Pith. sign in

REVIEW 4 major objections 5 minor 35 references

Physics-Guided Radiotherapy Treatment Planning with Deep Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read A dose-supervised second training stage brings deep-learning VMAT plans to a 90.5 percent high-dose gamma pass rate.

desk verdict A worthwhile two-stage training idea with strong 3D U-Net numbers, but the paper never says what dose engine produced the reported evaluation doses—if it is the same frozen learned dose predictor used in training, the headline results are not yet evidence of true dosimetric improvement. read the letter →

arxiv 2506.19880 v1 pith:T7U6FY3N submitted 2025-06-23 physics.med-ph cs.AI

classification physics.med-phcs.AI
keywords radiotherapytreatmentplanningvolumetricmodulatedarctherapyphysics-guideddeeplearningMLCapertureprediction3Ddoseprostatecanceradaptivegammapassrate
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

Radiotherapy planning currently requires solving a high-dimensional, non-convex optimization problem, and adaptive radiotherapy makes the time pressure worse. This paper tries to show that a two-stage deep-learning pipeline can produce clinically usable VMAT plans in under a second. Stage one trains a network to predict treatment plan parameters, the multileaf collimator apertures and monitor units, directly from the CT and structure masks; stage two adds dose-domain supervision by sending the predicted plan through a frozen, differentiable dose predictor and penalizing the resulting 3D dose against the clinical dose. The authors claim, based on 133 prostate patients treated with a uniform 62 Gy two-arc protocol, that this physics-guided second stage substantially improves dosimetric agreement, with the best model reaching a PTV $D_{95\%}$ error of $0.42 \pm 1.83$ Gy and a high-dose gamma pass rate of $90.5 \pm 7.3$ percent while keeping organ-at-risk doses lower. If true, the method gives adaptive radiotherapy an automatic replanning step scored on dose rather than on one arbitrary parameter solution.

What carries the argument

The load-bearing object is the RT Dose Predictor: a fully differentiable convolutional gated recurrent unit, pretrained on 350 patients with plans generated by one commercial planning system, that maps a CT and treatment plan to a 3D dose distribution. During stage two it is frozen and cascaded after the Deep RT Planner, so the $L_2$ distance between predicted and ground-truth dose backpropagates through the dose predictor into the MLC and MU decoders. This is what makes the training physics-guided: the planner is optimized for the dose its parameters imply rather than for matching a single MLC/MU ground truth, and the gradient flows through a model of dose transport instead of through the plan parameters alone.

What would settle it

Compute the frozen dose predictor's 2%/2 mm gamma pass rate on the 13 test patients by comparing its predicted dose with the recalculated clinical ground-truth dose; if the pass rate is far below the 99.6 percent cited from its original validation, the reported stage-2 gains could be an artifact of fitting the surrogate. A complementary experiment is to train stage 2 with a classical differentiable dose engine in place of the learned predictor and check whether the PTV $D_{95\%}$ error and high-dose gamma pass-rate gains persist.

Watch

Extended reading notes

Core claim

Direct supervision on MLC apertures and monitor units is hampered by non-uniqueness: many different parameter sets deliver clinically equivalent dose. The central discovery is that adding a second-stage dose-supervision term, computed by pushing the predicted plan through a pretrained, frozen RT Dose Predictor and comparing its 3D dose to the ground-truth dose with an L2 loss, systematically brings predictions closer to clinical plans. For the 3D U-Net, the PTV $D_{95\%}$ error falls from $1.75 \pm 3.55$ Gy to $0.42 \pm 1.83$ Gy, and the gamma pass rate above 90 percent of maximum dose rises from $56.46 \pm 32.68$ percent to $90.50 \pm 7.28$ percent; OAR mean and maximum doses shift lower or stay comparable. The same dose-guided stage also improves the UNETR variant, though less strongly, which supports the interpretation that the mechanism, not a single architecture, drives the gain. The paper concludes that training in the clinically relevant dose domain mitigates multi-arc redundancy and makes deep-learning plans feasible for adaptive radiotherapy.

