of material models. Journal of Elasticity61(1-3), 1–48 (2006) 25

2006
[29]

International Journal of Non-Linear Mechanics47(8), 938–949 (2012)

Buganza Tepole, A., Gosain, A.K., Kuhl, E.: Stretching skin: The physiological limit and beyond. International Journal of Non-Linear Mechanics47(8), 938–949 (2012)

2012
[30]

Acta Biomaterialia137, 136–146 (2022)

Han, T., Lee, T., Ledwon, J., Vaca, E., Turin, S., Kearney, A., Gosain, A.K., Tepole, A.B.: Bayesian calibration of a computational model of tissue expansion based on a porcine animal model. Acta Biomaterialia137, 136–146 (2022)

2022
[31]

Group Normalization

Wu, Y., He, K.: Group Normalization (2018). https://arxiv.org/abs/1803.08494

work page Pith review arXiv 2018
[32]

In: Advances in Neural Information Processing Systems (NeurIPS), vol

Kidger, P., Morrill, J., Foster, J., Lyons, T.: Neural controlled differential equations for irregular time series. In: Advances in Neural Information Processing Systems (NeurIPS), vol. 33, pp. 6696–6707 (2020)

2020
[33]

Progress of Theoretical Physics33(3), 423–455 (1965)

Mori, H.: Transport, collective motion, and brownian motion*). Progress of Theoretical Physics33(3), 423–455 (1965)

1965
[34]

Journal of Statistical Physics9(3), 215–220 (1973) https://doi.org/10.1007/BF01008729

Zwanzig, R.W.: Nonlinear generalized langevin equations. Journal of Statistical Physics9(3), 215–220 (1973) https://doi.org/10.1007/BF01008729

work page doi:10.1007/bf01008729 1973
[35]

arXiv preprint arXiv:1808.04258 (2018)

Ma, C., Wang, J., E, W.: Model reduction with memory and the machine learning of dynamical systems. arXiv preprint arXiv:1808.04258 (2018)

work page arXiv 2018
[36]

Journal of Computational Physics410, 109402 (2020)

Recurrent neural network closure of parametric pod-galerkin reduced-order mod- els based on the mori-zwanzig formalism. Journal of Computational Physics410, 109402 (2020)

2020
[37]

In: Pro- ceedings of the 5th Annual Conference on Learning for Dynamics and Control (L4DC)

Miao, K., Gatsis, K.: Learning robust state observers using neural odes. In: Pro- ceedings of the 5th Annual Conference on Learning for Dynamics and Control (L4DC). Proceedings of Machine Learning Research, vol. 211, pp. 1–12 (2023)

2023
[38]

Under review at ICLR 2026 (2026)

Anonymous Authors: Learning the Koopman Operator Using Attention-Free Transformers. Under review at ICLR 2026 (2026)

2026
[39]

Computers & Fluids307, 106958 (2026) https://doi.org/10.1016/j.compfluid.2025.106958

Xu, W., Karthikeyakannan, M., McComb, C., Grande Guti´ errez, N.: Energy- based feature extraction with adaptive local domain decomposition for prediction of transient and turbulence flow with operator regression models. Computers & Fluids307, 106958 (2026) https://doi.org/10.1016/j.compfluid.2025.106958

work page doi:10.1016/j.compfluid.2025.106958 2026
[40]

International Journal for Numerical Methods in Engineering (2022) https://doi

Kollepara, K.S., Navarro-Jim´ enez, J.M., Le Guennec, Y., Silva, L., Aguado, J.V.: On the limitations of low-rank approximations in contact mechanics problems. International Journal for Numerical Methods in Engineering (2022) https://doi. org/10.1002/nme.7118

work page doi:10.1002/nme.7118 2022
[41]

International Journal for Numerical Methods in Engineering126, 70041 (2025) 26 https://doi.org/10.1002/nme.70041

Ritzert, S., Kehls, J., Reese, S., Brepols, T.: Component-based model-order reduc- tion with mortar tied contact for nonlinear quasi-static mechanical problems. International Journal for Numerical Methods in Engineering126, 70041 (2025) 26 https://doi.org/10.1002/nme.70041

work page doi:10.1002/nme.70041 2025
[42]

arXiv preprint arXiv:2505.18050 (2025)

Manvelyan-Stroot, D., Filanova, Y., Duff, I.P., Benner, P., Wever, U.: Inference of substructured reduced-order models for dynamic contact from contact-free simulations. arXiv preprint arXiv:2505.18050 (2025)

work page arXiv 2025
[43]

Journal of Biomechanics49(14), 3727–3733 (2016) https://doi.org/10.1016/j.jbiomech.2016.09.020

Fey, N.P., Klute, G.K., Neptune, R.R.: A reduced-order computational model of a semi-active variable-stiffness foot prosthesis. Journal of Biomechanics49(14), 3727–3733 (2016) https://doi.org/10.1016/j.jbiomech.2016.09.020

work page doi:10.1016/j.jbiomech.2016.09.020 2016
[44]

Journal of Biomechanical Engineering144(12), 121005 (2022)

Han, T., Ahmed, K.S., Gosain, A.K., Tepole, A.B., Lee, T.: Multi-fidelity gaussian process surrogate modeling of pediatric tissue expansion. Journal of Biomechanical Engineering144(12), 121005 (2022)

2022
[45]

Computers in Biology and Medicine133, 104394 (2021) https: //doi.org/10.1016/j.compbiomed.2021.104394

Jiang, Z., Choi, J., Baek, S.: Machine learning approaches to surrogate multifi- delity growth and remodeling models for efficient abdominal aortic aneurysmal applications. Computers in Biology and Medicine133, 104394 (2021) https: //doi.org/10.1016/j.compbiomed.2021.104394

work page doi:10.1016/j.compbiomed.2021.104394 2021
[46]

Biomechanics and Modeling in Mechanobiology24, 169–183 (2025) https://doi

Reiber, M., Bensel, F., Zheng, Z., Nackenhorst, U.: A non-intrusive reduced-order model for finite element analysis of implant positioning in total hip replacements. Biomechanics and Modeling in Mechanobiology24, 169–183 (2025) https://doi. org/10.1007/s10237-024-01903-w

work page doi:10.1007/s10237-024-01903-w 2025
[47]

Archives of Plastic Surgery50(6), 685–692 (2023) https://doi.org/10.5999/aps.2023.00486 27

Celik, S.,et al.: The impact of subpectoral tissue expander applications on tho- racic deformity: Retrospective analytic study on mastectomy patients. Archives of Plastic Surgery50(6), 685–692 (2023) https://doi.org/10.5999/aps.2023.00486 27

work page doi:10.5999/aps.2023.00486 2023