Pith. sign in

REVIEW 3 major objections 5 minor 48 references

A 60-second, ungated MRI scan can produce five-dimensional cardiac motion maps in about seven minutes, fitting the planning window for MRI-guided arrhythmia ablation on a 1.5 T MR-linac.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 21:32 UTC pith:WTDKU7Y2

load-bearing objection Solid engineering with a credible 7-minute 5D-MRI pipeline, but the in-vivo validation is thinner than the abstract implies and the fixed cardiac frequency band may not hold for the intended VT population. the 3 major comments →

arxiv 2607.16033 v1 pith:WTDKU7Y2 submitted 2026-07-17 physics.med-ph

Fast ungated five-dimensional cardiac MRI on a 1.5 T MR-linac for MRI-guided radiotherapy

M.L. Terpstra , T.E. Olausson , M.M.N. Aubert , C. Beijst , A. Sbrizzi , C.A.T. van den Berg , M.F. Fast This is my paper
classification physics.med-ph
keywords 5D-MRIMR-linaccardiorespiratory motionlow-rank deformation vector fieldsungated MRImotion disentanglementstereotactic arrhythmia radioablationretrospective binning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper aims to make five-dimensional (5D) MRI, which resolves both cardiac and respiratory motion, fast enough for clinical use in stereotactic arrhythmia radio-ablation (STAR) on an MR-linac. The authors propose 5D CMR-MOTUS, a method that reconstructs a motion-corrected reference image and low-rank deformation vector fields (DVFs) from a one-minute free-running, ungated Cartesian acquisition, then splits the motion into cardiac and respiratory components by temporal frequency. The reconstruction takes six minutes, giving a total latency of roughly seven minutes, within the ten-minute planning window of adaptive STAR treatments. If this holds, 5D-MRI becomes a practical tool for personalized motion management in cardiac radiotherapy.

Core claim

The paper reports that accurate 5D-MRI can be obtained on a 1.5 T MR-linac with a 60-second acquisition and six minutes of reconstruction, yielding a total end-to-end latency of about seven minutes. The key is to jointly optimize a motion-corrected reference image and a low-rank factorization of deformation vector fields, where the temporal basis is explicitly split into respiratory (0.1–0.4 Hz) and cardiac (0.7–2 Hz) bands. This low-rank DVF then serves as a self-navigator: binning its temporal components reconstructs any desired number of respiratory and cardiac phases retrospectively, without re-running the reconstruction. Validation with digital and physical phantoms and ten healthy volu

What carries the argument

The central object is the low-rank deformation vector field (DVF), written as D ≈ ΦΨᵀ, where Φ is a spatial basis of cubic B-splines and Ψ is a temporal basis. The temporal basis is designed to separate physiology by frequency band: three components for respiratory motion (0.1–0.4 Hz), five for cardiac motion (0.7–2 Hz), and one for drift below 0.1 Hz. This low-rank model is what makes the joint reconstruction of reference image and motion data-efficient, and the same temporal basis serves as a self-navigator for assigning motion states, enabling retrospective selection of phase counts.

Load-bearing premise

The method assumes that cardiac and respiratory motion are perfectly separable by temporal frequency bands and that all relevant physiological motion can be represented by smooth, low-rank deformation vector fields; if heart–lung coupling or non-smooth motion occurs, the reconstructed 5D states will mislabel or miss motion.

What would settle it

Take a patient or phantom with coupled cardiorespiratory motion (e.g., a heart rate near 0.5–0.6 Hz overlapping the respiratory band, or a known cardiac-respiratory coupling) and compare 5D cardiac-phase images against a gated reference: if the heart boundary appears blurred across respiratory bins or the measured cardiac motion error relative to 2D cine exceeds roughly 3 mm, the frequency-based disentanglement and smooth-DVF assumptions would be shown to fail.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • 5D-MRI can be acquired and reconstructed within the ~10-minute planning window of MRI-guided STAR, enabling personalized cardiorespiratory motion models for target margins and gating decisions.
  • The number of cardiac and respiratory phases can be chosen after reconstruction without extra cost, so the same acquisition can serve different motion-management strategies.
  • Because all k-space data contribute to every motion state (via the reference image and DVFs), low motion-state-specific SNR is avoided.
  • The method removes the need for ECG gating or external respiratory navigation, simplifying the clinical workflow.
  • The publicly released k-space data and reconstructions for ten volunteers provide a benchmark for future 5D-MRI development on MR-linacs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the frequency-band separation holds in real patients with arrhythmia or structural heart disease, this approach could be extended to other moving organs on MR-linacs (e.g., lung or abdominal targets) with similar low-rank motion models.
  • A patient with significant heart–lung coupling, where cardiac and respiratory frequencies overlap or interact, would likely violate the separation assumption; testing on such a cohort would define the method's true clinical envelope.
  • The framework's assumption of smooth deformation fields means it will miss blood-flow-related intensity changes and sliding tissue boundaries; a natural extension is to add a residual image component or alternative bases.
  • The one-minute acquisition plus six-minute reconstruction time suggests that the same approach could support real-time motion monitoring if the optimization is accelerated, though the paper does not claim this.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes '5D CMR-MOTUS,' a method for free-running, ungated 5D cardiac-respiratory MRI on a 1.5 T MR-linac. The method extends CMR-MOTUS by jointly reconstructing a motion-corrected reference image and a low-rank deformation vector field (DVF) factorization, then explicitly partitioning the temporal basis into respiratory (0.1–0.4 Hz), cardiac (0.7–2 Hz), and drift (<0.1 Hz) components. After optimization, the temporal components are used as navigators to bin the DVFs into retrospectively adjustable cardiorespiratory motion states, yielding 5D-MRI by warping the reference image. Validation comprises a digital XCAT/MRXCAT phantom, a deformable physical phantom with simultaneous cardiac and respiratory motion, and 10 healthy volunteers. The authors report a 60-second acquisition and 6-minute reconstruction on an NVIDIA L40S, claim accurate 5D-MRI with a total latency of approximately 7 minutes, and make the volunteer k-space data and reconstructions publicly available.