Load-bearing premise

The load-bearing premise is that the frozen dose calculator, trained on plans from other patients and a different planning system, is accurate enough on the current cohort's recalculated doses, so the second-stage loss steers the planner toward true dose instead of toward the calculator's own errors.

Editorial extensions

If this is right

  • A single forward pass of the two-stage pipeline produces a full 144-control-point VMAT plan in under one second per patient, contrasted with minutes for conventional GPU-based planning systems.
  • Dose-domain supervision improves PTV and CTV coverage metrics such as $D_{95\%}$, $D_{98\%}$, and $V_{95\%}$ and high-dose gamma pass rates for both a convolutional and a transformer-based planner, so the benefit does not depend on the specific network.
  • After the physics-guided stage, rectum and femoral-head doses are lower than or clinically comparable to the clinical ground truth, suggesting the method could reduce organ-at-risk toxicity in replanning.
  • The largest gamma improvement occurs in the highest-dose region, where the 3D U-Net's pass rate roughly doubles, which is the region most relevant for target coverage.

Reading between the lines

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

  • The gains depend on the frozen dose predictor transferring to the target planning system; an unstated check would be to re-evaluate that predictor's 2%/2 mm gamma pass rate directly on the recalculated doses of the 13 test patients and to exclude patient overlap between its 350 training cases and this cohort.
  • A clean way to locate the source of the physics guidance is to replace the learned dose predictor with a classical differentiable dose calculator, for example a collapsed-cone or Monte Carlo-based engine, in the same two-stage loop; if the dosimetric gains vanish, they come from the learned surrogate rather than from dose-level supervision itself.
  • Because the non-uniqueness of plan parameters is generic, dose-level supervision should transfer across planning systems better than parameter-level supervision; a natural extension is to train the planner on plans recalculated with several different systems and test on a held-out system.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a two-stage, physics-guided deep learning pipeline for VMAT prostate radiotherapy planning. In the first stage, a 3D U-Net or UNETR is trained with direct supervision on MLC aperture masks and monitor units from clinical plans. In the second stage, a frozen, differentiable neural dose predictor (the RT Dose Predictor, after Witte and Sonke 2024) maps the predicted plan to a 3D dose distribution, and an additional MSE loss between predicted and ground-truth dose is backpropagated through the cascade. The authors evaluate on 133 prostate cancer patients (104 training, 16 validation, 13 test) treated with a uniform 2-arc VMAT protocol delivering 62 Gy, with all ground-truth plans recalculated in Pinnacle with standardized parameters. They report that the second stage improves PTV/CTV DVH metrics and gamma pass rates for both architectures, with the 3D U-Net achieving PTV D95% error of 0.42 ± 1.83 Gy and a high-dose (90% threshold) gamma pass rate of 90.5 ± 7.3%.

Significance. If the reported dosimetric gains are confirmed with an independent dose calculation engine, the paper would make a useful contribution to fast adaptive radiotherapy planning: it demonstrates an end-to-end differentiable training scheme that combines direct parameter supervision with a dose-domain objective, and it evaluates two competitive architectures on a clinically standardized dataset. The authors also receive credit for using a uniform treatment protocol, clinical review of the recalculated plans, and a reasonably sized test cohort. However, the central claim rests on the evaluation dose distributions being computed by an independent engine; the manuscript does not state what engine produced the DVH and gamma results for the predicted plans. Because the stage-2 loss is minimized in the space of the frozen learned dose predictor, evaluation in that same space would be circular and would not establish true agreement with Pinnacle dose. The significance of the work is therefore conditional on resolving this issue.

