Mechanistic Learning with Guided Diffusion Models to Predict Spatio-Temporal Brain Tumor Growth

Andreas Rauschecker; Catherine R. Jutzeler; Daria Laslo; Efthymios Georgiou; Marius George Linguraru; Sabine Muller; Sarah Bruningk

REVIEW 3 major objections 6 minor 25 references

Mechanistic Learning with Guided Diffusion Models to Predict Spatio-Temporal Brain Tumor Growth

T0 review · 3 major / 6 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read A mechanistic ODE forecast of tumor area, used as a conditioning target for a gradient-guided DDIM, generates follow-up brain MRIs whose tumor growth probability maps match observed growth direction, as measured by 95th percentile Hausdorff

desk verdict A sensible hybrid of ODE tumor-burden forecasting and guided DDIM, but the directionality claim needs a size-matched baseline before it stands. read the letter →

arxiv 2509.09610 v1 pith:DKZ2IKYY submitted 2025-09-11 cs.CV cs.AI

Daria Laslo , Efthymios Georgiou , Marius George Linguraru , Andreas Rauschecker , Sabine Muller , Catherine R. Jutzeler , Sarah Bruningk This is my paper · ORCID

classification cs.CVcs.AI

keywords diffusionmodelstumorgrowthpredictionmechanisticlearningpediatricdiffusemidlinegliomalongitudinalMRIsynthesisprobabilitymapsordinarydifferentialequationsHausdorffdistance

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

This paper tries to show that a hybrid pipeline can predict how a brain tumor will grow in space and time from a recent MRI and a mechanistic model of tumor dynamics. It combines an ordinary differential equation model, which forecasts tumor area after radiotherapy, with a denoising diffusion implicit model steered by a regressor gradient toward that forecasted size. The authors test the approach on pediatric diffuse midline glioma, a rare and fatal tumor with sparse longitudinal imaging data. They report that synthesized follow-up scans preserve anatomy, and that tumor growth probability maps produced by aggregating bootstrap-conditioned generations are significantly closer to the true follow-up segmentation than the starting scan is, by 95th percentile Hausdorff distance. The value, if correct, is that biologically informed image generation becomes possible in data-limited clinical settings.

What carries the argument

The load-bearing object is the gradient of a trained regressor R(x_l, l) that predicts tumor size relative to brain area from a noisy image. This gradient is added to the DDIM denoising step (Eq. 2), scaled by a dynamic factor s_R(dyn) that depends on the gap between the regressor's current output and the ODE-extrapolated target size, times a constant scale. The ODE model itself partitions the tumor at radiotherapy onset into surviving and dying compartments, with delayed exponential decay for the dying fraction, and is fitted per patient and per slice with bootstrap resampling to produce a distribution of target sizes. The machinery's job is to convert a scalar forecast (tumor area at follo

What would settle it

A direct test is to replace the ODE-extrapolated target size with a size-matched but direction-neutral target (e.g., the true final area, or a shuffled size from another patient) and measure whether thresholded growth probability maps still reduce HD95 relative to the initial segmentation. If an isotropic enlargement of the initial tumor mask to the same area achieves equal or better HD95, then the reported directionality reduces to a size effect.

Watch

Extended reading notes

Core claim

The central claim is that mechanistic model estimates of future tumor burden can serve as conditioning targets for a gradient-guided DDIM, and that doing so yields realistic follow-up MRI scans in which the direction of tumor growth is clinically relevant. The authors introduce tumor growth probability maps, computed by averaging binarized difference maps across generations using bootstrap-sampled target sizes, and show that thresholded versions of these maps have a significantly lower HD95 to the true target segmentation than the initial tumor mask does. They interpret this as evidence that the generated growth is both anatomically feasible and directionally aligned with observed progressio

Load-bearing premise

The whole directionality claim rests on the regressor's gradient, trained only on static single-time-point images, encoding spatial growth direction when pushed toward a larger tumor-area scalar; if that gradient encodes size increase without anatomical direction, the HD95 improvement is just a size artifact.

