Pith. sign in

REVIEW 3 major objections 6 minor 53 references

TumorTwin: A python framework for patient-specific digital twins in oncology

T0 review · 3 major / 6 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 TumorTwin presents a reusable, open-source pipeline that turns MRI data into patient-specific tumor digital twins for forecasting treatment response.

desk verdict A useful modular framework for tumor digital twins, but the validation is self-referential and the ADC-to-cellularity formula in Eq. (1) looks mis-specified. read the letter →

arxiv 2505.00670 v1 pith:CMWOY2F2 submitted 2025-05-01 physics.med-ph cs.MS

Michael Kapteyn , Anirban Chaudhuri , Ernesto A. B. F. Lima , Graham Pash , Rafael Bravo , Karen Willcox , Thomas E. Yankeelov , David A. Hormuth II This is my paper · ORCID
classification physics.med-phcs.MS
keywords digitaltwinoncologytumorgrowthmodelmagneticresonanceimagingcalibrationreaction-diffusionsoftwareframeworkpatient-specificmodeling
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

TumorTwin is a Python framework whose central claim is that a single open-source codebase can support the whole lifecycle of a patient-specific tumor digital twin, from raw MRI and treatment records to calibrated growth forecasts and treatment decisions. The paper argues that this is the first framework explicitly designed as an end-to-end, data-to-decisions digital-twin infrastructure working from tissue-scale imaging data. A sympathetic reader would care because this fills a practical gap: existing modeling libraries stop at simulation, leaving researchers to assemble their own data processing and calibration pipelines for each cancer site. The package demonstrates the workflow on synthetic high-grade-glioma and triple-negative-breast-cancer cases, recovering known model parameters to high relative accuracy and projecting tumor response under standard-of-care therapies.

What carries the argument

The load-bearing machinery is a modular abstraction over the imaging-to-decision pipeline, built on three cooperating objects. A validated PatientData structure encodes MRI and treatment history, using the ADC-to-cellularity mapping $N(x,t) = (ADC_w - ADC(x,t))/(ADC(x,t)-ADC_{\min})$ to turn diffusion-weighted images into tumor cellularity. A TumorGrowthModel3D implements the reaction-diffusion equation $\partial N/\partial t = \nabla \cdot (D\nabla N) + k N (1 - N/\theta)$ with chemotherapy decay and an instantaneous linear-quadratic radiotherapy term. The model is wrapped in a differentiable ODE solver whose adjoint backward pass computes gradients of any scalar output with respect to the patient parameters, and a suite of optimizers, including Levenberg-Marquardt, performs calibration with those gradients. The connection of these three objects, rather than any single equation, is what carries the framework's end-to-end claim.

What would settle it

Run the same calibration and holdout-prediction workflow on longitudinal MRI from real patients with recorded outcomes, freeing all model parameters rather than fixing them to generating truth; if the recovered parameters are unstable or the holdout predictions do not beat a simple tumor-volume extrapolation baseline, the framework's clinical forecasting claim is not supported.

Watch

Extended reading notes

Core claim

TumorTwin is introduced as the first framework explicitly designed as an end-to-end, data-to-decisions digital-twin infrastructure for oncology that works directly from tissue-scale imaging data. The framework's central data object holds imaging and treatment history; its model, solver, and optimizer modules are interchangeable; and its differentiable solver computes adjoint-based gradients, making gradient calibration feasible on grids of roughly two million voxels. In the demonstration, calibrating four patient-specific parameters (proliferation rate, diffusion coefficient, chemotherapy efficacy, and radiation sensitivity) against five synthetic high-grade-glioma imaging visits recovers the generating parameter values with relative errors between $10^{-4}$ and $10^{-6}$, and the calibrated model predicts the held-out imaging visits. The same workflow is repeated for a synthetic triple-negative-breast-cancer case, and the paper states that once calibrated, the model can be used to test alternative treatment schedules or be re-calibrated as new MRI data arrive.

Load-bearing premise

The demonstrations assume that synthetic longitudinal images, generated by evolving the framework's own reaction-diffusion model on atlas or real-patient anatomy, thresholded, converted to ADC, and calibrated with every non-fitted parameter fixed to its generating value, stand in for real patient imaging closely enough to support the claimed data-to-decisions utility.