major comments (4)
  1. [Section 3.1, Tables 1 and 2] The manuscript never specifies which dose calculation engine produced the dose distributions used to compute the DVH differences in Table 1 and the gamma pass rates in Table 2. Section 3.1 states only that evaluation used CUDA-accelerated preprocessing and model inference on an RTX A6000, which is consistent with running the frozen RT Dose Predictor on the predicted MLC/MU, not with a Pinnacle recalculation. If the evaluation doses come from the same learned RT Dose Predictor used in the stage-2 loss, the comparison is circular: the planner is trained to minimize error in the surrogate's dose space and then judged in that same space. The authors must state explicitly which dose engine was used for the evaluation of predicted plans and, if it was not Pinnacle or another independent engine, must recompute the reported metrics with an independent dose calculation.
  2. [Section 2.2, RT Dose Predictor] The RT Dose Predictor is the load-bearing component of the physics-guided stage, yet its accuracy is only asserted via a 99.6% gamma pass rate (2%/2 mm) from prior validation on 350 Monaco-generated plans. This paper evaluates on 133 Pinnacle-recalculated plans, a different treatment planning system, and does not revalidate the dose predictor on this cohort or report whether any of the 350 training patients overlap with the current 133 patients. A biased surrogate trained on Monaco plans could systematically disagree with Pinnacle dose, and the stage-2 loss would then steer the planner toward the surrogate's errors rather than toward true dose. The authors should report the dose predictor's gamma pass rate and DVH accuracy on the current Pinnacle cohort, including the 13 test patients, and clarify that no overlap exists with the dose predictor's training set.
  3. [Section 3.1, paragraph 2] The paper states that the second training stage 'significantly improved' key dosimetric metrics, but no statistical significance testing is reported. Many OAR metrics in Table 1 have standard deviations that exceed the mean differences (e.g., rectum Dmean for 3D U-Net: -0.82 ± 4.45 Gy; left femoral Dmean for UNETR: -0.72 ± 6.34 Gy), so the claimed OAR sparing is not established without paired tests or confidence intervals. The authors should provide paired statistical tests (e.g., Wilcoxon signed-rank or paired t-tests) across the 13 test patients for the key PTV/CTV and OAR metrics, with appropriate multiple-comparison awareness.
  4. [Section 3.1, paragraph 2 and Table 1] There is a numeric inconsistency: the text reports that the 3D U-Net achieved a mean absolute difference of D95% = 0.59 ± 2.23 Gy, D98% = 0.70 ± 2.14 Gy, and V95% = -0.42 ± 1.12%, but Table 1's PTV rows for the 3D U-Net second stage show D95% = 0.42 ± 1.83, D98% = -0.71 ± 2.12, and V95% = -0.22 ± 1.87. The values in the text match the CTV rows of Table 1, not the PTV rows. The authors should clarify which ROI the text refers to and correct the mismatch, since the abstract quotes PTV values and the reader should be able to trace them.
minor comments (5)
  1. [Abstract] The abstract says the approach 'consistently produces treatment plans that closely match clinical ground truths,' but Table 1 shows UNETR's second-stage PTV D95% is -1.70 ± 2.21 Gy. Consider adding 'on average' or reporting the range, because 'consistently' overstates the per-patient behavior suggested by the standard deviations.
  2. [Section 2.1] There are two unit errors: '7 MeV beam energy' should be '7 MV' (megavoltage, not megaelectronvolt), and 'an isotropic resolution of 3.5 mm3' should be '3.5 mm' (voxel spacing, not volume).
  3. [Section 2.2, first paragraph] The description of the five input channels is unclear: the first channel is the CT, and 'the remaining four channels contain the rotation and projection at each control point for the CT, PTV, CTV, and OARs.' Please clarify whether the four channels are rotated versions of the CT and three masks for each control point, and how the control point index is provided to the network.
  4. [Section 3.2] The gamma evaluation uses 3%/3 mm criteria but does not state the software or implementation used to compute the gamma index, nor whether the comparison is performed in the 3D dose grid after resampling. Adding this detail would improve reproducibility.
  5. [Discussion] The Discussion does not mention the reliance on a learned dose predictor as a limitation. Given that the stage-2 objective uses a surrogate dose engine, a sentence acknowledging that surrogate errors could propagate and that independent dose recalculation is needed for clinical validation would strengthen the paper.

Circularity Check

1 steps flagged · score 6.0 of 10

Dosimetric gains likely reflect the same frozen dose surrogate used for stage-2 supervision; evaluation never shows independent Pinnacle recalculation of predicted plans.

