pith. sign in

arxiv: 1906.08427 · v1 · pith:5UUH3YCTnew · submitted 2019-06-20 · ⚛️ physics.med-ph

Automatic target positioning and tracking for image-guided radiotherapy without implanted fiducials

Wei Zhao , Liyue Shen , Yan Wu , Bin Han , Yong Yang , Lei Xing This is my paper

Pith reviewed 2026-05-25 19:32 UTC · model grok-4.3

classification ⚛️ physics.med-ph
keywords prostate cancermarkerless localizationdeep learningkV X-rayimage-guided radiotherapytarget trackingVMATfiducial-free
0
0 comments X

The pith

Deep learning locates the prostate in routine kV X-ray images with 1-3 mm accuracy without fiducials.

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

This paper demonstrates that a deep learning model trained on thousands of annotated kV projection images can determine prostate position in new images. The approach eliminates the need for implanted fiducials in prostate radiotherapy. If the results hold, it reduces patient risks from invasive procedures while enabling accurate daily setup and tracking. The method was validated on three patients, showing consistency with fiducial-derived positions.

Core claim

The authors show that their pre-trained deep learning model identifies the prostate location in projection kV X-ray images. Deviations from annotations were 1.66 mm to 2.77 mm anterior-posterior and 1.15 mm to 2.88 mm lateral. Positions matched those from fiducials, establishing that highly accurate markerless localization is possible.

What carries the argument

The deep learning model trained to interpret projection kV X-ray images for prostate target identification.

Load-bearing premise

The annotations accurately represent the true prostate position and the model generalizes to new patients and image qualities.

What would settle it

A study comparing the model's predictions to fiducial positions or CBCT in a larger group of unseen patients would confirm or refute the accuracy claims.

Figures

Figures reproduced from arXiv: 1906.08427 by Bin Han, Lei Xing, Liyue Shen, Wei Zhao, Yan Wu, Yong Yang.

Figure 1
Figure 1. Figure 1: Overall flowchart of the proposed deep learning-based treatment target localization method. Abbreviation: DRR [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Schematic illustration of the deep learning model for treatment target detection. RPN = region proposed [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The prostate bounding boxes derived from the deep learning model (yellow dashed box) and their corresponding [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Predicted and actual positions of the prostate target overlaid on top of the anterior-posterior simulated kV [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
read the original abstract

Current image-guided prostate radiotherapy often relies on the use of implanted fiducials or transducers for target localization. Fiducial or transducer insertion requires an invasive procedure that adds cost and risks for bleeding, infection, and discomfort to some patients. We are developing a novel markerless prostate localization strategy using a pre-trained deep learning model to interpret routine projection kV X-ray images without the need for daily cone-beam computed tomography (CBCT). A deep learning model was first trained by using several thousand annotated projection X-ray images. The trained model is capable of identifying the location of the prostate target for a given input X-ray projection image. To assess the accuracy of the approach, three patients with prostate cancer received volumetric modulated arc therapy (VMAT) were retrospectively studied. The results obtained by using the deep learning model and the actual position of the prostate were compared quantitatively. The deviations between the target positions obtained by the deep learning model and the corresponding annotations ranged from 1.66 mm to 2.77 mm for anterior-posterior (AP) direction, and from 1.15 mm to 2.88 mm for lateral direction. Target position provided by deep learning model for the kV images acquired using OBI is found to be consistent that derived from the fiducials. This study demonstrates, for the first time, that highly accurate markerless prostate localization based on deep learning is achievable. The strategy provides a clinically valuable solution to daily patient positioning and real-time target tracking for image-guided radiotherapy (IGRT) and interventions.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper proposes a deep learning model, pre-trained on thousands of annotated kV projection X-ray images, to perform markerless prostate localization for image-guided radiotherapy. It reports quantitative evaluation on three retrospectively studied VMAT patients, with localization deviations from fiducial-derived annotations of 1.66–2.77 mm (AP) and 1.15–2.88 mm (lateral), concluding that highly accurate markerless localization is achievable and clinically valuable.