Editorial extensions

If this is right

  • Researchers can swap data sources, growth models, solvers, or optimizers without rebuilding the pipeline, so modeling choices can be compared systematically in one codebase.
  • A calibrated digital twin can simulate alternative chemoradiation schedules by editing dose and timing, giving a route to treatment optimization on a per-patient basis.
  • Performance numbers suggest clinically relevant turnaround: a one-year tumor growth simulation takes about a minute for a forward pass and one to two minutes for an adjoint backward pass on consumer-grade GPU hardware or a high-performance CPU.
  • Because the same workflow was applied to glioma and breast cancer, the framework is positioned as a pan-cancer foundation rather than a disease-site-specific code.
  • Open-source release with included datasets and tutorials lowers the barrier for other groups to reproduce, extend, or adapt the digital-twin pipeline.

Reading between the lines

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

  • The in silico demonstrations are closed-loop self-tests: the data are generated by the same model that is later calibrated, so they establish internal consistency and computational performance rather than predictive validity on real tissue; the decisive next step is calibration on real longitudinal MRI.
  • Fixing all non-calibrated parameters to their generating truth sidesteps identifiability and model-form error, so a useful stress test would be to calibrate all parameters simultaneously or allow spatially varying proliferation and diffusion fields.
  • If the modular abstractions hold, the same pipeline could be extended to other solid tumors wherever an imaging-derived cellularity or observable map can be defined, turning the framework into a general testbed for mechanistic oncology models.
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 / 6 minor

Summary. TumorTwin is an open-source Python/PyTorch framework for building image-guided oncology digital twins. The paper describes a modular architecture with PatientData, TumorGrowthModel, Solver, and Optimizer components; a reaction-diffusion tumor growth model with chemotherapy and radiotherapy terms; adjoint-based gradient computation; and gradient-based calibration using a Levenberg-Marquardt optimizer. The framework is demonstrated on two synthetic datasets, one for high-grade glioma and one for triple-negative breast cancer, generated by simulating the same reaction-diffusion model on an atlas or an I-SPY2 subject. The paper also reports CPU/GPU performance benchmarks for forward and backward solves, and claims that TumorTwin is the first end-to-end data-to-decisions digital twin framework for tissue-scale imaging data.

Significance. If the claims hold, TumorTwin would be a valuable community resource: it is publicly available, includes documentation and tutorials, and provides a modular, composable pipeline that could lower the barrier to experimenting with different tumor models, solvers, and optimization algorithms. The performance measurements are careful, with multiple runs and standard deviations, and the use of GPU acceleration and adjoint methods is a practical strength. However, the scientific demonstrations are limited to synthetic data generated by the same equations being calibrated, with non-calibrated parameters fixed to their generating values. The central claim of patient-specific predictive utility therefore rests on self-consistency rather than on independent validation. The potential mis-specification of the ADC-to-cellularity map in Eq. (1) further undermines the imaging-to-model link that is central to the framework's stated purpose.

major comments (3)
  1. [Section 2.2.2, Eq. (1)] The observation operator N = (ADCw − ADC)/(ADC − ADCmin) is unbounded: as ADC approaches ADCmin, N diverges, whereas the model in Eq. (4) and the total tumor cell count in Eq. (8) treat N as a bounded normalized density. The standard relation in the cited protocol literature (Jarrett et al., Nat. Protoc. 2021) uses the denominator (ADCw − ADCmin), which is bounded and has the correct limits at ADCw and ADCmin. Because Eq. (1) is used both to generate the synthetic ADC images (Methods 4.2) and to convert real patient ADC into cellularity, the unbounded form changes the mapping between imaging and model state, making the calibrated parameters {k, D, α1, αRT} not comparable with previous work. The synthetic demonstrations use the same operator for both generation and calibration, so they cannot distinguish a self-consistent but incorrect observation model from a correct one. Please correct the formula or provide a justification for the unbounded form, and rerun the demonstrations if the formula changes.
  2. [Sections 2.2.6 and 4.2] The calibration and holdout 'prediction' results are self-consistency checks rather than independent validation. The synthetic longitudinal datasets are produced by integrating the same reaction-diffusion equations (4)-(5) that are later calibrated, and all non-calibrated parameters are fixed to their ground-truth generating values (Section 2.2.6; Methods 4.2). The held-out data are therefore generated by the very equations being tested, so the demonstrations show that the software correctly recovers known parameters and reproduces its own forward simulations, but they do not establish that TumorTwin can forecast real patient-specific tumor growth or treatment response. The abstract and discussion should be tempered accordingly, or the paper should include validation on independent real clinical data.
  3. [Section 2.2.6] The calibration demonstration is artificially well-posed: the optimizer is initialized at 20% of the true parameter values and all other parameters are fixed to their generating truth values. This does not reflect the realistic situation in which parameters such as the carrying capacity, chemotherapy decay rate, and radiosensitivity βRT are unknown or spatially heterogeneous, and no identifiability or sensitivity analysis is provided. The manuscript should acknowledge this limitation explicitly and, ideally, include tests with misspecified fixed parameters or unknown additional parameters to assess the robustness of the calibration pipeline.
minor comments (6)
  1. [Eq. (8)] The units in the total tumor cell count formula are inconsistent: if N is a normalized density and θ is the carrying capacity, then θ should have units of cells/mm^3 rather than cells, and the formula TTC = sum_i N_i θ V_voxel should be clarified; if N is already a cell density, the factor θ is unnecessary.
  2. [Methods 4.2] The text states 'D = 0.1 mm2/s' for both the HGG and TNBC cases, but Eq. (4) defines D with units mm^2/day; a diffusion coefficient of 0.1 mm^2/s is physically implausible for tumor invasion, and the numerical timestep is given as 'δt = 0.1 day−1' instead of 'day'. Please correct these units.
  3. [Methods 4.2] There is a duplicated phrase in the TNBC preprocessing description: 'using a rigid registration using a rigid registration using imregtform' should be 'using a rigid registration performed with imregtform'.
  4. [Section 3, Discussion] The sentence 'The are several opportunities for further development' contains a typo and should read 'There are several opportunities'.
  5. [Table 1] On the server, the forward GPU solves are slower than the CPU solves (speedup 0.78–0.86), which is not highlighted in the text; the claim of general GPU acceleration should be qualified by noting that the benefit depends on problem size and hardware.
  6. [Section 1, Background] The claim that TumorTwin is 'the first' end-to-end digital twin framework for tissue-scale imaging data is strong and is not supported by a systematic comparison with existing software frameworks; please substantiate this novelty claim or soften it.