Significance. If the result holds, the method would be an important step toward integrating 5D-MRI into MRI-guided STAR workflows, where the 10-minute planning window is currently a bottleneck. The use of low-rank DVF disentanglement with retrospective state binning is a novel and potentially powerful alternative to k-space binning methods, and the public release of data is a strength. The phantom experiments are independent and the physical phantom DICE of 0.96 is impressive. However, the central claim of 'accurate 5D-MRI' is validated only under conditions that match the method's assumptions — sinusoidal phantom motion and healthy-volunteer sinus rhythm — and the in-vivo validation is limited to motion-magnitude agreement with 2D cine, not state-level ground truth. The specific clinical target population (VT patients) may violate the fixed cardiac frequency band and smooth-DVF assumptions, so the reported accuracy is not yet established for the intended use.

major comments (3)
  1. [Methods — Estimating motion states] The disentanglement into respiratory and cardiac components relies on fixed temporal frequency bands (respiratory 0.1–0.4 Hz, cardiac 0.7–2 Hz) with five cardiac components. All experiments use sinus rhythm or sinusoidal 1 Hz cardiac motion, which falls comfortably within the assumed passband. The intended STAR population includes ventricular tachycardia patients, who may have heart rates above 120 bpm (>2 Hz) and irregular rhythms. Under those conditions the cardiac motion would be partially or completely outside the modeled band, so the five cardiac components would fail to capture it, and energy could leak into the respiratory components, mislabeling the reconstructed 5D states. The in-vivo comparison (95th-percentile DVF magnitudes against 2D cine) does not test state correctness or phase alignment. To support the central 7-minute/accuracy claim, I would require a sensitivity analysi
  2. [Methods — Physical phantom reconstruction] The physical phantom experiment uses rank-one translational motion for respiratory deformation and a rank-one cubic B-spline basis for cardiac deformation. This does not exercise the actual 9-component temporal bandpass disentanglement that is the methodological novelty; it validates only that a low-rank DVF model can represent simple sinusoidal motions. The digital phantom uses sinusoidal waveforms whose frequencies exactly match the assumed bands (1 Hz cardiac, 0.2 Hz respiratory). Neither experiment tests the behavior when cardiac and respiratory frequencies overlap or when higher-order/coupled motion is present. Given that the Discussion acknowledges 'cardiorespiratory motion coupling can prevent this separation,' a phantom or simulation with coupled or overlapping frequency content would be a more load-bearing validation of the disentanglement claim.
  3. [Methods — In-vivo MRI reconstruction] The in-vivo validation reports respiratory and cardiac motion-magnitude errors of 0.2 ± 2.9 mm and 0.07 ± 0.9 mm, respectively, but the respiratory mean absolute error is 2.31 ± 1.47 mm and the standard deviation of the respiratory amplitude difference is 5.5 mm. These metrics are based on 95th-percentile DVF magnitudes computed in a single coronal slice and compared with a separately acquired 2D cine, not with ground-truth 5D motion states. This does not establish that the binned 5D states are correct in phase or in 3D spatial distribution. The Discussion partially acknowledges the amplitude discrepancy, but the conclusion's phrase 'accurate 5D-MRI can be obtained' is stronger than what the in-vivo data support. A state-level validation (e.g., ventricular volume curves, landmark tracking, or comparison against ECG-triggered high-resolution images) would be needed to fully support the ce
minor comments (5)
  1. [Author affiliations] Typo: 'Computatinal Imaging Group' should be 'Computational Imaging Group.'
  2. [Results — Digital phantom] The wording 'PSNR is minimal when the spatial resolution is >3 mm' is awkward; 'lowest' or 'minimal PSNR is observed' would be clearer.
  3. [Figure 8 caption] The caption states 'identical colors indicate identical volunteers' but the figure is not shown in full color in the manuscript text; consider ensuring the color legend is visible in the printed version.
  4. [Throughout] DICE/Dice capitalization is inconsistent; use 'Dice' consistently as a proper noun.
  5. [Abstract vs. Results] The abstract reports 'cardiac motion error of 0.1 ± 0.9 mm' while the Results report 'cardiac motion magnitude error of 0.07 ± 0.9 mm'; unify the terminology and values.