Editorial extensions

If this is right

Follow-up MRIs can be generated from a recent scan plus an ODE-based area forecast, without needing a large longitudinal training set for the target population.
Tumor growth probability maps can be thresholded to produce predicted future tumor segmentations for treatment planning.
The per-slice, per-patient ODE fits give time-continuous forecasts of tumor response to radiotherapy, including the timing of delayed cell death.
Bootstrapped target sizes yield dynamic probability maps with lower HD95 variability than static maps produced with a single target size.
The framework offers generative-space-time predictions that incorporate mechanistic priors, which is relevant for data-scarce pediatric cancers like DMG.

Reading between the lines

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

If the regressor gradient is indeed the carrier of directional information, a testable extension is to visualize the gradient field on held-out images and check whether its support concentrates on tumor-adjacent tissue; diffuse support would predict limited transfer to 3D volumes or tumors with less distinct margins.
The bootstrap-to-probability-map aggregation is a general recipe: any scalar biomarker forecast from a mechanistic model could be converted into a spatial probability map over image changes, potentially useful for other diseases with sparse longitudinal imaging.
A natural next experiment is to compare the generated growth maps against a purely data-driven spatio-temporal diffusion model trained on the same longitudinal pairs; the mechanistic prior should improve data efficiency and directional consistency, and that comparison would isolate the contribution of the ODE.
Because the ODE parameters are per patient and per slice, the framework yields a quantitative, time-continuous tumor response curve that could be correlated with clinical endpoints beyond imaging, such as progression-free survival.

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.

Desk Editor's Note

read the letter

The paper is a plausible integration of a per-slice ODE tumor-growth model with gradient-guided DDIM, plus bootstrap-aggregated growth probability maps, evaluated on a small in-house longitudinal DMG cohort. The combination is new, the code link appears, and the idea of using a mechanistic size forecast as a guidance target is reasonable. The probability-map aggregation is a nice touch. Credit where due: the ODE module is simple but evaluated with bootstrap and nRMSE on held-out time points, and the diffusion training on BraTS is standard.

The headline claim about directionality, though, is not yet supported. The guidance signal is the gradient of a scalar regressor that predicts tumor area relative to brain area. That gradient carries no explicit spatial-direction information. The diffusion model can satisfy it by expanding the tumor in almost any direction, as long as the area increases. The reported HD95 improvement over the initial mask is then expected even for directionally wrong growth, because a larger overlapping mask reduces HD95 purely by size. Without a size-matched baseline—e.g., isotropically dilating the initial mask to the same target area—the HD95 comparison cannot isolate directionality. That is the load-bearing flaw.

There is also a concrete inconsistency: Eq. (6) defines lambda'(t) = lambda_decay * tanh(...), while Algorithm 1 uses lambda_eff = -lambda * tanh(...). Those are different models, and the sign conventions differ. The paper never reconciles them. The static evaluation in Sec. 3.3 conditions on the true target size, which removes the ODE's forecast role; the dynamic maps use ODE sizes but still suffer the same size-effect issue. The hyperparameter grid search was done on the same Zurich cohort; the paper says the 8 evaluation patients are 'additional,' but it should state explicitly that they are disjoint from the 29 used for tuning.

These are fixable. The right fix is to add a non-directional size-control baseline (isotropic dilation, or a diffusion run with the same target size but no guidance beyond area) and show the generated direction beats that baseline. Also fix the lambda inconsistency and report error bars over slices. The writing is honest about limitations, and the work is serious, not a toy.

Who is this for: researchers working on mechanistic generative models for longitudinal medical imaging. As written, I would not let the directionality claim stand, but the framework deserves referee time. Recommendation: send to peer review with a request for the size-matched baseline and the equation fix.

