Modality-agnostic, patient-specific digital twins modeling temporally varying digestive motion

Ergys D. Subashi; Harini Veeraraghavan; James M. Balter; Jorge Tapias Gomez; Jue Jiang; Lando S. Bosma; Mert R Sabuncu; Neelam Tyagi; Nishant Nadkarni; William P. Segars

arxiv: 2507.01909 · v3 · pith:JF6LC5NRnew · submitted 2025-07-02 · 💻 cs.CV

Modality-agnostic, patient-specific digital twins modeling temporally varying digestive motion

Jorge Tapias Gomez , Nishant Nadkarni , Lando S. Bosma , Jue Jiang , Ergys D. Subashi , William P. Segars , James M. Balter , Mert R Sabuncu

show 2 more authors

Neelam Tyagi Harini Veeraraghavan

This is my paper

Pith reviewed 2026-05-19 05:38 UTC · model grok-4.3

classification 💻 cs.CV

keywords digital twinsdeformable image registrationgastrointestinal motionpatient-specific modeling4D sequencesdose accumulationMRICT

0 comments

The pith

Patient-specific digital twins synthesized from static 3D scans model realistic time-varying gastrointestinal motion to test deformable image registration accuracy.

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

The paper describes a pipeline that turns ordinary static 3D patient scans into patient-specific digital twins simulating digestive motion across the gastrointestinal tract. These twins are produced by feeding published analytical motion models into a semi-automated process that outputs twenty-one motion phases as 4D sequences. A sympathetic reader would care because validating registration tools for highly mobile organs has been limited by the scarcity of real 4D data and the difficulty of placing manual landmarks. If the pipeline performs as claimed, it supplies a practical ground-truth reference for measuring how well different registration methods handle complex motion and for quantifying errors when dose distributions are warped and summed. The work demonstrates the approach on MRI and CT datasets and reports motion statistics that fall within 0.8 mm of values taken from independent real-patient gastric motion studies.

Core claim

The authors built a semi-automated pipeline that generates 4D digital twins from static 3D scans by applying published analytical models of gastrointestinal motion. Across eleven datasets spanning T2-weighted MRI, T1-weighted 4D golden-angle stack-of-stars MRI, and contrast-enhanced CT, the resulting twins produce mean and maximum motion amplitudes and mean log Jacobian determinant values within 0.8 mm and 0.01 of published real-patient gastric motion data. These twins then serve as reference to evaluate six deformable image registration methods through target registration error, Dice similarity coefficient, and 95th-percentile Hausdorff distance, with additional dose-warping and dose-accum-

What carries the argument

A semi-automated pipeline that applies published analytical gastrointestinal motion models to static 3D patient scans to produce 4D motion sequences.

If this is right

The digital twins supply quantitative target registration error, Dice, and Hausdorff metrics for six registration methods without manual landmarks.
Voxel-level visualizations become available to show where each registration method succeeds or fails in mobile anatomy.
Dose distributions can be warped and accumulated to measure dosimetric errors in low- and high-dose regions on a patient-specific basis.
The same pipeline supports rigorous testing of registration tools in other dynamic, anatomically complex regions.

Where Pith is reading between the lines

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

The method could be extended by substituting different analytical motion models for other organs or disease sites.
Routine use might reduce reliance on scarce 4D acquisitions during algorithm development.
Patient-specific error maps could inform adaptive radiotherapy planning that accounts for motion variability.

Load-bearing premise

The published analytical models of gastrointestinal motion accurately capture the range and temporal patterns of real digestive movement in patients.

What would settle it

Direct comparison of the generated twins against additional independent 4D MRI scans from the same patients that reveals mean motion amplitude differences larger than 0.8 mm.

Figures

Figures reproduced from arXiv: 2507.01909 by Ergys D. Subashi, Harini Veeraraghavan, James M. Balter, Jorge Tapias Gomez, Jue Jiang, Lando S. Bosma, Mert R Sabuncu, Neelam Tyagi, Nishant Nadkarni, William P. Segars.

read the original abstract

Objective: Clinical implementation of deformable image registration (DIR) requires voxel-based spatial accuracy metrics such as manually identified landmarks, which are challenging to implement for highly mobile gastrointestinal (GI) organs. To address this, patient-specific digital twins (DT) modeling temporally varying motion were created to assess the accuracy of DIR methods. Approach: 21 motion phases simulating digestive GI motion as 4D sequences were generated from static 3D patient scans using published analytical GI motion models through a semi-automated pipeline. Eleven datasets, including six T2w FSE MRI (T2w MRI), two T1w 4D golden-angle stack-of-stars, and three contrast-enhanced CT scans. The motion amplitudes of the DTs were assessed against real patient stomach motion amplitudes extracted from independent 4D MRI datasets. The generated DTs were then used to assess six different DIR methods using target registration error, Dice similarity coefficient, and the 95th percentile Hausdorff distance using summary metrics and voxel-level granular visualizations. Finally, for a subset of T2w MRI scans from patients treated with MR-guided radiation therapy, dose distributions were warped and accumulated to assess dose warping errors, including evaluations of DIR performance in both low- and high-dose regions for patient-specific error estimation. Main results: Our proposed pipeline synthesized DTs modeling realistic GI motion, achieving mean and maximum motion amplitudes and a mean log Jacobian determinant within 0.8 mm and 0.01, respectively, similar to published real-patient gastric motion data. It also enables the extraction of detailed quantitative DIR performance metrics and rigorous validation of dose mapping accuracy. Significance: The pipeline enables rigorously testing DIR tools for dynamic, anatomically complex regions enabling granular spatial and dosimetric accuracies.