Circularity Check

0 steps flagged

No significant circularity: the 5D CMR-MOTUS feasibility result is validated against independent phantoms and 2D cine MRI, not derived from fitted inputs or self-citations.

full rationale

The paper's derivation chain is not circular. The core reconstruction jointly optimizes a reference image and low-rank DVFs against measured k-space (Eq. 1); the respiratory/cardiac disentanglement is obtained by band-limiting the temporal basis Ψ (Methods, 'Estimating motion states'), and the 5D states are constructed by binning Ψ and warping the reference image. This is a model-based reconstruction, not a prediction derived from fitted parameters. The central accuracy claims are checked against independent targets: XCAT/MRXCAT ground-truth DVFs, a physical phantom with static ground-truth scans and DICE, and 2D cine MRI in volunteers. In each case the comparison quantity is measured separately from the reconstructed 5D fields. The paper's self-citations to CMR-MOTUS (ref. 13) and low-rank MR-MOTUS (refs. 24, 25) supply the base model and a rank/number-of-components heuristic, but they do not by themselves force the 60-second/6-minute feasibility result, which is established empirically in this paper. The acknowledged limitations—frequency-band separability and smooth-DVF assumptions—are stated assumptions, not circular reductions. No equation or fitted parameter is relabeled as a prediction. Therefore no circular step is present.

Axiom & Free-Parameter Ledger

7 free parameters · 5 axioms · 0 invented entities

The method adds no new physical constants or entities. It depends on a set of reconstruction hyperparameters (listed as free parameters) and three modeling assumptions: low-rank smooth deformation, frequency-separable cardiorespiratory motion, and the CMR-MOTUS forward model. The most fragile is the frequency-separation assumption, which the authors themselves flag in the Discussion.

free parameters (7)
  • Low-rank component counts = 3 respiratory + 5 cardiac + 1 drift = 9
    Determines how much motion the model can represent; chosen based on prior work (ref 24) and empirical tuning.
  • Temporal bandpass cutoffs = resp 0.1–0.4 Hz, cardiac 0.7–2 Hz
    Assumes cardiac and respiratory frequency separation; fails if the bands overlap.
  • L1 wavelet penalty weight λ = 1e-3
    Regularization strength set empirically during algorithm development.
  • Optimizer learning rates and schedule = 0.02 for Φ/Ψ, 0.075 for reference/coils, gamma=0.8
    Empirically established; affects convergence and final image/motion quality.
  • Spline basis resolution = 15 splines/dim increasing to d/2 over 10 epochs
    Coarse-to-fine spatial regularization, hand-chosen.
  • OPRA phase encodes per leaflet = 26
    Sets temporal resolution (~10 Hz) vs sampling density; chosen as a balance.
  • Number of motion-state bins = 10×10, also 20×20
    Retrospective choice; affects state definition but not reconstruction cost.
axioms (5)
  • ad hoc to paper Respiratory and cardiac motion occupy non-overlapping temporal frequency bands (0.2–0.4 Hz vs 0.8–2 Hz).
    Used to partition the temporal basis into Ψ_resp and Ψ_cardiac (Methods, 'Estimating motion states'); breaks down for cardiorespiratory coupling or high respiratory rates.
  • domain assumption All physiological motion can be represented by smooth cubic B-spline DVFs in a low-rank subspace.
    The model D ≈ ΦΨ^T with spline spatial basis cannot capture blood flow or sliding organ motion; acknowledged in Discussion.
  • domain assumption A single motion-corrected reference image plus time-resolved DVFs explains the measured k-space through the warped Fourier forward operator.
    Core CMR-MOTUS signal model inherited from refs 13, 24, 25.
  • domain assumption OPRA Cartesian sampling provides temporally incoherent coverage sufficient for joint reconstruction.
    Assumes sampling artifacts are acceptable at the tested resolutions, scan times, and motion amplitudes.
  • standard math ESPIRiT coil sensitivity maps and the bSSFP/GRE signal equations describe the acquired data.
    Used in the forward operator; standard MRI reconstruction assumptions.

pith-pipeline@v1.3.0-alltime-deepseek · 12004 in / 14986 out tokens · 136682 ms · 2026-08-01T21:32:14.098922+00:00 · methodology