  1. fitted input called prediction [Section 2.3 (second-stage loss) and Section 3.1/3.2 (evaluation)]
    "The second-stage loss function is defined as: L = L_BCE(M_pred, M_true) + λ1·∥MU_pred−MU_true∥1 + λ2·∥D_pred−D_true∥2^2 ... To ensure efficient evaluation, we used CUDA-accelerated preprocessing [7], including rotation and projection operations, and performed model inference on a single NVIDIA RTX A6000 GPU in under one second per patient."

    D_pred comes from the frozen RT Dose Predictor, the paper's only dose engine. Evaluation is described only as 'model inference ... in under one second per patient', with no independent Pinnacle recalculation of predicted plans; therefore Tables 1-2 compare D_pred from the same surrogate to D_true. Stage 2 already minimizes ∥D_pred−D_true∥2^2, so the gamma/DVH improvements re-measure the training objective in the surrogate's output space. The 99.6% gamma from [33] was obtained on a different Monaco-based dataset and does not validate the surrogate on this Pinnacle-recalculated cohort. A biased surrogate lets the planner reduce surrogate error without improving true dose.

full rationale

The central training contribution is a differentiable dose-based loss using a pretrained neural dose predictor. If the predictor were only used for training and the final plans were independently recalculated in Pinnacle, the comparison would be a valid external benchmark. However, the evaluation text describes CUDA-accelerated preprocessing and model inference in under one second per patient, with no statement that predicted plans were recalculated with Pinnacle, while the dataset section explicitly says ground-truth plans were Pinnacle-recalculated. Since the only dose predictor in the method is the frozen RT Dose Predictor, the reported dosimetric metrics appear to compare surrogate dose to ground-truth Pinnacle dose. This makes the improvement partly by construction: the planner is optimized to minimize exactly the surrogate-vs-truth discrepancy that the gamma/DVH tables then report. The self-citation [33] is an independently validated journal result (Monte Carlo training), so it is not circular by itself; the circularity lies in using the same learned function as both training target and evaluation engine without an independent dose recalculation. If the authors confirm separate Pinnacle recalculations for evaluation, the score would drop to 0-2.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim is empirical and rests on the accuracy of a pre-trained learned dose surrogate, the validity of clinical ground truth plans, and an unrestricted plan representation. No new physical entities are introduced; the 'physics guidance' is a learned dose model from prior work rather than a first-principles dose calculation.

free parameters (2)
  • lambda_1 (MU loss weight) = 100
    Hand-chosen balance between BCE MLC loss and L1 MU loss in both stages (Section 2.3); no sensitivity analysis.
  • lambda_2 (dose loss weight) = 10
    Hand-chosen weight for the MSE dose term in stage 2 (Section 2.3); no sensitivity analysis.
assumptions (5)
  • domain assumption Pinnacle-recalculated clinical plans, reviewed by an oncologist, are a valid and unambiguous ground truth for plan quality.
    The supervised losses in Section 2.3 and all DVH comparisons in Table 1 assume the clinical plan is the correct target, despite the paper's own statement that optimal plans are non-unique.
  • domain assumption The frozen RT Dose Predictor transfers from Monaco-generated training plans to Pinnacle-recalculated plans in this cohort without revalidation.
    Section 2.2 reports prior validation on 350 Monaco patients with a 99.6% gamma pass rate, but does not evaluate the predictor on the 133-patient Pinnacle dataset or analyze possible patient overlap.
  • domain assumption A plan is fully represented by 144 binary MLC aperture masks and 144 MU values, with no additional deliverability constraints.
    The input and output representation in Sections 2.1 and 2.2 ignores MLC speed, leaf collision, tongue-and-groove, and other mechanical delivery limits.
  • domain assumption Gradients backpropagated through the frozen differentiable dose predictor are useful for training the planner.
    Section 2.3 relies on end-to-end differentiability, but no analysis of gradient magnitude or bias is provided.
  • standard math Standard deep learning training assumptions (AdamW, cosine annealing, early stopping) lead to a converged solution.
    Training details in Section 2.3 are standard; no claim-specific mathematical derivation depends on them.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Physics-Guided Radiotherapy Treatment Planning with Deep Learning." pith.science (2026). https://pith.science/paper/T7U6FY3N