Circularity Check

2 steps flagged · score 6.0 of 10

Holdout 'predictions' are generated by the same reaction-diffusion model and the same ADC-to-cellularity operator used in calibration, so the in silico demonstrations reduce to self-consistency checks.

  1. fitted input called prediction [Section 2.2.6 (Model calibration) and Methods 4.2 (Generation of synthetic longitudinal imaging studies)]
    "For this example, we know the ground-truth model parameters (used to generate the dataset), but the optimizer is initiated with an initial guess of k = 0.01, D = 0.02, α1 = 0.01, αRT = 0.04 (20% of the respective truth values). Other parameters are fixed to their ground truth values. ... we calibrate the model to the first five imaging timepoints, while holding out the remaining images to assess predictive accuracy."

    The held-out images are produced by integrating the same reaction-diffusion model, Eqs. (4)-(5), with exactly these ground-truth parameter values (Methods 4.2: 'seeded an artificial tumor ... and evolved it using Eqs. (4)-(5) with the following parameter values'). Calibrating to a subset of such data and then forecasting the remaining same-model timepoints guarantees agreement up to numerical optimization tolerance. The 'predictive accuracy' demonstrated in Figures 4 and 6 is therefore a code-correctness check; the fitted parameters return to ground truth precisely because the data were generated from those parameters. This does not externally validate the model or the DT's ability to forecast real patient tumors.

  2. self definitional [Equation (1) and Methods 4.2]
    "Using the ADC to cellularity function, the observation at each visit {ti} is defined as o(ti) = N(x,ti) = (ADCw − ADC(x,ti)) / (ADC(x,ti) − ADCmin). ... We then used the segmentations, N(x,t), and Equation (1) to calculate the ADC for each corresponding N(x,t) map."

    The same Eq. (1) is inverted to manufacture the synthetic ADC images from N and then applied forward to convert those images back into N for calibration. Thus the observation operator used in the loss function is, by construction, the inverse of the operator used to generate the data. Any error in the assumed ADC-to-cellularity map is invisible in this loop: the calibrator sees exactly the state that produced the images, so agreement between predicted and observed cellularity maps is enforced by definition rather than established by independent imaging evidence.

full rationale

The paper is primarily a software contribution, and its architecture, modular abstractions, solver interfaces, gradient computation, and CPU/GPU performance benchmarks are self-contained and do not reduce to their inputs. However, the demonstration of DT 'predictive accuracy' is circular in two linked ways. First, the synthetic HGG and TNBC longitudinal datasets are generated by numerically integrating the very reaction-diffusion PDE (Eqs. 4-5) that is later calibrated to them, with all non-fitted parameters fixed to the generating ground-truth values. Fitting parameters to the first few timepoints and then 'predicting' later timepoints is therefore a guaranteed recovery of the generating trajectory. Second, the imaging-to-model observation map, Eq. (1), is used in both directions: it converts model cellularity into synthetic ADC images, and it converts those ADC images back into cellularity observations for calibration. The prediction and the observations are linked by construction, so the demonstrations cannot detect misspecification of the observation operator or of the growth model. The paper is honest that the datasets are synthetic and framed as demonstrations of functionality, which keeps this from being a fraudulent or fully tautological validation, but it is still a partial circularity: the claimed demonstration of patient-specific prediction reduces to verifying numerical self-consistency of the code. Self-citations to prior work for the model and ADC mapping are present but are not the main circular mechanism; the reduction is the data-generation/calibration loop itself. No external real-patient validation is performed, so the biological predictive claim remains unsupported by this paper's demonstrations.