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.

Desk Editor's Note private letter to a colleague

A practical pipeline turns static scans into patient-specific GI motion twins for DIR testing, but realism rests on matching summary amplitudes and one Jacobian number to literature values.

read the letter

This paper describes a pipeline for creating patient-specific digital twins that simulate time-varying digestive motion in the gastrointestinal tract. The twins are built from ordinary static 3D scans using published analytical motion models, and the authors show how to use them to test deformable image registration and dose accumulation in areas where landmarks are scarce. They processed eleven datasets spanning T2-weighted MRI, golden-angle stack-of-stars MRI, and contrast CT. From each static scan they produced 21 motion phases. The resulting motion amplitudes and mean log Jacobian determinant stayed within 0.8 mm and 0.01 of values reported in independent real-patient gastric studies. They then ran six different DIR algorithms on the twins and reported target registration error, Dice scores, and Hausdorff distances, along with some dose-warping results for MR-guided cases. The practical side is the main strength. Turning static images into modality-agnostic 4D sequences in a semi-automated way gives a concrete resource for validation work that has been missing for highly mobile abdominal organs. The granular visualizations they provide should make it easier to diagnose where registration methods break down. One limitation is that the realism argument rests on matching summary statistics from the literature. The abstract notes that amplitudes came from independent 4D MRI, yet there is no mention of comparing the full spatial deformation patterns or organ-specific paths directly to those same datasets. If the analytical models get the overall scale right but produce incorrect directions or non-physiological deformations, the DIR performance numbers could look better than they would on real motion. It would also be useful to see how much the results change when the motion model parameters are varied within reasonable ranges. Readers working on DIR validation or radiation therapy planning for abdominal sites will find this directly relevant. The paper supplies a method and some baseline numbers that others can replicate or extend. I would recommend sending it to peer review. The topic addresses a genuine gap, and the work is grounded enough that referees can evaluate the validation approach in detail.

Referee Report

1 major / 0 minor

Summary. The manuscript presents a semi-automated pipeline that generates modality-agnostic, patient-specific digital twins (DTs) from static 3D scans (T2w MRI, T1w 4D golden-angle stack-of-stars, and contrast-enhanced CT) by applying published analytical GI motion models to produce 21-phase 4D sequences simulating temporally varying digestive motion. Motion realism is assessed by comparing mean and maximum amplitudes and mean log Jacobian determinant of the DTs to values extracted from independent real-patient 4D MRI datasets (agreement within 0.8 mm and 0.01). The resulting DTs are then used to evaluate six DIR methods via target registration error, Dice similarity coefficient, and 95th-percentile Hausdorff distance, and to quantify dose-warping errors (including low- and high-dose regions) for a subset of MR-guided radiotherapy patients.

Significance. If the DTs are shown to reproduce not only summary statistics but also spatially accurate deformation fields, the pipeline would provide a practical, patient-specific framework for rigorous DIR validation and dose-accumulation testing in highly mobile GI anatomy where manual landmarks are difficult to obtain. The use of independent 4D MRI datasets for amplitude extraction and published analytical models for motion synthesis are positive elements that support reproducibility.

major comments (1)

[Approach / Main results] Approach section (and Main results): validation of 'realistic GI motion' is performed exclusively via aggregate statistics (mean/max motion amplitudes and mean log Jacobian determinant) matched to published literature values. No direct voxel-wise or field-level comparison of the generated deformation vector fields, organ-specific trajectories, or phase-to-phase consistency against the independent 4D MRI datasets is reported. Because the central claim is that the DTs enable reliable DIR and dose-warping assessment, this indirect validation leaves open the possibility that plausible summary metrics coexist with non-physiological spatial patterns.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive review and for recognizing the potential of our pipeline for DIR and dose-warping validation in mobile GI anatomy. We address the major comment below and have revised the manuscript to clarify the validation strategy and its limitations.

read point-by-point responses

Referee: [Approach / Main results] Approach section (and Main results): validation of 'realistic GI motion' is performed exclusively via aggregate statistics (mean/max motion amplitudes and mean log Jacobian determinant) matched to published literature values. No direct voxel-wise or field-level comparison of the generated deformation vector fields, organ-specific trajectories, or phase-to-phase consistency against the independent 4D MRI datasets is reported. Because the central claim is that the DTs enable reliable DIR and dose-warping assessment, this indirect validation leaves open the possibility that plausible summary metrics coexist with non-physiological spatial patterns.

