Abstract
Respiratory motion models support thoracic radiotherapy and deep-learning methods for registration, reconstruction, and motion compensation, but training data depend on scarce four-dimensional computed tomography (4DCT) images and error-prone deformable image registrations (DIRs). We evaluated an anatomy-driven heuristic model of thoracic respiratory motion. At generation, it uses a static reference CT, organ structures, and a breathing preset without considering patient-specific 4DCT or DIR information. Its interpretable parameters were fitted by gradient descent to DIR-derived motion from 100 training patients. The fits yielded a compact breathing preset catalogue for future simulation. A population principal component analysis (PCA) model built on the same cohort provided an oracle representational-capacity benchmark. On 32 held-out patients the PCA reconstruction achieved a whole-lung mean end-point error (EPE) of 2.30 mm. The heuristic with per-patient differentiable fitting approached this accuracy on the training cohort (EPE 2.79 mm). On the same 32 held-out patients the default configuration scored 3.51 mm, while the registration-free population preset reached 3.21 mm and the best-fitting breathing-type preset 3.06 mm. Preset fields folded less than the registration references. Respiratory motion Synthetic data 4DCT Deformable registration PCA.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/RIME_013.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=ONdf59Rhme
BibTex
@InProceedings{ZimLuk_An_MICCAISAT2026,
author = { Zimmermann, Lukas AND Arnold, Julius AND Rauter, Michael AND Renner, Andreas AND Knäusl, Barbara},
title = { { An Anatomy-Driven Thoracic Respiratory Motion Model with Population-Calibrated Presets } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17277},
month = {pending},
page = {pending}
}