Significance. If the central claim holds under proper validation, the work would offer a non-invasive alternative to fiducial implantation, eliminating associated procedural risks and costs while enabling daily positioning and real-time tracking from routine kV images. The approach leverages standard OBI hardware and demonstrates consistency with fiducial ground truth on the tested cases.

major comments (2)
  1. [patient study / results] Evaluation on three patients (abstract and patient-study section): the reported deviations support mm-level accuracy on this small retrospective cohort, but the absence of cross-validation, inter-patient error breakdown, prospective hold-out testing, or details on whether test images come from training patients undermines the generalization claim required for the assertion that the model works on 'new patients without fiducials.' This is load-bearing for the headline result.
  2. [methods / abstract] Methods and abstract: no model architecture, training hyperparameters, loss function, data split strategy, or statistical tests (error bars, p-values) are supplied, preventing assessment of whether the 1.15–2.88 mm range reflects robust performance or overfitting to the limited annotated data.
minor comments (1)
  1. [introduction] The claim of demonstrating the result 'for the first time' would benefit from explicit comparison to prior markerless DL or template-matching approaches in the introduction.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive comments on our manuscript. We provide point-by-point responses to the major comments and indicate where revisions will be made.

read point-by-point responses

  1. Referee: [patient study / results] Evaluation on three patients (abstract and patient-study section): the reported deviations support mm-level accuracy on this small retrospective cohort, but the absence of cross-validation, inter-patient error breakdown, prospective hold-out testing, or details on whether test images come from training patients undermines the generalization claim required for the assertion that the model works on 'new patients without fiducials.' This is load-bearing for the headline result.

    Authors: We note that the deep learning model was pre-trained on several thousand annotated kV projection images, which forms the basis for its ability to generalize to new patients. The evaluation on three patients was performed retrospectively to compare against fiducial-derived positions, and the test images are from these patients not involved in the pre-training. We agree that additional analyses such as cross-validation and inter-patient breakdowns would strengthen the paper. We will revise to include an inter-patient error breakdown and clarify the data usage. However, as this is a retrospective study, prospective hold-out testing is not feasible in the current work, and we will discuss this limitation explicitly. revision: partial

  2. Referee: [methods / abstract] Methods and abstract: no model architecture, training hyperparameters, loss function, data split strategy, or statistical tests (error bars, p-values) are supplied, preventing assessment of whether the 1.15–2.88 mm range reflects robust performance or overfitting to the limited annotated data.

    Authors: We agree with this observation. The current version of the manuscript does not provide these details in the abstract or methods. We will revise the manuscript to include a detailed description of the model architecture, training hyperparameters, loss function, data split strategy, and incorporate statistical tests with error bars and p-values where appropriate to allow proper assessment of the results' robustness. revision: yes

standing simulated objections not resolved
  • Prospective hold-out testing on new patients (retrospective study design prevents this)

Circularity Check

0 steps flagged

No circularity; standard supervised DL training evaluated on fiducial ground truth

full rationale

The paper describes training a deep learning model on several thousand annotated kV projection images and then quantitatively comparing its outputs to the corresponding annotations on data from three retrospective VMAT patients. No equations, derivations, or parameter-fitting steps are presented that reduce a claimed prediction to its own inputs by construction. The evaluation relies on fiducial-derived annotations as external ground truth, which is independent of the model's learned mapping. No self-citations are invoked as load-bearing uniqueness theorems, and no ansatz or renaming of known results occurs. The result is therefore self-contained empirical validation rather than a circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on supervised learning from annotated projection images; no explicit free parameters, new physical entities, or ad-hoc axioms beyond standard assumptions of deep learning applicability to medical images are introduced.

axioms (1)
  • domain assumption Deep neural networks trained on annotated kV projection images can learn to localize the prostate in unseen images.
    The entire pipeline depends on this supervised-learning premise.

pith-pipeline@v0.9.0 · 5817 in / 1264 out tokens · 40544 ms · 2026-05-25T19:32:41.964266+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

  1. [1]

    Measurements of intrafraction motion and interfraction and intrafraction rotation of prostate by three- dimensional analysis of daily portal imaging with radiopaque markers,

    Aubry, J.-F., Beaulieu, L., Girouard, L.-M., Aubin, S., Tremblay, D., Laverdi` ere, J., and Vigneault, E., “Measurements of intrafraction motion and interfraction and intrafraction rotation of prostate by three- dimensional analysis of daily portal imaging with radiopaque markers,” International Journal of Radiation Oncology* Biology* Physics 60(1), 30–39 (2004)

    work page 2004

  2. [2]

    Prostate motion during standard radiother- apy as assessed by fiducial markers,

    Crook, J., Raymond, Y., Salhani, D., Yang, H., and Esche, B., “Prostate motion during standard radiother- apy as assessed by fiducial markers,” Radiotherapy and Oncology 37(1), 35–42 (1995)

    work page 1995

  3. [3]

    Intrafraction motion of the prostate during external-beam radiation therapy: analysis of 427 patients with implanted fiducial markers,