Referee Report

3 major / 6 minor

Summary. The paper proposes a hybrid 'mechanistic learning' framework for spatio-temporal brain tumor growth prediction. A mechanistic ODE model, including exponential pre-RT growth, compartmental survival/death fractions at RT onset, and a tanh-delayed decay term, is fitted per patient and per 2D axial slice to longitudinal tumor area measurements. The fitted model extrapolates future tumor burden, which is then used as a target for a gradient-guided DDIM: a regressor trained to predict tumor size relative to brain area steers the denoising process toward images with the desired tumor size. Follow-up FLAIR slices are synthesized, and tumor growth probability maps are computed by aggregating difference maps from multiple generated images with varying target sizes (static: true target size; dynamic: ODE bootstrap estimates). The method is trained on BraTS adult/pediatric data and evaluated on 60 axial slices from 8 in-house pediatric DMG patients. The main reported results are region-specific SSIM for image quality and a significant reduction in HD95 between generated thresholded probability maps and the true target segmentation compared with initial-to-target HD95, which the authors interpret as evidence that the framework captures both extent and directionality of tumor growth.

Significance. If the claims hold, the paper would make a useful contribution: it combines a biologically motivated ODE model with a modern generative model in a data-sparse clinical setting, and it reports an in-house longitudinal pediatric DMG evaluation rather than relying only on public datasets. The public code availability and the explicit use of bootstrap uncertainty quantification are strengths. The mechanistic module is evaluated separately from the generative module, which is good practice. However, the central claim about growth 'directionality' is currently supported only by a metric that is confounded by tumor-size increase, and the ODE model has an internal mathematical inconsistency. Because these issues are load-bearing for the main conclusions, the paper requires substantive revision before the claims can be accepted.

major comments (3)

[Section 2.2, Eq. (6) and Algorithm 1 line 7] The delayed decay rate is defined inconsistently. Eq. (6) states lambda'(t) = lambda_decay * tanh((t - t_RTstart - delta) * slope), while Algorithm 1 line 7 sets lambda_eff = -lambda * tanh((t - t_RTstart - delay)/slope). These differ in sign, in the magnitude parameter (lambda_decay vs lambda), in the delay symbol, and in the slope scaling. More importantly, both Eqs. (6) and Algorithm 1 write the dying compartment as A_d(t) = A_d(t_RTstart) * exp(lambda_eff(t) * (t - t_RTstart)) with a time-dependent rate. This is not the solution of dA_d/dt = lambda_eff(t) * A_d; the correct exponential factor is exp(integral lambda_eff(s) ds). Since the ODE predictions are the conditioning targets for image generation, this inconsistency and integration error must be corrected and the resulting fits re-evaluated.
[Section 3.3, Fig. 4B (HD95 evaluation)] The claim that the framework captures growth directionality is confounded by tumor-size increase. The HD95 of the generated (thresholded probability) mask against the true target segmentation is compared with the HD95 of the initial (smaller) segmentation against the target. Because the guidance signal is a scalar regressor of tumor size relative to brain area, the generated image is explicitly pushed toward larger tumor area; any mask that enlarges the tumor and overlaps the target will reduce HD95 relative to the smaller initial mask, even if the direction of growth is wrong. A size-matched baseline is required: for example, isotropically dilating the initial tumor mask to the target area, or using a non-spatial growth baseline, and comparing HD95 against that baseline. Without such a control, the reported HD95 improvement does not demonstrate correct directionality.
[Section 3.3 (static vs. dynamic evaluation)] The static probability-map evaluation conditions on the true tumor size at the next imaging session and then compares the generated mask with the true segmentation at that same session. This is a valid oracle test of the generative model's ability to produce a target-sized tumor, but it removes the mechanistic ODE from the spatial prediction path. The dynamic evaluation uses ODE-estimated sizes, but it still lacks a size-matched spatial control. Consequently, the specific contribution of the mechanistic model to spatial accuracy is not identified. Please report the results separately for the static and dynamic settings with the control baseline, and state explicitly which claim each comparison is intended to support.