@misc{pith2026250619880,
  author       = {Pith},
  title        = {Pith review of: Physics-Guided Radiotherapy Treatment Planning with Deep Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T7U6FY3N}},
  note         = {Machine review of arXiv:2506.19880}
}
read the original abstract

Radiotherapy (RT) is a critical cancer treatment, with volumetric modulated arc therapy (VMAT) being a commonly used technique that enhances dose conformity by dynamically adjusting multileaf collimator (MLC) positions and monitor units (MU) throughout gantry rotation. Adaptive radiotherapy requires frequent modifications to treatment plans to account for anatomical variations, necessitating time-efficient solutions. Deep learning offers a promising solution to automate this process. To this end, we propose a two-stage, physics-guided deep learning pipeline for radiotherapy planning. In the first stage, our network is trained with direct supervision on treatment plan parameters, consisting of MLC and MU values. In the second stage, we incorporate an additional supervision signal derived from the predicted 3D dose distribution, integrating physics-based guidance into the training process. We train and evaluate our approach on 133 prostate cancer patients treated with a uniform 2-arc VMAT protocol delivering a dose of 62 Gy to the planning target volume (PTV). Our results demonstrate that the proposed approach, implemented using both 3D U-Net and UNETR architectures, consistently produces treatment plans that closely match clinical ground truths. Our method achieves a mean difference of D95% = 0.42 +/- 1.83 Gy and V95% = -0.22 +/- 1.87% at the PTV while generating dose distributions that reduce radiation exposure to organs at risk. These findings highlight the potential of physics-guided deep learning in RT planning.

Figures

Figures reproduced from arXiv: 2506.19880 by the authors.

Figure 1
Figure 1. Two-stage physics-guided training framework: (a) The Deep RT Planner is trained with ground truth RT plans. (b) Dose supervision via the RT Dose Predictor. OARs, lower dose values are preferable, meaning a negative difference relative to the ground truth indicates better sparing of healthy tissue. The second training stage significantly improved key dosimetric metrics for the PTV/CTV, including D98, D95, and V95%. T… view at source ↗
Figure 2
Figure 2. DVHs comparing predicted dose distributions after the 1st (left) and 2nd (right) training stages of the 3D U-Net for the same patient. Ground truth (G.T.) doses are shown as dashed lines, while predicted (Pred.) doses are shown as solid lines. 3D U-Net: 1st Stage Patient 1 Patient 2 Ground Truth 3D U-Net: 2nd Stage [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Isodose distributions (25%, 50%, 75%, 90% of max dose) for two patients. The ground truth (left) is compared to predictions from the 1st (middle) and 2nd (right) training stages of the 3D U-Net [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 22 canonical work pages

  1. [1]

    Medical Physics22(3), 379–388 (1995)

    Ahnesjö, A.: Collapsed cone convolution of radiant energy for photon dose calcu- lation in heterogeneous media. Medical Physics22(3), 379–388 (1995)

  2. [2]

    Medical Physics 36(11), 5128–5138 (2009)

    Bedford, J.L.: Treatment planning for volumetric modulated arc therapy. Medical Physics 36(11), 5128–5138 (2009). https://doi.org/10.1118/1.3240488

  3. [3]

    Physics in Medicine & Biology51(13), R363–R379 (2006)

    Bortfeld, T.: Imrt: a review and preview. Physics in Medicine & Biology51(13), R363–R379 (2006). https://doi.org/10.1088/0031-9155/51/13/R21

  4. [4]

    Bronstein, M.M., Bruna, J., Cohen, T., Veličković, P.: Geometric deep learning: Grids, groups, graphs, geodesics, and gauges (2021), https://arxiv.org/abs/2104.13478

  5. [5]

    Leaver,M.T.:Washington&Leaver’s Principles and Practice of Radiation Therapy

    CharlesM.Washington,DennisT. Leaver,M.T.:Washington&Leaver’s Principles and Practice of Radiation Therapy. Mosby, 5th edn. (2020)

  6. [6]

    Cho, K., van Merrienboer, B., Bahdanau, D., Bengio, Y.: On the prop- erties of neural machine translation: Encoder-decoder approaches (2014), https://arxiv.org/abs/1409.1259