    Kotte, A. N., Hofman, P., Lagendijk, J. J., van Vulpen, M., and van der Heide, U. A., “Intrafraction motion of the prostate during external-beam radiation therapy: analysis of 427 patients with implanted fiducial markers,” International Journal of Radiation Oncology* Biology* Physics 69(2), 419–425 (2007)

    work page 2007

  4. [4]

    Prostate intrafraction motion assessed by simultaneous kilovoltage fluoroscopy at megavoltage delivery i: Clinical observations and pattern analysis,

    Adamson, J. and Wu, Q., “Prostate intrafraction motion assessed by simultaneous kilovoltage fluoroscopy at megavoltage delivery i: Clinical observations and pattern analysis,” International Journal of Radiation Oncology* Biology* Physics 78(5), 1563–1570 (2010)

    work page 2010

  5. [5]

    Overview of image-guided radiation therapy,

    Xing, L., Thorndyke, B., Schreibmann, E., Yang, Y., Li, T.-F., Kim, G.-Y., Luxton, G., and Koong, A., “Overview of image-guided radiation therapy,” Medical Dosimetry 31(2), 91–112 (2006)

    work page 2006

  6. [6]

    Advances in image-guided radiation therapy,

    Dawson, L. A. and Jaffray, D. A., “Advances in image-guided radiation therapy,” Journal of clinical oncol- ogy 25(8), 938–946 (2007)

    work page 2007

  7. [7]

    Clinical use of stereoscopic x-ray positioning of patients treated with confor- mal radiotherapy for prostate cancer,

    Soete, G., Verellen, D., Michielsen, D., Vinh-Hung, V., Van de Steene, J., Van den Berge, D., De Roover, P., Keuppens, F., and Storme, G., “Clinical use of stereoscopic x-ray positioning of patients treated with confor- mal radiotherapy for prostate cancer,”International Journal of Radiation Oncology* Biology* Physics 54(3), 948–952 (2002)

    work page 2002

  8. [8]

    Combined kv and mv imaging for real-time tracking of implanted fiducial markers a,

    Wiersma, R., Mao, W., and Xing, L., “Combined kv and mv imaging for real-time tracking of implanted fiducial markers a,” Medical physics 35(4), 1191–1198 (2008)

    work page 2008

  9. [9]

    Real-time 3d internal marker tracking during arc radiotherapy by the use of combined mv–kv imaging,

    Liu, W., Wiersma, R., Mao, W., Luxton, G., and Xing, L., “Real-time 3d internal marker tracking during arc radiotherapy by the use of combined mv–kv imaging,” Physics in Medicine & Biology 53(24), 7197 (2008)

    work page 2008

  10. [10]

    Flat-panel cone-beam computed tomography for image-guided radiation therapy,

    Jaffray, D. A., Siewerdsen, J. H., Wong, J. W., and Martinez, A. A., “Flat-panel cone-beam computed tomography for image-guided radiation therapy,” International Journal of Radiation Oncology* Biology* Physics 53(5), 1337–1349 (2002)

    work page 2002

  11. [11]

    Image-guided radiotherapy (igrt) for prostate cancer comparing kv imaging of fiducial markers with cone beam computed tomography (cbct),

    Barney, B. M., Lee, R. J., Handrahan, D., Welsh, K. T., Cook, J. T., and Sause, W. T., “Image-guided radiotherapy (igrt) for prostate cancer comparing kv imaging of fiducial markers with cone beam computed tomography (cbct),” International Journal of Radiation Oncology* Biology* Physics 80(1), 301–305 (2011)

    work page 2011

  12. [12]

    Online magnetic resonance image guided adaptive radiation therapy: first clinical applications,

    Acharya, S., Fischer-Valuck, B. W., Kashani, R., Parikh, P., Yang, D., Zhao, T., Green, O., Wooten, O., Li, H. H., Hu, Y., et al., “Online magnetic resonance image guided adaptive radiation therapy: first clinical applications,” International Journal of Radiation Oncology* Biology* Physics 94(2), 394–403 (2016)

    work page 2016

  13. [13]

    Review of ultrasound image guidance in external beam radiotherapy: I. treatment planning and inter-fraction motion management,

    Fontanarosa, D., Van der Meer, S., Bamber, J., Harris, E., OShea, T., and Verhaegen, F., “Review of ultrasound image guidance in external beam radiotherapy: I. treatment planning and inter-fraction motion management,” Physics in Medicine & Biology 60(3), R77 (2015)

    work page 2015

  14. [14]

    Improved clinical outcomes with high-dose image guided radiotherapy compared with non-igrt for the treatment of clinically localized prostate cancer,