minor comments (6)

[Throughout] Typographical errors: 'cnetral' should be 'central' (Section 3.2), 'Institue' should be 'Institute' (affiliations), and 'bootstrap' capitalization is inconsistent.
[Eq. (2)] The notation epsilon_theta^{(l)} is not defined clearly; it should distinguish the network's noise prediction at diffusion step l from the noise terms in Eq. (1).
[Section 2.3] The dynamic guidance scaling s_R(dyn) is described verbally but no update rule is given. Please specify the formula used to adapt the gradient scale during sampling.
[Figure 4B] The y-axis label 'HD95 Distance' lacks units and the figure does not report the number of paired samples, exact p-values, or effect sizes. Please include these details in the text or caption.
[Section 3.3] The criterion for using 'tumor sizes below 90th percentile of the bootstrap-estimated values' as targets is not justified. Please clarify how the target sizes were sampled and why the upper tail was excluded.
[Figure 3B and Section 3.2] The labels 'all' and 'train' are used without a definition in the figure caption. Define them in the caption or in the main text for clarity.

Circularity Check

1 steps flagged · score 4.0 of 10

Static probability-map arm conditions on the true follow-up size and is then scored against that same follow-up, so the HD95 'directionality' evidence is partly built from the target value; the dynamic ODE arm remains the only independent prediction.

other [Section 3.3, Implementation Details / Evaluation]
"As a comparison, the framework was applied with the target set to the true tumor size. ... We thresholded both the generated static (true tumor size) and dynamic (ODE-estimated tumor size) probability maps to generate binary masks. These were compared with the true target segmentation using the 95th percentile of the Hausdorff distance (HD95)."

Static probability maps are produced by setting the conditioning target in Eq. (2) to the actual follow-up tumor area ('the framework was applied with the target set to the true tumor size'), and then the HD95 evaluation is run against that same follow-up segmentation ('These were compared with the true target segmentation'). Because R in Eq. (2) is a scalar size regressor, the guidance field ∇_x R carries no explicit spatial-direction information; the only future information injected is the true area. Consequently, any generated mask that merely enlarges the tumor and overlaps the larger target will lower HD95 relative to the initial-scan mask, regardless of whether the growth direction is clinically correct. The reported HD95 drop is therefore built into the static evaluation by construc

full rationale

Most of the derivation chain is self-contained: the DDPM/DDIM and regressor-guidance components are standard external methods [2,20,23]; the ODE model (Eqs. 3-6) is a conventional exponential-growth/compartmental-decay model fitted only to measurements before the follow-up time; and the dynamic probability maps are driven by bootstrap ODE extrapolations rather than by the true future size, so they constitute a genuine temporal extrapolation. No load-bearing argument reduces to a self-citation chain or a uniqueness theorem imported from the authors' prior work. The main circularity risk is the static probability-map arm: the generator is explicitly given the true tumor size at the target session, and the identical follow-up segmentation is then used as the HD95 reference. Since the guidance signal in Eq. (2) is the gradient of a scalar size regressor, the spatial direction is not directly encoded, so a size increase alone can produce the reported HD95 improvement. The paper labels this arm 'as a comparison', so the central mechanistic-learning claim retains independent content, but the conclusion that HD95 'supports ... correct direction' leans on this oracle-conditioned arm. A size-matched isotropic-expansion baseline would be needed to separate size effect from directionality; its absence is a validity concern rather than an additional circular step.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The central claim rests on per-slice ODE parameters fitted to sparse longitudinal data (A0, lambda, S, delta, slope, and a decay rate whose definition differs between Eq. 6 and Algorithm 1), plus grid-searched inference hyperparameters (NL=200, sR=50k). The modeling axioms are standard exponential-growth assumptions and ad hoc choices for the shape of radiation-induced death. No new physical entities are introduced; the tumor growth probability maps are postprocessing aggregations, not invented entities.

free parameters (8)