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

The central demonstrations rest on 4 fitted parameters and 3 domain assumptions. No new physical entities are introduced. The main evidential burden is that the synthetic data are produced by the same equations being fitted, so the fitted parameters and holdout predictions inherit the model's assumptions rather than testing them.

free parameters (4)
  • proliferation rate k = HGG truth 0.05 day^-1, recovered to 10^-4 to 10^-6 relative error
    Calibrated by Levenberg-Marquardt against synthetic MRI-derived cellularity maps, starting from 20% of the truth value.
  • diffusion coefficient D = HGG truth 0.1 mm2/s as stated in Methods, recovered to similar tolerance
    Calibrated in both HGG and TNBC demonstrations; the unit is inconsistent with the model definition in Section 2.2.3.
  • chemotherapy efficacy alpha_1 = HGG truth 0.2 day^-1, recovered to 10^-4 to 10^-6 relative error
    Calibrated in both demonstrations; represents the sensitivity of tumor cells to the chemotherapy agent.
  • radiosensitivity alpha_RT = HGG truth 0.05 Gy^-1, recovered to similar tolerance
    Calibrated in the HGG case; fixed to zero in the TNBC case because no radiation is delivered.
assumptions (3)
  • domain assumption ADC-to-cellularity mapping in Eq. (1) relates measured ADC to tumor cell density N.
    Used to initialize u0 and define observational data; no independent validation of this mapping is provided in this paper.
  • domain assumption Reaction-diffusion model with logistic growth, exponential chemotherapy decay, and linear-quadratic radiation adequately represents tumor response.
    Core model in Eqs. (4)-(5) is taken from prior author work and is not derived or independently validated here.
  • ad hoc to paper Synthetic imaging data generated using the same model are representative of real longitudinal patient data.
    Section 4.2 generates 'observations' by integrating the model to be calibrated, then seeds the optimizer from perturbed parameters; this makes the demonstration a code-correctness test rather than an external validation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TumorTwin: A python framework for patient-specific digital twins in oncology." pith.science (2026). https://pith.science/paper/CMWOY2F2

@misc{pith2026250500670,
  author       = {Pith},
  title        = {Pith review of: TumorTwin: A python framework for patient-specific digital twins in oncology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CMWOY2F2}},
  note         = {Machine review of arXiv:2505.00670}
}
read the original abstract

Background: Advances in the theory and methods of computational oncology have enabled accurate characterization and prediction of tumor growth and treatment response on a patient-specific basis. This capability can be integrated into a digital twin framework in which bi-directional data-flow between the physical tumor and the digital tumor facilitate dynamic model re-calibration, uncertainty quantification, and clinical decision-support via recommendation of optimal therapeutic interventions. However, many digital twin frameworks rely on bespoke implementations tailored to each disease site, modeling choice, and algorithmic implementation. Findings: We present TumorTwin, a modular software framework for initializing, updating, and leveraging patient-specific cancer tumor digital twins. TumorTwin is publicly available as a Python package, with associated documentation, datasets, and tutorials. Novel contributions include the development of a patient-data structure adaptable to different disease sites, a modular architecture to enable the composition of different data, model, solver, and optimization objects, and CPU- or GPU-parallelized implementations of forward model solves and gradient computations. We demonstrate the functionality of TumorTwin via an in silico dataset of high-grade glioma growth and response to radiation therapy. Conclusions: The TumorTwin framework enables rapid prototyping and testing of image-guided oncology digital twins. This allows researchers to systematically investigate different models, algorithms, disease sites, or treatment decisions while leveraging robust numerical and computational infrastructure.

Figures

Figures reproduced from arXiv: 2505.00670 by the authors.