  7. [7]

    Elsevier (2013)

    Cook, S.: CUDA Programming: A Developer’s Guide to Parallel Computing with GPUs. Elsevier (2013)

  8. [8]

    Cancer104(6), 1129–1137 (2005)

    Delaney, G., Jacob, S., Featherstone, C., Barton, M.: The role of radiotherapy in cancer treatment: estimating optimal utilization from a review of evidence-based clinical guidelines. Cancer104(6), 1129–1137 (2005)

Show all 35 references
  1. [9]

    arXiv preprint arXiv:2010.11929 (2021), https://arxiv.org/abs/2010.11929

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., Houlsby, N.: An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929 (...

  2. [10]

    Elekta: Monaco treatment planning system

  3. [11]

    Faroughi, S.A., Pawar, N., Fernandes, C., Raissi, M., Das, S., Kalantari, N.K., Mahjour, S.K.: Physics-guided, physics-informed, and physics-encoded neural net- works in scientific computing (2023), https://arxiv.org/abs/2211.07377

  4. [12]

    MIT Press (2016)

    Goodfellow, I., Bengio, Y., Courville, A.: Deep Learning. MIT Press (2016)

  5. [13]

    arXiv preprint arXiv:2103.10504 (2022) 10 S

    Hatamizadeh, A., Yin, H., Kautz, J., Molchanov, P.: Unetr: Transformers for 3d medical image segmentation. arXiv preprint arXiv:2103.10504 (2022) 10 S. Achlatis et al

  6. [14]

    Medical Physics 50(8), 5088–5094 (Aug 2023)

    Heilemann, G., Zimmermann, L., Schotola, R., Lechner, W., Peer, M., Wid- der, J., Goldner, G., Georg, D., Kuess, P.: Generating deliverable dicom rt treatment plans for prostate vmat by predicting mlc motion sequences with an encoder-decoder network. Medical Physics 50(8), 508...

  7. [15]

    Medical Physics51(6), 3972– 3984 (2024)

    Hrinivich, W.T., Bhattacharya, M., Mekki, L., McNutt, T., Jia, X., Li, H., Song, D.Y., Lee, J.: Clinical vmat machine parameter optimization for localized prostate cancer using deep reinforcement learning. Medical Physics51(6), 3972– 3984 (2024). https://doi.org/10.1002/mp.17100

  8. [16]

    Medical Physics47(12), 6140–6150 (2020)

    Hrinivich, W.T., Lee, J.: Artificial intelligence-based radiotherapy machine param- eter optimization using reinforcement learning. Medical Physics47(12), 6140–6150 (2020). https://doi.org/10.1002/mp.14544

  9. [17]

    Physics in Medicine & Biology 67(18), 185017 (2022)

    Jhanwar, G., Dahiya, N., Ghahremani, P., Zarepisheh, M., Nadeem, S.: Domain knowledge driven 3d dose prediction using moment-based loss function. Physics in Medicine & Biology 67(18), 185017 (2022). https://doi.org/10.1088/1361- 6560/ac8d45

  10. [18]

    Cancers (Basel) 16(6) (2024)

    Lemus, O.M.D., Cao, M., Cai, B., Cummings, M., Zheng, D.: Adaptive radiotherapy: Next-generation radiotherapy. Cancers (Basel) 16(6) (2024). https://doi.org/10.3390/cancers16061206

  11. [19]

    arXiv preprint arXiv:1509.02971 (2019)

    Lillicrap, T., Hunt, J., Pritzel, A., Heess, N., Erez, T., Tassa, Y., Silver, D., Wierstra, D.: Continuous control with deep reinforcement learning. arXiv preprint arXiv:1509.02971 (2019)

  12. [20]

    Liu, F., Jansson, N., Podobas, A., Fredriksson, A., Markidis, S.: Ac- celerating radiation therapy dose calculation with nvidia gpus (2021), https://arxiv.org/abs/2103.09683

  13. [21]

    Loshchilov, I., Hutter, F.: Decoupled weight decay regularization (2019), https://arxiv.org/abs/1711.05101