A0 = per-slice, not reported
Initial axial tumor area in Eq. (3); estimated by least squares for each slice.
lambda = per-slice, not reported
Net exponential growth rate in Eqs. (3) and (5); controls pre-RT and surviving-compartment growth.
S = per-slice, not reported
Surviving fraction at RT onset in Eq. (4); partitions tumor into living and dying compartments.
delta = per-slice, not reported
Delay parameter in the tanh transition of Eq. (6) / Algorithm 1; estimated from data.
slope = per-slice, not reported
Steepness parameter of the tanh transition; estimated from data.
lambda_decay = per-slice, not reported
Decay rate of the dying compartment in Eq. (6), while Algorithm 1 instead uses -lambda; both forms are fit-dependent.
noise_level_NL = 200
Number of forward diffusion steps used at inference; grid-searched on 185 paired DMG images with the true next tumor size as target.
regressor_scale_sR(ct) = 50,000
Constant guidance scaling in Eq. (2); grid-searched on the same validation pairs.

assumptions (5)

domain assumption Untreated tumor growth follows exponential kinetics before and after RT for the surviving population
Eqs. (3) and (5) assume no growth saturation during the prediction window, a standard but untested assumption for these DMG slices.
ad hoc to paper Radiation-induced delayed cell death follows a tanh transition with fitted delay and slope parameters
Section 2.2 states this functional form was chosen to reflect gradual engagement of death pathways; it is not derived from data or a cited mechanistic law.
domain assumption Each 2D axial slice is an independent dynamical system
Section 3.2 treats all 60 slices independently, ignoring through-plane coupling and shared patient effects.
domain assumption A regressor trained to predict tumor size from static noisy images provides spatially useful guidance for growth direction
Eq. (2) and Section 2.3 use only the tumor-size regressor gradient to steer generation; the regressor is never trained on longitudinal growth direction.
ad hoc to paper The bootstrap noise model of 10% multiplicative Gaussian observation error captures measurement uncertainty
Algorithm 1 applies N(0, sigma^2) perturbation with 10% scale without comparing to an actual measurement-error distribution.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Mechanistic Learning with Guided Diffusion Models to Predict Spatio-Temporal Brain Tumor Growth." pith.science (2026). https://pith.science/paper/DKZ2IKYY

@misc{pith2026250909610,
  author       = {Pith},
  title        = {Pith review of: Mechanistic Learning with Guided Diffusion Models to Predict Spatio-Temporal Brain Tumor Growth},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DKZ2IKYY}},
  note         = {Machine review of arXiv:2509.09610}
}

read the original abstract

Predicting the spatio-temporal progression of brain tumors is essential for guiding clinical decisions in neuro-oncology. We propose a hybrid mechanistic learning framework that combines a mathematical tumor growth model with a guided denoising diffusion implicit model (DDIM) to synthesize anatomically feasible future MRIs from preceding scans. The mechanistic model, formulated as a system of ordinary differential equations, captures temporal tumor dynamics including radiotherapy effects and estimates future tumor burden. These estimates condition a gradient-guided DDIM, enabling image synthesis that aligns with both predicted growth and patient anatomy. We train our model on the BraTS adult and pediatric glioma datasets and evaluate on 60 axial slices of in-house longitudinal pediatric diffuse midline glioma (DMG) cases. Our framework generates realistic follow-up scans based on spatial similarity metrics. It also introduces tumor growth probability maps, which capture both clinically relevant extent and directionality of tumor growth as shown by 95th percentile Hausdorff Distance. The method enables biologically informed image generation in data-limited scenarios, offering generative-space-time predictions that account for mechanistic priors.

Figures

Figures reproduced from arXiv: 2509.09610 by the authors.