0 comments
read the original abstract

Background: Stereotactic arrhythmia radio-ablation (STAR) for patients with ventricular tachycardia is currently limited by complex cardiorespiratory motion. Current 5D-MRI motion models require long acquisition and reconstruction times, limiting clinical viability. Objective: To develop a fast, ungated 5D-MRI reconstruction method for personalized motion characterization to support MRI-guided STAR treatments. Methods: We propose a fast, ungated 5D-MRI reconstruction method based on the CMR-MOTUS framework. The method uses a 3D Cartesian acquisition with a joint optimization framework to reconstruct a motion-corrected reference image and low-rank deformation vector fields (DVFs). By exploiting the low rank structure, we explicitly disentangle respiratory and cardiac motion during optimization. Then, the DVFs are used for 5D-MRI reconstruction with a retrospectively adjustable number of motion states. Validation was performed using digital and physical cardiorespiratory phantoms. Furthermore, the approach was evaluated using 10 healthy volunteers, comparing motion consistency with 2D cine MRI. Results: Validation of 5D CMR-MOTUS using digital and physical phantoms demonstrated accurate 5D-MRI reconstruction. In the physical phantom, 5D CMR-MOTUS achieved a left-ventricle DICE of 0.96 +/- 0.01. In the volunteer cohort, the 5D-MRI scans showed strong motion to 2D cine MRI, with a cardiac motion error of 0.1 +/- 0.9 mm and a respiratory motion error of 0.2 +/- 2.9 mm. Crucially, 5D-MRI data were acquired in 1 minute and reconstructed in 6 minutes. Conclusions: The proposed 5D-MRI method enables rapid, high-quality, and personalized motion characterization, demonstrating potential for integration into MRI-guided STAR treatments. Data Availability: The 3D k-space data and 5D reconstructions for the ten volunteers are publicly available at https://doi.org/10.5281/zenodo.21278894

Figures

Figures reproduced from arXiv: 2607.16033 by A. Sbrizzi, C.A.T. van den Berg, C. Beijst, M.F. Fast, M.L. Terpstra, M.M.N. Aubert, T.E. Olausson.

Figure 1
Figure 1. Figure 1: Method overview. To reconstruct our 5D-MRI, free-running time-resolved k￾space is acquired using the OPRA sampling pattern (left). Using the 5D CMR-MOTUS algorithm (middle), the motion-corrected reference image and a low-rank approximation of time-resolved DVFs (green), consisting of explicitly disentangled spatial (𝜱) and temporal (𝜳) bases, are jointly optimized to minimize the difference to the sampled … view at source ↗
Figure 2
Figure 2. Figure 2: Experiment setup. The deformable cardiac phantom is placed onto a motion platform, enabling simultaneous cardiac and respiratory motion (left). The coil was placed over the phantom “ventricles”, while the motor drives the phantom with 0.4Hz 20 mm respiratory amplitude, and 1Hz 10 mm cardiac deformation amplitude. An example static coronal MRI scan of the phantom is shown on the right. The dynamic data were… view at source ↗
Figure 3
Figure 3. Figure 3: Example MRXCAT reconstruction. On the left, the ground-truth MRXCAT digital phantom simulations, while the right side shows 5D CMR-MOTUS reconstructions using a one-minute acquisition with a 10Hz temporal resolution [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Quantitative digital phantom results. Comparing the estimated 5D-MRI to the ground-truth MRXCAT 5D-MRI, the PSNR is minimal when the spatial resolution is > 3 mm, (top). The image quality significantly decreases when the acquisition is shorter than 60 seconds (middle). To resolve cardiac and respiratory motion, sufficient performance is obtained when the temporal resolution is at least 4 Hz (bottom) [PITH… view at source ↗
Figure 5
Figure 5. Figure 5: Physical phantom results. The top row shows the images in systole, while the bottom row show diastole motion state. The columns of each reconstruction show exhale and inhale, respectively. The XD-GRASP reconstruction suffers from significant artifacts, whereas the 5D CMR-MOTUS accurately captures cardiorespiratory motion. The 5D CMR￾MOTUS and XD-GRASP reconstructions suffer from banding artifacts due to th… view at source ↗
Figure 6
Figure 6. Figure 6: Reconstruction overview. Here, an overview of the short-axis reconstructions for all 10 volunteers across four extreme motion states (Diastole/inhale, systole/inhale, diastole/exhale, and systole/exhale) is shown, illustrating motion-corrected images and cardiorespiratory disentanglement between the motion states. An animated figure showing all motion states is provided in Supplementary [PITH_FULL_IMAGE:f… view at source ↗
Figure 7
Figure 7. Figure 7: Cardiac orientations. The four-chamber, short-axis, and two-chamber orientations are shown for one volunteer across four extreme motion states (Diastole/inhale, systole/inhale, diastole/exhale, and systole/exhale), demonstrating high image quality and motion disentanglement between motion states. An animated figure showing all motion states is provided in Supplementary [PITH_FULL_IMAGE:figures/full_fig_p0… view at source ↗

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

48 extracted references · 28 canonical work pages · 2 internal anchors

  1. [1]

    Terpstra1,†, T.E

    1 Fast ungated five-dimensional cardiac MRI on a 1.5 T MR-linac for MRI-guided radiotherapy M.L. Terpstra1,†, T.E. Olausson1,†, M.M.N. Aubert2, C. Beijst2, A. Sbrizzi1, C.A.T. van den Berg1, M.F. Fast2 1Computatinal Imaging Group for MRI Therapy & Diagnostics, Center of Image Sciences, University Medical Center Utrecht, Utrecht, the Netherlands 2Departmen...