Figure 1
Figure 1. TumorTwin workflow key components. TumorTwin takes as inputs the patient configuration file, MRI data, and an initial guess of model parameters. The patient configuration file and MRI data are used to construct a PatientData object which serves as the central data object throughout TumorTwin. The PatientData object is used to initialize a patient-specific TumorGrowthModel, which can be combined with a ForwardSolver … view at source ↗
Figure 2
Figure 2. Patient data summary graphics. Example output of the patient data summary function applied to the in silico HGG dataset. This graphic shows treatment and imaging schedule (top panel), anatomical T1-weighted MRI with associated tumor segmentations (middle panel), and the apparent diffusion coefficient (ADC) map for the same imaging slices (bottom panel). In practice, this summary can be used to visually confirm longi… view at source ↗
Figure 3
Figure 3. Example tumor growth prediction for the in silico HGG patient. The predicted total tumor cell count over time is shown (top panel), with 2D slices of the model solution at three snapshots in time (bottom panel). Note the invasion, logistic growth, chemotherapy, and radiotherapy effects in the solution. 2.2.5 Efficient gradient computation In addition to solving a model forward in time to predict tumor growth for a g… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Model calibration to patient-specific MRI data. Top: Total tumor cell count (TTC) time series for each iteration of the calibration (black lines; opacity increased with iteration) compared with the observed TTC (purple stars). Middle: Evolution of a central tumor slice…
Figure 5
Figure 5. Figure 5: Solver performance profiling for CPU and GPU architectures. Top: Mean and standard deviation of the wall-clock time required to compute a forward model prediction for 180 days. Bottom: Results for a backward adjoint solve for the same period. Two CPU-based and two GPU-…
Figure 6
Figure 6. Figure 6: Model calibration to patient-specific MRI data - Triple-negative breast cancer example. Top: Total tumor cell count (TTC) time series for each iteration of the calibration (black lines; opacity increased with iteration) compared with the observed TTC (purple stars). Mi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 48 canonical work pages

  1. [1]

    Abbasi, S

    A. Abbasi, S. Amjad-Iranagh, and B. Dabir. Cellsys: An open-source tool for building initial structures for bio-membranes and drug-delivery systems. Journal of Computational Chemistry , 43(5):331–339, 2022

    work page 2022

  2. [2]

    Bakas, M

    S. Bakas, M. Reyes, A. Jakab, S. Bauer, M. Rempfler, A. Crimi, R. T. Shinohara, C. Berger, S. M. Ha, M. Rozycki, et al. Identifying the best machine learning algorithms for brain tumor segmentation, progres- sion assessment, and overall survival prediction in the brats challenge. arXiv preprint arXiv:1811.02629 , 2018

    arXiv 2018

  3. [3]

    R. R. Bravo, E. Baratchart, J. West, R. O. Schenck, A. K. Miller, J. Gallaher, C. D. Gatenbee, D. Bas- anta, M. Robertson-Tessi, and A. R. Anderson. Hybrid automata library: A flexible platform for hybrid modeling with real-time visualization. PLoS computational biology, 16(3):e1007635, 2020

    work page 2020

  4. [4]

    Chaudhuri, B

    A. Chaudhuri, B. Kramer, M. Norton, J. O. Royset, and K. Willcox. Certifiable risk-based engineering design optimization. AIAA Journal, 60(2):551–565, 2022

    work page 2022

  5. [5]

    Chaudhuri, G

    A. Chaudhuri, G. Pash, D. A. Hormuth II, G. Lorenzo, M. Kapteyn, C. Wu, E. A. B. F. Lima, T. E. Yan- keelov, and K. Willcox. Predictive Digital Twin for Optimizing Patient-Specific Radiotherapy Regimens under Uncertainty in High-Grade Gliomas. Frontiers in Artificial Intelligence , 6, 2023

    work page 2023

  6. [6]

    R. T. Q. Chen. torchdiffeq, 2018

    2018

  7. [7]

    X. Chen, R. M. Summers, and J. Yao. Kidney tumor growth prediction by coupling reaction–diffusion and biomechanical model. IEEE Transactions on Biomedical Engineering , 60(1):169–173, 2013

    work page 2013

  8. [8]

    Christenson, C

    C. Christenson, C. Wu, D. A. Hormuth II, C. E. Stowers, M. LaMonica, J. Ma, G. M. Rauch, and T. E. Yankeelov. Fast model calibration for predicting the response of breast cancer to chemotherapy using proper orthogonal decomposition. Journal of Computational Science , 82:102400, 2024

    work page 2024

Show all 53 references
  1. [9]

    J. R. Dormand and P. J. Prince. A family of embedded runge-kutta formulae. Journal of computational and applied mathematics , 6(1):19–26, 1980

    1980

  2. [10]

    Ghaffarizadeh, R

    A. Ghaffarizadeh, R. Heiland, S. H. Friedman, S. M. Mumenthaler, and P. Macklin. Physicell: An open source physics-based cell simulator for 3-d multicellular systems. PLoS computational biology , 14(2):e1005991, 2018

    2018

  3. [11]

    Ghattas and K

    O. Ghattas and K. Willcox. Learning physics-based models from data: perspectives from inverse problems and model reduction. Acta Numerica, 30:445–554, 2021. 14 Figure 6: Model calibration to patient-specific MRI data - Triple-negative breast cancer example . Top: Total tumor c...