**Figure 1.** Overview of the proposed method. 3 Experimental Results 3.1 Optimization of Guided Denoising Diffusion Model Implementation Details. For training our guided diffusion framework, we used multiparametric MRI scans from 1125 adult and 105 pediatric high-grade glioma patients from BraTS 2023 Challenge[13,5]. All 2D axial slices containing brain tissue were included (n∼140,000). The DDPM was trained with 1,000 diffusion… view at source ↗

**Figure 2.** Result for the grid search optimization of noise level (NL) and regressor [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** A.Growth curve fitting (red) and estimation(gray). B. Mechanistic mod [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: A. Generating tumor growth probability maps based on boostrap [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 2 canonical work pages

[1]

Brüningk, S.C., Peacock, J., Whelan, C.J., Brady-Nicholls, R., Yu, H.H.M., Sa- hebjam, S., Enderling, H.: Intermittent radiotherapy as alternative treatment for recurrent high grade glioma: a modeling study based on longitudinal tumor measurements. Sci. Rep.11(1), 20219 (12 Oct 2021).https://doi.org/10.1038/ s41598-021-99507-2

2021
[2]

Advances in neural information processing systems34, 8780–8794 (2021)

Dhariwal, P., Nichol, A.: Diffusion models beat gans on image synthesis. Advances in neural information processing systems34, 8780–8794 (2021)

2021
[3]

Lancet Oncol.21(6), e317–e329 (Jun 2020).https://doi.org/10.1016/ S1470-2045(20)30173-X

Erker,C.,Tamrazi,B.,Poussaint,T.Y.,Mueller,S.,Mata-Mbemba,D.,Franceschi, E., Brandes, A.A., Rao, A., Haworth, K.B., Wen, P.Y., Goldman, S., Vezina, G., MacDonald, T.J., Dunkel, I.J., Morgan, P.S., Jaspan, T., Prados, M.D., War- ren, K.E.: Response assessment in paediatric high-grade glioma: recommenda- tions from the response assessment in pediatric neuro...

2020
[4]

arXiv [q-bio.QM] (1 Jul 2019)

Ezhov, I., Lipkova, J., Shit, S., Kofler, F., Collomb, N., Lemasson, B., Barbier, E., Menze, B.: Neural parameters estimation for brain tumor growth modeling. arXiv [q-bio.QM] (1 Jul 2019)

2019
[5]

Fathi Kazerooni, A., Khalili, N., Liu, X., Haldar, D., Jiang, Z., Zapaishchykova, A., et al.: BraTS-PEDs: Results of the multi-consortium international pediatric braintumorsegmentationchallenge2023.J.Mach.Learn.Biomed.Imaging3(June 2025), 72–87 (26 Jun 2025).https://doi.org/10.59275/j.melba.2025-f6fg

doi:10.59275/j.melba.2025-f6fg 2025
[6]

Hazout, S., Oehler, C., Zwahlen, D.R., Taussky, D.: Historical view of the effects of radiation on cancer cells. Oncol. Rev.19, 1527742 (30 Apr 2025).https://doi. org/10.3389/or.2025.1527742

arXiv 2025
[7]

Advances in neural information processing systems33, 6840–6851 (2020)

Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020)

2020
[8]

arXiv [cs.LG] (25 Jul 2022)

Ho, J., Salimans, T.: Classifier-free diffusion guidance. arXiv [cs.LG] (25 Jul 2022)

2022

Show all 25 references

[9]

NPJ Precis

Khalighi, S., Reddy, K., Midya, A., Pandav, K.B., Madabhushi, A., Abedaltha- gafi, M.: Artificial intelligence in neuro-oncology: advances and challenges in brain tumor diagnosis, prognosis, and precision treatment. NPJ Precis. Oncol.8(1), 80 (29 Mar 2024).https://doi.org/10.1...

2024 doi
[10]

arXiv [eess.IV] (7 Feb 2024)

Konz, N., Chen, Y., Dong, H., Mazurowski, M.A.: Anatomically-controllable med- ical image generation with segmentation-guided diffusion models. arXiv [eess.IV] (7 Feb 2024)

2024
[11]

IEEE Trans

Lipkova, J., Angelikopoulos, P., Wu, S., Alberts, E., Wiestler, B., Diehl, C., Preibisch, C., Pyka, T., Combs, S.E., Hadjidoukas, P., Van Leemput, K., Koumout- sakos, P., Lowengrub, J., Menze, B.: Personalized radiotherapy design for glioblas- toma: Integrating mathematical tu...