  2. [2]

    The 5D CMR-MOTUS framework was implemented by extending the CMR-MOTUS framework (Olausson, Terpstra, Ahmad, et al., 2026). Using the autodifferentiation capabilities of the PyTorch framework, we perform joint stochastic optimization in time of the reference image 𝒒 and the low-rank deformation motion field components 𝚽𝚿%for 15 epochs using minibatches of ...

    2026

  3. [3]

    2D-cine MRI

    The field of view was chosen to include the body contour in transverse view, which is essential for dose calculation when considering STAR treatments. In total, approximately 15800 k-space phase encodes were obtained for each volunteer for each 5D-MRI acquisition, and reconstructed using the 5D CMR-MOTUS algorithm into 10 respiratory motion states and 10 ...

  4. [4]

    Stereotactic Arrhythmia Radioablation (STAR): Assessment of cardiac and respiratory heart motion in ventricular tachycardia patients - A STOPSTORM.eu consortium review

    Stevens RRF, Hazelaar C, Fast MF, et al. Stereotactic Arrhythmia Radioablation (STAR): Assessment of cardiac and respiratory heart motion in ventricular tachycardia patients - A STOPSTORM.eu consortium review. Radiother Oncol. 2023;188:109844. doi:10.1016/j.radonc.2023.109844

    arXiv 2023

  5. [5]

    Analyzing Cardiorespiratory Motion and Its Dosimetric Effect on Stereotactic Arrhythmia Radio-Ablation: A STOPSTORM.eu Consortium Study

    Van Der Pol LHG, Mandija S, Balgobind BV, et al. Analyzing Cardiorespiratory Motion and Its Dosimetric Effect on Stereotactic Arrhythmia Radio-Ablation: A STOPSTORM.eu Consortium Study. Int J Radiat Oncol. 2026;124(3):799-809. doi:10.1016/j.ijrobp.2025.09.039 21

    work page doi:10.1016/j.ijrobp.2025.09.039 2026

  6. [6]

    Noninvasive Cardiac Radiation for Ablation of Ventricular Tachycardia

    Cuculich PS, Schill MR, Kashani R, et al. Noninvasive Cardiac Radiation for Ablation of Ventricular Tachycardia. N Engl J Med. 2017;377(24):2325-2336. doi:10.1056/nejmoa1613773

    doi:10.1056/nejmoa1613773 2017

  7. [7]

    Acknowledgments We gratefully acknowledge the help of Katrinus Keijnemans and Edwin Versteeg for their technical support

    Both supplementary figures can be found at https://doi.org/10.5281/zenodo.21293931. Acknowledgments We gratefully acknowledge the help of Katrinus Keijnemans and Edwin Versteeg for their technical support. This work was supported by the Dutch Research Council (NWO) through the PAMPER (19003) and MEGAHERTZ (19484) projects. Author contributions CRediT: ML ...

    work page doi:10.5281/zenodo.21293931

  8. [8]

    Stereotactic arrhythmia radioablation (STAR)—A systematic review and meta-analysis of prospective trials on behalf of the STOPSTORM.eu consortium

    Miszczyk M, Hoeksema WF, Kuna K, et al. Stereotactic arrhythmia radioablation (STAR)—A systematic review and meta-analysis of prospective trials on behalf of the STOPSTORM.eu consortium. Heart Rhythm. 2025;22(1):80-89. doi:10.1016/j.hrthm.2024.07.029

    doi:10.1016/j.hrthm.2024.07.029 2025

  9. [9]

    Stereotactic arrhythmia radioablation for refractory ventricular tachycardia: the STOPSTORM.eu study

    Van Der Pol LHG, Tomasik B, Hoeksema WF, et al. Stereotactic arrhythmia radioablation for refractory ventricular tachycardia: the STOPSTORM.eu study. Eur Heart J. Published online April 20, 2026:ehag338. doi:10.1093/eurheartj/ehag338

    work page doi:10.1093/eurheartj/ehag338 2026

  10. [10]

    Magnetic Resonance-Guided Stereotactic Radioablation for Septal Ventricular Tachycardias

    Bianchi S, Marchesano D, Magnocavallo M, et al. Magnetic Resonance-Guided Stereotactic Radioablation for Septal Ventricular Tachycardias. JACC Clin Electrophysiol. 2024;10(12):2569-2580. doi:10.1016/j.jacep.2024.08.008

    doi:10.1016/j.jacep.2024.08.008 2024

  11. [11]