    2021

  4. [12]

    Hairer, S

    E. Hairer, S. P. Nørsett, and G. Wanner. Solving ordinary differential equations I (2nd revised. ed.): nonstiff problems. Springer-Verlag, Berlin, Heidelberg, 1993

    1993

  5. [13]

    Hawkins-Daarud, S

    A. Hawkins-Daarud, S. Prudhomme, K. G. van der Zee, and J. T. Oden. Bayesian calibration, validation, and uncertainty quantification of diffuse interface models of tumor growth. Journal of Mathematical Biology, 67(6):1457–1485, 2013

    2013

  6. [14]

    D. A. Hormuth, K. A. Al Feghali, A. M. Elliott, T. E. Yankeelov, and C. Chung. Image-based personaliza- tion of computational models for predicting response of high-grade glioma to chemoradiation. Scientific reports, 11(1):8520, 2021

    2021

  7. [15]

    D. A. Hormuth, M. Farhat, C. Christenson, B. Curl, C. Chad Quarles, C. Chung, and T. E. Yankeelov. Opportunities for improving brain cancer treatment outcomes through imaging-based mathematical mod- eling of the delivery of radiotherapy and immunotherapy. Advanced Drug Deliver...

  8. [16]

    D. A. Hormuth, A. M. Jarrett, and T. E. Yankeelov. Forecasting tumor and vasculature response dynamics to radiation therapy via image based mathematical modeling. Radiation Oncology, 15:1–14, 2020

    2020

  9. [17]

    D. A. Hormuth, J. A. Weis, S. L. Barnes, M. I. Miga, E. C. Rericha, V. Quaranta, and T. E. Yankeelov. A mechanically coupled reaction–diffusion model that incorporates intra-tumoural heterogeneity to predict in vivo glioma growth. Journal of The Royal Society Interface , 14(12...

    2017

  10. [18]

    D. A. Hormuth II, K. A. A. Feghali, A. M. Elliott, T. Yankeelov, and C. Chung. Image-based personaliza- tion of computational models for predicting response of high-grade glioma to chemoradiation. Scientific Reports, 11:1–14, 2021. Publisher: Nature Publishing Group UK ISBN: 4...

    2021

  11. [19]

    T. M. Inc. MATLAB version: R2024b, 2024

    2024

  12. [20]

    A. M. Jarrett, A. S. Kazerouni, C. Wu, J. Virostko, A. G. Sorace, J. C. DiCarlo, D. A. Hormuth, D. A. Ekrut, D. Patt, B. Goodgame, S. Avery, and T. E. Yankeelov. Quantitative magnetic resonance imaging and tumor forecasting of breast cancer patients in the community setting. N...

    2021

  13. [21]

    Kidger, R

    P. Kidger, R. T. Q. Chen, and T. J. Lyons. ”hey, that’s not an ode”: Faster ode adjoints via seminorms. International Conference on Machine Learning , 2021

    2021

  14. [22]

    A. R. Kutuva, J. J. Caudell, K. Yamoah, H. Enderling, and M. U. Zahid. Mathematical modeling of radiotherapy: impact of model selection on estimating minimum radiation dose for tumor control. Frontiers in Oncology, Volume 13 - 2023, 2023

    2023

  15. [23]

    Laubenbacher, B

    R. Laubenbacher, B. Mehrad, I. Shmulevich, and N. Trayanova. Digital twins in medicine. Nature Computational Science, 4(3):184–191, 2024

    2024

  16. [24]

    Leder, K

    K. Leder, K. Pitter, Q. LaPlant, D. Hambardzumyan, B. D. Ross, T. A. Chan, E. C. Holland, and F. Mi- chor. Mathematical modeling of pdgf-driven glioblastoma reveals optimized radiation dosing schedules. Cell, 156(3):603–616, 2014

    2014

  17. [25]

    R. J. LeVeque. Finite difference methods for ordinary and partial differential equations: steady-state and time-dependent problems. SIAM, 2007

    2007

  18. [26]

    W. Li, D. C. Newitt, J. Gibbs, L. J. Wilmes, E. F. Jones, V. A. Arasu, F. Strand, N. Onishi, A. A.- T. Nguyen, J. Kornak, B. N. Joe, E. R. Price, H. Ojeda-Fournier, M. Eghtedari, K. W. Zamora, S. A. Woodard, H. Umphrey, W. Bernreuter, M. Nelson, and N. M. Hylton. I-spy 2 breas...