2019
[12]

arXiv [eess.IV] (18 Jun 2024)

Litrico, M., Guarnera, F., Giuffirda, V., Ravì, D., Battiato, S.: TADM: Temporally- aware diffusion model for neurodegenerative progression on brain MRI. arXiv [eess.IV] (18 Jun 2024)

2024
[13]

IEEE Trans

Menze, B.H., Jakab, A., Bauer, S., Kalpathy-Cramer, J., Farahani, K., Kirby, J., et al.: The multimodal brain tumor image segmentation benchmark (BRATS). IEEE Trans. Med. Imaging34(10), 1993–2024 (Oct 2015).https://doi.org/10. 1109/TMI.2014.2377694

1993
[14]

Metzcar, J., Jutzeler, C.R., Macklin, P., Köhn-Luque, A., Brüningk, S.C.: A review of mechanistic learning in mathematical oncology. Front. Immunol.15, 1363144 (12 Mar 2024).https://doi.org/10.3389/fimmu.2024.1363144

2024
[15]

Nichol, A., Dhariwal, P.: Improved denoising diffusion probabilistic models (18 Feb 2021)

2021
[16]

Brainlesion11993, 380–394 (19 May 2020).https://doi.org/10.1007/978-3-030-46643-5_38

Pati, S., Singh, A., Rathore, S., Gastounioti, A., Bergman, M., Ngo, P., Ha, S.M., Bounias, D., Minock, J., Murphy, G., Li, H., Bhattarai, A., Wolf, A., Sridaran, P., Kalarot, R., Akbari, H., Sotiras, A., Thakur, S.P., Verma, R., Shinohara, R.T., Yushkevich, P., Fan, Y., Konto...

2020 doi
[17]

Peng, W., Adeli, E., Zhao, Q., Pohl, K.M.: Generating realistic 3D brain MRIs using a conditional diffusion probabilistic model (15 Dec 2022)

2022
[18]

arXiv [eess.IV] (15 Sep 2022)

Pinaya, W.H.L., Tudosiu, P.D., Dafflon, J., da Costa, P.F., Fernandez, V., Nachev, P., Ourselin, S., Cardoso, M.J.: Brain imaging generation with latent diffusion models. arXiv [eess.IV] (15 Sep 2022)

2022
[19]

In: Medical Image Computing and Computer-Assisted In- tervention – MICCAI 2015

Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomedi- cal image segmentation. In: Medical Image Computing and Computer-Assisted In- tervention – MICCAI 2015. pp. 234–241. Springer International Publishing (2015). https://doi.org/10.1007/978-3-319-2...

2015 doi
[20]

Song, J., Meng, C., Ermon, S.: Denoising diffusion implicit models (6 Oct 2020)

2020
[21]

Nature Methods17, 261–272 (2020).https://doi.org/10.1038/ s41592-019-0686-2

Virtanen, P., Gommers, R., Oliphant, T.E., Haberland, M., Reddy, T., Courna- peau, D., Burovski, E., Peterson, P., Weckesser, W., Bright, J., van der Walt, S.J., Brett, M., Wilson, J., Millman, K.J., Mayorov, N., Nelson, A.R.J., Jones, E., Kern, R., Larson, E., Carey, C.J., Po...

2020
[22]

Wen, P.Y., Macdonald, D.R., Reardon, D.A., Cloughesy, T.F., Sorensen, A.G., Galanis, E., DeGroot, J., Wick, W., Gilbert, M.R., Lassman, A.B., Tsien, C., Mikkelsen, T., Wong, E.T., Chamberlain, M.C., Stupp, R., Lamborn, K.R., Vo- gelbaum, M.A., van den Bent, M.J., Chang, S.M.: ...