    A motion analysis of cardiac substructures for guiding stereotactic arrhythmia radiotherapy motion management

    Wang Y, McKeown T, Hao Y, et al. A motion analysis of cardiac substructures for guiding stereotactic arrhythmia radiotherapy motion management. Med Phys. 2025;52(9). doi:10.1002/mp.18115

    work page doi:10.1002/mp.18115 2025

  12. [12]

    The Alberta Rotating Biplanar Linac-MR, a.k.a., Aurora-RTTM

    Fallone BG, Rathee S, de Zanche N, Yip E, Wachowicz K, Yun J. The Alberta Rotating Biplanar Linac-MR, a.k.a., Aurora-RTTM. In: A Practical Guide to MR-Linac. Springer International Publishing; 2024:193-215. doi:10.1007/978-3-031-48165-9_11

    work page doi:10.1007/978-3-031-48165-9_11 2024

  13. [13]

    The ViewRay System: Magnetic Resonance–Guided and Controlled Radiotherapy

    Mutic S, Dempsey JF. The ViewRay System: Magnetic Resonance–Guided and Controlled Radiotherapy. Semin Radiat Oncol. 2014;24(3):196-199. doi:10.1016/j.semradonc.2014.02.008

    work page doi:10.1016/j.semradonc.2014.02.008 2014

  14. [14]

    First patients treated with a 1.5 T MRI-Linac: clinical proof of concept of a high-precision, high-field MRI guided radiotherapy treatment

    Raaymakers BW, Jürgenliemk-Schulz IM, Bol GH, et al. First patients treated with a 1.5 T MRI-Linac: clinical proof of concept of a high-precision, high-field MRI guided radiotherapy treatment. Phys Med Biol. 2017;62(23):L41-L50. doi:10.1088/1361-6560/aa9517

    work page doi:10.1088/1361-6560/aa9517 2017

  15. [15]

    Deducing cardiorespiratory motion of cardiac substructures using a novel 5D-MRI workflow for radiotherapy

    Ruff C, Naren T, Wieben O, et al. Deducing cardiorespiratory motion of cardiac substructures using a novel 5D-MRI workflow for radiotherapy. Phys Med Biol. 2026;71(9):095007. doi:10.1088/1361-6560/ae5752

    work page doi:10.1088/1361-6560/ae5752 2026

  16. [16]

    First magnetic resonance imaging-guided cardiac radioablation of sustained ventricular tachycardia

    Mayinger M, Kovacs B, Tanadini-Lang S, et al. First magnetic resonance imaging-guided cardiac radioablation of sustained ventricular tachycardia. Radiother Oncol. 2020;152:203-207. doi:10.1016/j.radonc.2020.01.008

    doi:10.1016/j.radonc.2020.01.008 2020

  17. [17]

    A hybrid 2D/4D-MRI methodology using simultaneous multislice imaging for radiotherapy guidance

    Keijnemans K, Borman PTS, Uijtewaal P, Woodhead PL, Raaymakers BW, Fast MF. A hybrid 2D/4D-MRI methodology using simultaneous multislice imaging for radiotherapy guidance. Med Phys. 2022;49(9):6068-6081. doi:10.1002/mp.15802

    work page doi:10.1002/mp.15802 2022

  18. [18]

    MRI-guidance for motion management in external beam radiotherapy: current status and future challenges

    Paganelli C, Whelan B, Peroni M, et al. MRI-guidance for motion management in external beam radiotherapy: current status and future challenges. Phys Med Biol. 2018;63(22):22TR03. doi:10.1088/1361-6560/aaebcf

    doi:10.1088/1361-6560/aaebcf 2018

  19. [19]

    Margins to account for cardiac and respiratory motion in cardiac radioablation

    Marshall J, Poon J, Bergman A, et al. Margins to account for cardiac and respiratory motion in cardiac radioablation. Med Phys. 2025;52(10):e70041. doi:10.1002/mp.70041

    work page doi:10.1002/mp.70041 2025

  20. [20]

    5D whole-heart sparse MRI

    Feng L, Coppo S, Piccini D, et al. 5D whole-heart sparse MRI. Magn Reson Med. 2018;79(2):826-838. doi:10.1002/mrm.26745

    doi:10.1002/mrm.26745 2018

  21. [21]

    First experimental exploration of real-time cardiorespiratory motion management for future stereotactic arrhythmia radioablation treatments on the MR-linac

    Akdag O, Borman PTS, Woodhead P, et al. First experimental exploration of real-time cardiorespiratory motion management for future stereotactic arrhythmia radioablation treatments on the MR-linac. Phys Med Biol. 2022;67(6):065003. doi:10.1088/1361-6560/ac5717 22

    work page doi:10.1088/1361-6560/ac5717 2022

  22. [22]

    Radiation Therapy Workflow and Dosimetric Analysis from a Phase 1/2 Trial of Noninvasive Cardiac Radioablation for Ventricular Tachycardia