    2022

  19. [27]

    Liang, J

    B. Liang, J. Tan, L. Lozenski, D. A. Hormuth, T. E. Yankeelov, U. Villa, and D. Faghihi. Bayesian inference of tissue heterogeneity for individualized prediction of glioma growth. IEEE Transactions on Medical Imaging, 42(10):2865–2875, 2023

    2023

  20. [28]

    E. Lima, J. Oden, B. Wohlmuth, A. Shahmoradi, D. Hormuth II, T. Yankeelov, L. Scarabosio, and T. Horger. Selection and validation of predictive models of radiation effects on tumor growth based on noninvasive imaging data. Computer methods in applied mechanics and engineering ...

    2017

  21. [29]

    Lipkova, P

    J. Lipkova, P. Angelikopoulos, S. Wu, E. Alberts, B. Wiestler, C. Diehl, C. Preibisch, T. Pyka, S. E. Combs, P. Hadjidoukas, K. Van Leemput, P. Koumoutsakos, J. Lowengrub, and B. Menze. Personalized radiotherapy design for glioblastoma: Integrating mathematical tumor models, m...

    2019

  22. [30]

    Lipkov´ a, B

    J. Lipkov´ a, B. Menze, B. Wiestler, P. Koumoutsakos, and J. S. Lowengrub. Modelling glioma progres- sion, mass effect and intracranial pressure in patient anatomy. Journal of the Royal Society Interface , 19(188):20210922, 2022

    2022

  23. [31]

    Lorenzo, D

    G. Lorenzo, D. A. Hormuth II, C. Wu, G. Pash, A. Chaudhuri, E. A. Lima, L. C. Okereke, R. Patel, K. Willcox, and T. E. Yankeelov. Validating the predictions of mathematical models describing tumor growth and treatment response. arXiv preprint arXiv:2502.19333 , 2025

    2025 arXiv

  24. [32]

    S. J. McMahon. The linear quadratic model: usage, interpretation and challenges. Physics in Medicine & Biology, 64(1):01TR01, 2018

    2018

  25. [33]

    Metzcar, C

    J. Metzcar, C. R. Jutzeler, P. Macklin, A. K¨ ohn-Luque, and S. C. Br¨ uningk. A review of mechanistic learning in mathematical oncology. Frontiers in Immunology, 15, 2024

    2024

  26. [34]

    G. R. Mirams, C. J. Arthurs, M. O. Bernabeu, R. Bordas, J. Cooper, A. Corrias, Y. Davit, S.-J. Dunn, A. G. Fletcher, D. G. Harvey, et al. Chaste: an open source c++ library for computational physiology and biology. PLoS computational biology, 9(3):e1002970, 2013

    2013

  27. [35]

    National Academies of Sciences, and Medicine

    National Academy of Engineering and E. National Academies of Sciences, and Medicine. Foundational Research Gaps and Future Directions for Digital Twins. The National Academies Press, Washington, DC, 2023

    2023

  28. [36]

    D. C. Newitt, S. C. Partridge, Z. Zhang, J. Gibbs, T. Chenevert, M. Rosen, P. Bolan, H. Marques, J. Romanoff, L. Cimino, B. N. Joe, H. Umphrey, H. Ojeda-Fournier, B. Dogan, K. Y. Oh, H. Abe, J. Drukteinis, L. J. Esserman, and N. M. Hylton. Acrin 6698/i-spy2 breast dwi [data se...

    2021

  29. [37]

    Newman, J

    W. Newman, J. Verweij, H. Rosing, K. Grunberg, S. Chattopadhyay, E. Gamelin, J. Klastersky, and E. K. Rowinsky. Pharmacokinetics of temozolomide: an oral cytotoxic agent with activity in the central nervous system. Clinical Cancer Research, 2(8):1105–1111, 1996

    1996

  30. [38]

    Niarakis, R

    A. Niarakis, R. Laubenbacher, G. An, Y. Ilan, J. Fisher, ˚A. Flobak, K. Reiche, M. Rodr´ ıguez Mart´ ınez, L. Geris, L. Ladeira, et al. Immune digital twins for complex human pathologies: applications, limitations, and challenges. NPJ systems biology and applications , 10(1):141, 2024