Authors: We thank the referee for this observation. The independent 4D MRI datasets were used only to extract summary amplitude statistics because they come from separate patient cohorts and lack corresponding static 3D scans from the same individuals, precluding direct voxel-wise or field-level matching. The analytical GI motion models are taken from published studies that characterize physiological motion; matching mean and maximum amplitudes (within 0.8 mm) together with the mean log Jacobian determinant (within 0.01) therefore provides population-level consistency with real data. The primary purpose of the DTs is to supply known ground-truth deformation fields for DIR and dose-warping evaluation, which is possible by construction once the models are applied. In the revised manuscript we have added organ-specific trajectory examples, qualitative DVF visualizations, and an expanded Discussion section that explicitly acknowledges the indirect nature of aggregate validation while noting its practicality and reproducibility. These changes address the concern without overstating the spatial fidelity of the current DTs. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation relies on external published models and independent validation data

full rationale

The paper constructs digital twins by applying published analytical GI motion models to static 3D patient scans through a semi-automated pipeline, then evaluates the resulting motion amplitudes and mean log Jacobian determinant against values extracted from independent 4D MRI datasets and published real-patient gastric motion data. No load-bearing step reduces by construction to a quantity defined or fitted within the present work; the models and comparison benchmarks originate outside the paper and are not redefined or tuned to force the reported similarity metrics.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim depends on the fidelity of external analytical models and the assumption that synthesized motion matches real physiology; limited information is available from the abstract alone.

free parameters (1)

GI motion model parameters
The pipeline relies on published analytical models whose internal parameters are not detailed here and may have been fitted to prior data.

axioms (1)

domain assumption Published analytical GI motion models accurately represent real patient digestive motion patterns
The DTs are generated directly from these models applied to static scans.

invented entities (1)

Patient-specific digital twins for GI motion no independent evidence
purpose: Provide synthetic 4D sequences with known ground-truth motion for DIR validation
These are constructed models rather than directly measured 4D data.

pith-pipeline@v0.9.0 · 5895 in / 1379 out tokens · 77415 ms · 2026-05-19T05:38:11.091541+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

[1]

inner shell

Introduction Deformable image registration (DIR) is a critical component of both computed tomography (CT) and magnetic resonance (MR) image-guided adaptive radiotherapy [1-3]. It enables automated propagation of organ-at-risk (OAR) segmentations across treatment fractions and facilitates accurate estimation of the radiation dose delivered to both OARs and...

work page doi:10.1016/j.ijrobp.2025.04.004 2025
[2]

A realistic spline-based dynamic heart phantom,

Bosma LS, Zachiu C, Ries M, Denis de Senneville B, Raaymakers BW. Quantitative investigation of dose accumulation errors from intra-fraction motion in MRgRT for prostate cancer. Phys Med Biol. 2021 Mar 2;66(6):065002. doi: 10.1088/1361-6560/abe02a. PMID: 33498036. [12] Subashi E, Segars P, Veeraraghavan H, Deasy J, Tyagi N. A model for gastrointestinal tr...

work page doi:10.1088/1361-6560/abe02a 2021
[3]

VoxelMorph: A Learning Framework for Deformable Medical Image Registration,

G. Balakrishnan, A. Zhao, M. R. Sabuncu, J. Guttag and A. V. Dalca, "VoxelMorph: A Learning Framework for Deformable Medical Image Registration," in IEEE Transactions on Medical Imaging, vol. 38, no. 8, pp. 1788-1800, Aug. 2019, doi: 10.1109/TMI.2019.2897538. [37] Jiang J, Hong J, Tringale K, et al. Progressively refined deep joint registration segmentati...

work page doi:10.1109/tmi.2019.2897538 2019

[1] [1]

inner shell

Introduction Deformable image registration (DIR) is a critical component of both computed tomography (CT) and magnetic resonance (MR) image-guided adaptive radiotherapy [1-3]. It enables automated propagation of organ-at-risk (OAR) segmentations across treatment fractions and facilitates accurate estimation of the radiation dose delivered to both OARs and...

work page doi:10.1016/j.ijrobp.2025.04.004 2025

[2] [2]

A realistic spline-based dynamic heart phantom,

Bosma LS, Zachiu C, Ries M, Denis de Senneville B, Raaymakers BW. Quantitative investigation of dose accumulation errors from intra-fraction motion in MRgRT for prostate cancer. Phys Med Biol. 2021 Mar 2;66(6):065002. doi: 10.1088/1361-6560/abe02a. PMID: 33498036. [12] Subashi E, Segars P, Veeraraghavan H, Deasy J, Tyagi N. A model for gastrointestinal tr...

work page doi:10.1088/1361-6560/abe02a 2021

[3] [3]

VoxelMorph: A Learning Framework for Deformable Medical Image Registration,

G. Balakrishnan, A. Zhao, M. R. Sabuncu, J. Guttag and A. V. Dalca, "VoxelMorph: A Learning Framework for Deformable Medical Image Registration," in IEEE Transactions on Medical Imaging, vol. 38, no. 8, pp. 1788-1800, Aug. 2019, doi: 10.1109/TMI.2019.2897538. [37] Jiang J, Hong J, Tringale K, et al. Progressively refined deep joint registration segmentati...

work page doi:10.1109/tmi.2019.2897538 2019