    Knutson NC, Samson PP, Hugo GD, et al. Radiation Therapy Workflow and Dosimetric Analysis from a Phase 1/2 Trial of Noninvasive Cardiac Radioablation for Ventricular Tachycardia. Int J Radiat Oncol. 2019;104(5):1114-1123. doi:10.1016/j.ijrobp.2019.04.005

    work page doi:10.1016/j.ijrobp.2019.04.005 2019

  23. [23]

    We have demonstrated that 5D CMR-MOTUS accurately captures cardiorespiratory motion using digital and physical phantoms

    On the other hand, OPRA provided a good trade-off between reduced computational cost, temporal incoherency, and greater resilience to system imperfections due to its minimal jumps in k-space. We have demonstrated that 5D CMR-MOTUS accurately captures cardiorespiratory motion using digital and physical phantoms. For MRI-guided STAR, motion quality and targ...

  24. [24]

    An automated approach to fully self-gated free-running cardiac and respiratory motion-resolved 5D whole-heart MRI

    Di Sopra L, Piccini D, Coppo S, Stuber M, Yerly J. An automated approach to fully self-gated free-running cardiac and respiratory motion-resolved 5D whole-heart MRI. Magn Reson Med. 2019;82(6):2118-2132. doi:10.1002/mrm.27898

    work page doi:10.1002/mrm.27898 2019

  25. [25]

    Nonrigid 3D motion estimation at high temporal resolution from prospectively undersampled k-space data using low-rank MR-MOTUS

    Huttinga NRF, Bruijnen T, van den Berg CAT, Sbrizzi A. Nonrigid 3D motion estimation at high temporal resolution from prospectively undersampled k-space data using low-rank MR-MOTUS. Magn Reson Med. 2021;85(4):2309-2326. doi:10.1002/mrm.28562

    work page doi:10.1002/mrm.28562 2021

  26. [26]

    Optimizing 4-Dimensional Magnetic Resonance Imaging Data Sampling for Respiratory Motion Analysis of Pancreatic Tumors

    Stemkens B, Tijssen RHN, De Senneville BD, et al. Optimizing 4-Dimensional Magnetic Resonance Imaging Data Sampling for Respiratory Motion Analysis of Pancreatic Tumors. Int J Radiat Oncol. 2015;91(3):571-578. doi:10.1016/j.ijrobp.2014.10.050

    work page doi:10.1016/j.ijrobp.2014.10.050 2015

  27. [27]

    Accelerated reconstruction of 5D free-running MRI with variable projection-augmented Lagrangian (VPAL)

    Yang Y, Naeem M, Van Assen M, et al. Accelerated reconstruction of 5D free-running MRI with variable projection-augmented Lagrangian (VPAL). Magn Reson Imaging. 2026;129:110643. doi:10.1016/j.mri.2026.110643

    arXiv 2026

  28. [28]

    On NUFFT-based gridding for non-Cartesian MRI

    Fessler JA. On NUFFT-based gridding for non-Cartesian MRI. J Magn Reson. 2007;188(2):191-195. doi:10.1016/j.jmr.2007.06.012

    work page doi:10.1016/j.jmr.2007.06.012 2007

  29. [29]

    MR-MOTUS: model-based non-rigid motion estimation for MR-guided radiotherapy using a reference image and minimal k -space data

    Huttinga NRF, Van Den Berg CAT, Luijten PR, Sbrizzi A. MR-MOTUS: model-based non-rigid motion estimation for MR-guided radiotherapy using a reference image and minimal k -space data. Phys Med Biol. 2020;65(1):015004. doi:10.1088/1361-6560/ab554a

    work page doi:10.1088/1361-6560/ab554a 2020

  30. [31]

    4D XCAT phantom for multimodality imaging research

    Segars WP, Sturgeon G, Mendonca S, Grimes J, Tsui BMW. 4D XCAT phantom for multimodality imaging research. Med Phys. 2010;37(9):4902-4915. doi:10.1118/1.3480985

    doi:10.1118/1.3480985 2010

  31. [32]

    Technical Report (v1.0)--Pseudo-random Cartesian Sampling for Dynamic MRI

    Joshi M, Pruitt A, Chen C, Liu Y, Ahmad R. Technical Report (v1.0)--Pseudo-random Cartesian Sampling for Dynamic MRI. arXiv. Preprint posted online June 8, 2022:arXiv:2206.03630. doi:10.48550/arXiv.2206.03630 23

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2206.03630 2022

  32. [33]

    ESPIRiT--an eigenvalue approach to autocalibrating parallel MRI: where SENSE meets GRAPPA

    Uecker M, Lai P, Murphy MJ, et al. ESPIRiT--an eigenvalue approach to autocalibrating parallel MRI: where SENSE meets GRAPPA. Magn Reson Med. 2014;71(3):990-1001. doi:10.1002/mrm.24751

    doi:10.1002/mrm.24751 2014

  33. [34]