    2024

  31. [39]

    Quarteroni and A

    A. Quarteroni and A. Valli. Numerical approximation of partial differential equations, volume 23. Springer Science & Business Media, 2008

    2008

  32. [40]

    Rohlfing, N

    T. Rohlfing, N. M. Zahr, E. V. Sullivan, and A. Pfefferbaum. The sri24 multichannel atlas of normal adult human brain structure. Human Brain Mapping , 31(5):798–819, 2010

    2010

  33. [41]

    K. Sel, A. Hawkins-Daarud, A. Chaudhuri, D. Osman, A. Bahai, D. Paydarfar, K. Willcox, C. Chung, and R. Jafari. Survey and perspective on verification, validation, and uncertainty quantification of digital twins for precision medicine. npj Digital Medicine , 8(1):40, 2025. 17

    2025

  34. [42]

    E. A. Stahlberg, M. Abdel-Rahman, B. Aguilar, A. Asadpoure, R. A. Beckman, L. L. Borkon, J. N. Bryan, C. M. Cebulla, Y. H. Chang, A. Chatterjee, J. Deng, S. Dolatshahi, O. Gevaert, E. J. Greenspan, W. Hao, T. Hernandez-Boussard, P. R. Jackson, M. Kuijjer, A. Lee, P. Macklin, S...

    2022

  35. [43]

    Starruß, W

    J. Starruß, W. De Back, L. Brusch, and A. Deutsch. Morpheus: a user-friendly modeling environment for multiscale and multicellular systems biology. Bioinformatics, 30(9):1331–1332, 2014

    2014

  36. [44]

    Stupp, W

    R. Stupp, W. P. Mason, M. J. van den Bent, M. Weller, B. Fisher, M. J. Taphoorn, K. Belanger, A. A. Brandes, C. Marosi, U. Bogdahn, J. Curschmann, R. C. Janzer, S. K. Ludwin, T. Gorlia, A. Allgeier, D. Lacombe, G. Cairncross, E. Eisenhauer, and R. O. Mirimanoff. Radiotherapy p...

    2005

  37. [45]

    A. Swan, T. Hillen, J. C. Bowman, and A. D. Murtha. A patient-specific anisotropic diffusion model for brain tumour spread. Bulletin of Mathematical Biology , 80(5):1259–1291, May 2018

    2018

  38. [46]

    M. H. Swat, G. L. Thomas, J. M. Belmonte, A. Shirinifard, D. Hmeljak, and J. A. Glazier. Multi-scale modeling of tissues using compucell3d. In Methods in cell biology , volume 110, pages 325–366. Elsevier, 2012

    2012

  39. [47]

    Tisue and U

    S. Tisue and U. Wilensky. Netlogo: A simple environment for modeling complexity. In International conference on complex systems, volume 21, pages 16–21. Citeseer, 2004

    2004

  40. [48]

    Tyrrell Rockafellar and J

    R. Tyrrell Rockafellar and J. O. Royset. Engineering decisions under risk averseness. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering , 1(2):04015003, 2015

    2015

  41. [49]

    Woodhams, S

    R. Woodhams, S. Ramadan, P. Stanwell, S. Sakamoto, H. Hata, M. Ozaki, S. Kan, and Y. Inoue. Diffusion- weighted imaging of the breast: principles and clinical applications. Radiographics, 31(4):1059–1084, 2011

    2011

  42. [50]

    C. Wu, D. A. Hormuth, G. Lorenzo, A. M. Jarrett, F. Pineda, F. M. Howard, G. S. Karczmar, and T. E. Yankeelov. Towards patient-specific optimization of neoadjuvant treatment protocols for breast cancer based on image-guided fluid dynamics. IEEE Transactions on Biomedical Engin...

    2022

  43. [51]

    C. Wu, G. Lorenzo, D. A. Hormuth, E. A. B. F. Lima, K. P. Slavkova, J. C. DiCarlo, J. Virostko, C. M. Phillips, D. Patt, C. Chung, and T. E. Yankeelov. Integrating mechanism-based modeling with biomedical imaging to build practical digital twins for clinical oncology. Biophysi...

    2022

  44. [52]

    Zahid, A

    M. Zahid, A. Mohamed, J. Caudell, L. Harrison, C. Fuller, E. Moros, and H. Enderling. Dynamics- adapted radiotherapy dose (dard) for head and neck cancer radiotherapy dose personalization. Journal of Personalized Medicine, 11:1124, 2021. 18

    2021

  45. [2022]

    Publisher: Elsevier B.V

Pith tools

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