  14. [22]

    Physics in Medicine & Biology 67(15) (jul 2022)

    Ni, Y., Chen, S., Hibbard, L., Voet, P.: Fast vmat planning for prostate radiotherapy: dosimetric validation of a deep learning-based initial seg- ment generation method. Physics in Medicine & Biology 67(15) (jul 2022). https://doi.org/10.1088/1361-6560/ac80e5

  15. [23]

    Medical Physics 35(1), 310–317 (2008)

    Otto, K.: Volumetric modulated arc therapy: Imrt in a single gantry arc. Medical Physics 35(1), 310–317 (2008). https://doi.org/10.1118/1.2818738

  16. [24]

    Philips Healthcare: Pinnacle treatment planning system

  17. [25]

    Medical Physics47(6), 2329–2336 (2020)

    Shen, C., Nguyen, D., Chen, L., Gonzalez, Y., McBeth, R., Qin, N., Jiang, S.B., Jia, X.: Operating a treatment planning system using a deep-reinforcement learning-based virtual treatment planner for prostate cancer intensity-modulated radiation therapy treatment planning. Medi...

  18. [26]

    MIT Press, 2nd edn

    Sutton, R.S., Barto, A.G.: Reinforcement Learning: An Introduction. MIT Press, 2nd edn. (2018)

  19. [27]

    Physics in Medicine & Biology 60(2), 623 (2015)

    Unkelbach, J., Chan, T., Bortfeld, T.: Accounting for range uncertainties in the optimization of intensity modulated proton therapy. Physics in Medicine & Biology 60(2), 623 (2015)

  20. [28]

    Physics in Medicine & Biology (2023)

    Vandewinckele, L., Reynders, T., Weltens, C., Maes, F., Crijns, W.: Deep learning based mlc aperture and monitor unit prediction as a warm start for breast vmat optimisation. Physics in Medicine & Biology (2023)

  21. [29]

    Physics in Medicine & Biology48(21), R107–R164 (2003) Physics-Guided Radiotherapy Treatment Planning with Deep Learning 11

    Verhaegen, F., Seuntjens, J.: Monte carlo modelling of external radiotherapy pho- ton beams. Physics in Medicine & Biology48(21), R107–R164 (2003) Physics-Guided Radiotherapy Treatment Planning with Deep Learning 11

  22. [30]

    de Vries, L., Van Herten, R.L.M., Hoving, J.W., Isgum, I., Emmer, B., Majoie, C.B., Marquering, H., Gavves, S.: Accelerating physics-informed neural fields for fast ct perfusion analysis in acute ischemic stroke. In: Burgos, N., Petitjean, C., Vakalopoulou, M., Christodoulidis...

  23. [31]

    Medical Physics34(5), 1647–1654 (2007)

    Wendling, M., Zijp, L.J., McDermott, L.N., Smit, E.J., Sonke, J.J., Mijnheer, B.J., van Herk, M.: A fast algorithm for gamma evaluation in 3d. Medical Physics34(5), 1647–1654 (2007). https://doi.org/10.1118/1.2721657

  24. [32]

    Lil’Log (2024), https://lilianweng.github.io/posts/2024-11-28-reward-hacking/

    Weng, L.: Reward hacking in reinforcement learning. Lil’Log (2024), https://lilianweng.github.io/posts/2024-11-28-reward-hacking/

  25. [33]

    Physics and Imaging in Ra- diation Oncology 30, 100575 (2024)

    Witte, M., Sonke, J.J.: A deep learning based dynamic arc radiotherapy photon dose engine trained on monte carlo dose distributions. Physics and Imaging in Ra- diation Oncology 30, 100575 (2024). https://doi.org/10.1016/j.phro.2024.100575

  26. [34]

    Özgün Çiçek, Abdulkadir, A., Lienkamp, S.S., Brox, T., Ronneberger, O.: 3d u-net: Learning dense volumetric segmentation from sparse annotation (2016), https://arxiv.org/abs/1606.06650

  27. [1626]

    PMLR (03–05 Jul 2024), https://proceedings.mlr.press/v250/vries24a.html

Pith tools

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