1963
[23]

Wolleb, J., Sandkühler, R., Bieder, F., Cattin, P.C.: The swiss army knife for image-to-image translation: Multi-task diffusion models (6 Apr 2022)

2022
[24]

Yankeelov, T.E., Atuegwu, N., Hormuth, D., Weis, J.A., Barnes, S.L., Miga, M.I., Rericha, E.C., Quaranta, V.: Clinically relevant modeling of tumor growth and treatment response. Sci. Transl. Med.5(187), 187ps9 (29 May 2013).https:// doi.org/10.1126/scitranslmed.3005686

2013 doi
[25]

Zheng, D., Preuss, K., Milano, M.T., He, X., Gou, L., Shi, Y., Marples, B., Wan, R., Yu, H., Du, H., Zhang, C.: Mathematical modeling in radiotherapy for cancer: a comprehensive narrative review. Radiat. Oncol.20(1), 49 (4 Apr 2025).https: //doi.org/10.1186/s13014-025-02626-7

2025 doi

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

Brüningk, S.C., Peacock, J., Whelan, C.J., Brady-Nicholls, R., Yu, H.H.M., Sa- hebjam, S., Enderling, H.: Intermittent radiotherapy as alternative treatment for recurrent high grade glioma: a modeling study based on longitudinal tumor measurements. Sci. Rep.11(1), 20219 (12 Oct 2021).https://doi.org/10.1038/ s41598-021-99507-2

2021

[2] [2]

Advances in neural information processing systems34, 8780–8794 (2021)

Dhariwal, P., Nichol, A.: Diffusion models beat gans on image synthesis. Advances in neural information processing systems34, 8780–8794 (2021)

2021

[3] [3]

Lancet Oncol.21(6), e317–e329 (Jun 2020).https://doi.org/10.1016/ S1470-2045(20)30173-X

Erker,C.,Tamrazi,B.,Poussaint,T.Y.,Mueller,S.,Mata-Mbemba,D.,Franceschi, E., Brandes, A.A., Rao, A., Haworth, K.B., Wen, P.Y., Goldman, S., Vezina, G., MacDonald, T.J., Dunkel, I.J., Morgan, P.S., Jaspan, T., Prados, M.D., War- ren, K.E.: Response assessment in paediatric high-grade glioma: recommenda- tions from the response assessment in pediatric neuro...

2020

[4] [4]

arXiv [q-bio.QM] (1 Jul 2019)

Ezhov, I., Lipkova, J., Shit, S., Kofler, F., Collomb, N., Lemasson, B., Barbier, E., Menze, B.: Neural parameters estimation for brain tumor growth modeling. arXiv [q-bio.QM] (1 Jul 2019)

2019

[5] [5]

Fathi Kazerooni, A., Khalili, N., Liu, X., Haldar, D., Jiang, Z., Zapaishchykova, A., et al.: BraTS-PEDs: Results of the multi-consortium international pediatric braintumorsegmentationchallenge2023.J.Mach.Learn.Biomed.Imaging3(June 2025), 72–87 (26 Jun 2025).https://doi.org/10.59275/j.melba.2025-f6fg

doi:10.59275/j.melba.2025-f6fg 2025

[6] [6]

Hazout, S., Oehler, C., Zwahlen, D.R., Taussky, D.: Historical view of the effects of radiation on cancer cells. Oncol. Rev.19, 1527742 (30 Apr 2025).https://doi. org/10.3389/or.2025.1527742

arXiv 2025

[7] [7]

Advances in neural information processing systems33, 6840–6851 (2020)

Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems33, 6840–6851 (2020)

2020

[8] [8]

arXiv [cs.LG] (25 Jul 2022)

Ho, J., Salimans, T.: Classifier-free diffusion guidance. arXiv [cs.LG] (25 Jul 2022)

2022