    Zelefsky, M. J., Kollmeier, M., Cox, B., Fidaleo, A., Sperling, D., Pei, X., Carver, B., Coleman, J., Love- lock, M., and Hunt, M., “Improved clinical outcomes with high-dose image guided radiotherapy compared with non-igrt for the treatment of clinically localized prostate cancer,” International Journal of Radiation Oncology* Biology* Physics 84(1), 125–...

    work page 2012

  15. [15]

    Automated target tracking in kilovoltage images using dynamic templates of fiducial marker clusters,

    Campbell, W. G., Miften, M., and Jones, B. L., “Automated target tracking in kilovoltage images using dynamic templates of fiducial marker clusters,” Medical physics 44(2), 364–374 (2017)

    work page 2017

  16. [16]

    A magnetic resonance imaging study of prostate deformation relative to implanted gold fiducial markers,

    Nichol, A. M., Brock, K. K., Lockwood, G. A., Moseley, D. J., Rosewall, T., Warde, P. R., Catton, C. N., and Jaffray, D. A., “A magnetic resonance imaging study of prostate deformation relative to implanted gold fiducial markers,” International Journal of Radiation Oncology* Biology* Physics 67(1), 48–56 (2007)

    work page 2007

  17. [17]

    Deep learning mr imaging–based attenuation correction for pet/mr imaging,

    Liu, F., Jang, H., Kijowski, R., Bradshaw, T., and McMillan, A. B., “Deep learning mr imaging–based attenuation correction for pet/mr imaging,” Radiology 286(2), 676–684 (2017)

    work page 2017

  18. [18]

    Segmentation of organs-at-risks in head and neck ct images using convolutional neural networks,

    Ibragimov, B. and Xing, L., “Segmentation of organs-at-risks in head and neck ct images using convolutional neural networks,” Medical physics 44(2), 547–557 (2017)

    work page 2017

  19. [19]

    Superpixel-based and boundary-sensitive convolutional neural network for automated liver segmentation,

    Qin, W., Wu, J., Han, F., Yuan, Y., Zhao, W., Ibragimov, B., Gu, J., and Xing, L., “Superpixel-based and boundary-sensitive convolutional neural network for automated liver segmentation,” Physics in Medicine & Biology 63(9), 095017 (2018)

    work page 2018

  20. [20]

    Fast r-cnn,

    Girshick, R., “Fast r-cnn,” in [ Proceedings of the IEEE international conference on computer vision ], 1440– 1448 (2015)

    work page 2015

  21. [21]

    Faster r-cnn: towards real-time object detection with region proposal networks,

    Ren, S., He, K., Girshick, R., and Sun, J., “Faster r-cnn: towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis & Machine Intelligence (6), 1137–1149 (2017)

    work page 2017

  22. [22]

    Visualiz- ing the invisible in prostate radiation therapy: markerless prostate target localization via a deep learning model and monoscopic kv projection x-ray image,

    Zhao, W., Han, B., Yang, Y., Buyyounouski, M., Hancock, S., Bagshaw, H., and Xing, L., “Visualiz- ing the invisible in prostate radiation therapy: markerless prostate target localization via a deep learning model and monoscopic kv projection x-ray image,” International Journal of Radiation Oncology? Biology? Physics 102(3), S128–S129 (2018)

    work page 2018

  23. [23]

    Incorporating deep layer image information into image guided radiation therapy,

    Zhao, W., Han, B., Yang, Y., Buyyounouski, M., Hancock, S., Bagshaw, H., and Xing, L., “Incorporating deep layer image information into image guided radiation therapy,” in [MEDICAL PHYSICS ], 45(6), E686– E686 (2018)

    work page 2018

  24. [24]

    Deep convolutional neural networks for computer-aided detection: Cnn architectures, dataset characteris- tics and transfer learning,

    Hoo-Chang, S., Roth, H. R., Gao, M., Lu, L., Xu, Z., Nogues, I., Yao, J., Mollura, D., and Summers, R. M., “Deep convolutional neural networks for computer-aided detection: Cnn architectures, dataset characteris- tics and transfer learning,” IEEE transactions on medical imaging 35(5), 1285 (2016)

    work page 2016

  25. [25]

    Mark- erless pancreatic tumor target localization enabled by deep learning,

    Zhao, W., Shen, L., Han, B., Yang, Y., Cheng, K., Toesca, D., Koong, A., Chang, D., and Xing, L., “Mark- erless pancreatic tumor target localization enabled by deep learning,” International Journal of Radiation Oncology? Biology? Physics , In press (2019)

    work page 2019