    A deep learning framework for unsupervised affine and deformable image registration

    De Vos BD, Berendsen FF, Viergever MA, Sokooti H, Staring M, Išgum I. A deep learning framework for unsupervised affine and deformable image registration. Med Image Anal. 2019;52:128-143. doi:10.1016/j.media.2018.11.010

    doi:10.1016/j.media.2018.11.010 2019

  34. [36]

    Free-running time-resolved first-pass myocardial perfusion using a multi-scale dynamics decomposition: CMR-MOTUS

    Olausson TE, Terpstra ML, Huttinga NRF, et al. Free-running time-resolved first-pass myocardial perfusion using a multi-scale dynamics decomposition: CMR-MOTUS. Magma. 2026;39(2):173-186. doi:10.1007/s10334-025-01291-x

    work page doi:10.1007/s10334-025-01291-x 2026

  35. [37]

    MRXCAT: Realistic numerical phantoms for cardiovascular magnetic resonance

    Wissmann L, Santelli C, Segars WP, Kozerke S. MRXCAT: Realistic numerical phantoms for cardiovascular magnetic resonance. J Cardiovasc Magn Reson. 2014;16(1):63. doi:10.1186/s12968-014-0063-3

    work page doi:10.1186/s12968-014-0063-3 2014

  36. [38]

    Characterization of a deformable beating cardiac phantom with real-time dosimetric capabilities for validation of MRI-guided heart radiotherapy

    Aubert MMN, Uijtewaal P, Penev KI, et al. Characterization of a deformable beating cardiac phantom with real-time dosimetric capabilities for validation of MRI-guided heart radiotherapy. Med Phys. 2026;53(2):e70313. doi:10.1002/mp.70313

    work page doi:10.1002/mp.70313 2026

  37. [39]

    Deformable medical image registration: a survey

    Sotiras A, Davatzikos C, Paragios N. Deformable medical image registration: a survey. IEEE Trans Med Imaging. 2013;32(7):1153-1190. doi:10.1109/TMI.2013.2265603 24

    arXiv 2013

  38. [41]

    Spiral imaging: A critical appraisal

    Block KT, Frahm J. Spiral imaging: A critical appraisal. J Magn Reson Imaging. 2005;21(6):657-668. doi:10.1002/jmri.20320

    doi:10.1002/jmri.20320 2005

  39. [42]

    Cardiorespiratory-resolved magnetic resonance imaging: measuring respiratory modulation of cardiac function

    Thompson RB, McVeigh ER. Cardiorespiratory-resolved magnetic resonance imaging: measuring respiratory modulation of cardiac function. Magn Reson Med. 2006;56(6):1301-1310. doi:10.1002/mrm.21075

    work page doi:10.1002/mrm.21075 2006

  40. [43]

    Estimation of slipping organ motion by registration with direction-dependent regularization

    Schmidt-Richberg A, Werner R, Handels H, Ehrhardt J. Estimation of slipping organ motion by registration with direction-dependent regularization. Med Image Anal. 2012;16(1):150-159. doi:10.1016/j.media.2011.06.007

    work page doi:10.1016/j.media.2011.06.007 2012

  41. [46]

    Evaluation of the impact of cardiac implantable electronic devices on cine MRI for real-time adaptive cardiac radioablation on a 1.5 T MR-linac

    Akdag O, Mandija S, Borman PTS, et al. Evaluation of the impact of cardiac implantable electronic devices on cine MRI for real-time adaptive cardiac radioablation on a 1.5 T MR-linac. Med Phys. 2025;52(1):99-112. doi:10.1002/mp.17438

    work page doi:10.1002/mp.17438 2025

  42. [47]

    Principles and applications of balanced SSFP techniques

    Scheffler K, Lehnhardt S. Principles and applications of balanced SSFP techniques. Eur Radiol. 2003;13(11):2409-2418. doi:10.1007/s00330-003-1957-x

    doi:10.1007/s00330-003-1957-x 2003

  43. [101]

    2010;37(8):4078-4101

    Med Phys. 2010;37(8):4078-4101. doi:10.1118/1.3438081

    work page doi:10.1118/1.3438081 2010

  44. [2014]

    doi:10.48550/ARXIV.1412.6980

    Pith/arXiv arXiv doi:10.48550/arxiv.1412.6980

  45. [2015]

    doi:10.5281/ZENODO.31907

    work page doi:10.5281/zenodo.31907

  46. [2023]

    Lecture Notes in Computer Science

    Vol 14229. Lecture Notes in Computer Science. Springer Nature Switzerland; 2023:419-427. doi:10.1007/978-3-031-43999-5_40

    work page doi:10.1007/978-3-031-43999-5_40 2023

  47. [2025]

    doi:10.48550/ARXIV.2503.08373

    Pith/arXiv arXiv doi:10.48550/arxiv.2503.08373

  48. [2026]

    doi:10.48550/ARXIV.2603.04233

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2603.04233