Coronary Mask Guided Registration for Continuous Time 4D Cardiac CT Dataset Construction
Authors: Yuang Wang, Shuo Wang, Changyu Chen, Dufan Wu, Pengfei Jin, Yunqiang An, Yang Gao, Bin Lu, Dongrui Dai, Muge Du, Yan Yan, Dong Li, Liang Li, Li Zhang, Zhiqiang Chen
Organizations: Department of Engineering Physics, Tsinghua University, Beijing 100084, China · Department of Radiology, The Ohio State University, 2050 Kenny Rd, Columbus, OH 43221, Ohio, USA · Center for Advanced Medical Computing and Analysis, Massachusetts General Hospital and Harvard Medical School, 399 Revolution Dr, Somerville, 02145, Massachusetts, USA · Department of Radiology, Fuwai Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China · Department of Radiology, Tianjin Medical University General Hospital, Tianjin, China
Objective: Clinical cardiac CT multiphase reconstructions generally provide acceptable image quality in end-diastole (ED) or end-systole (ES) phases, but in other phases may exhibit motion artifacts, especially in the right coronary artery (RCA). This limits ground-truth availability in 4D cardiac CT imaging research. We aim to construct a 4D cardiac CT dataset that is generally suitable to serve as pseudo ground truth. Methods: We propose Coronary Mask Guided Registration (CMGR) to produce a motion-preserved, artifact-reduced, and continuous-time 4D cardiac CT sequence from the clinical multiphase reconstruction of each patient. For artifact reduction, CMGR uses the ED or ES phase as the reference phase and warps the reference volume with deformation fields to produce the sequence. For motion preservation, CMGR registers the reference phase to each non-reference phase of the multiphase reconstruction. To capture the motion of both the RCA and other cardiac structures in each registration, CMGR regularizes RCA masks and incorporates them into image-domain registration. Time-continuity is achieved by interpolating the deformation fields for non-reference phases to arbitrary times. Results: CMGR outperformed representative image-domain registration methods in capturing RCA motion and providing reasonable RCA shape, and showed competitive performance in capturing whole-heart motion. Additionally, CMGR reduced motion artifacts from clinical multiphase reconstructions, and intermediate CMGR frames generally provided plausible transitions between discrete cardiac phases. Conclusion: CMGR provides an effective approach for constructing continuous-time 4D cardiac CT datasets. Significance: The dataset can be used in system design simulations and in reconstruction algorithm development, thereby facilitating advances in cardiac CT imaging.
Figures & tables
Fig. 1: CMGR framework. CMGR is designed to produce a motion-preserved, artifact-reduced, and continuous-time 4D cardiac CT sequence from a clinical multiphase reconstruction. CMGR uses the cardiac phase with minimal motion artifacts as the reference phase and registers the reference phase to each non-reference phase. The two yellow boxes labeled “Multi-Resolution Registration” in Stages 1 and 2 represent the same registration strategy applied to different inputs. Specifically, the non-reference RCA mask in Stage 2 is the RCA mask produced after geometry propagation.
Original Cardiac Phases
10 × Denser Cardiac Phases
RCA
Entire Volume
RCA
Entire Volume
Method
Dice ↑
HD ↓
MSD ↓
RMSE ↓
SSIM ↑
LPIPS ↓
Dice ↑
HD ↓
MSD ↓
RMSE ↓
SSIM ↑
LPIPS ↓
FFD
0.48 ± 0.33
9.30 ± 7.64
2.99 ± 2.84
36.3 ±4.1
0.91 ± 0.01
0.20 ± 0.03
0.48 ± 0.33
9.17 ± 7.40
2.95 ± 2.76
35.9 ± 4.2
0.91 ± 0.01
0.19 ± 0.02
DARTEL
0.55 ± 0.31
8.65 ± 8.46
2.48 ± 2.78
35.4 ± 3.3
0.91 ± 0.01
0.20 ± 0.03
0.56 ± 0.31
8.59 ± 8.19
2.43 ± 2.68
34.9 ± 3.3
0.91 ± 0.01
0.19 ± 0.02
UGICON
0.75 ± 0.15
3.39 ± 3.41
0.54 ± 0.66
34.9 ± 2.3
0.91 ±0.01
0.18 ± 0.02
0.78 ± 0.14
3.30 ± 3.20
0.48 ± 0.60
34.1 ± 2.1
0.91 ± 0.01
0.18 ± 0.02
Ours
0.78 ± 0.11
1.91 ± 0.86
0.34 ± 0.16
34.4 ± 2.1
0.91 ± 0.01
0.18 ± 0.01
0.82 ± 0.08
1.77 ± 0.76
0.28 ± 0.12
33.0 ± 1.6
0.91 ± 0.01
0.18 ± 0.01
TABLE I: Quantitative results in XCAT simulations. HD and MSD are in mm, and RMSE in HU. STD was computed across cardiac phases for each case and then averaged across cases. Bold : best; underlined : second best.
Fig. 2: Visualization results for XCAT simulations. The cardiac phases close to the reference phase (45%) were omitted from the visualization. The regions enclosed by blue, green and yellow boxes are magnified for visual clarity. The box positions may vary across cardiac phases and are fixed across the original multiphase reconstruction, all tested methods, and the ground truth. The display window is [-250, 450] HU.
Fig. 3: Visualization results for CMGR intermediate frames in XCAT simulations. The regions enclosed by blue and yellow boxes are magnified for visual clarity. The box positions are fixed for each view. The display window is [-250, 450] HU.
RCA
Entire Volume
Method
Dice ↑
HD ↓
MSD ↓
RMSE ↓
SSIM ↑
LPIPS ↓
FFD
0.30 ± 0.23
14.3 ± 4.1
3.44 ± 2.41
50.2 ±4.4
0.93 ± 0.02
0.17 ± 0.04
DARTEL
0.45 ± 0.28
13.4 ± 4.7
2.81 ± 2.55
49.7 ± 4.3
0.93 ± 0.02
0.17 ± 0.04
UGICON
0.73 ± 0.24
13.5 ± 4.1
1.99 ± 2.12
43.6 ± 5.8
0.94 ± 0.02
0.17 ± 0.04
Ours
0.83 ± 0.04
8.6 ± 3.4
0.33 ± 0.09
46.4 ± 4.6
0.93 ± 0.02
0.17 ± 0.04
TABLE II: Quantitative results on clinical high-quality ED and ES pairs. HD and MSD are in mm, and RMSE in HU. STD was computed across pairs. Bold : best; underlined : second best.
Original Cardiac Phases
10 × Denser Cardiac Phases
Method
FOR ↑
NC ↑
FOR ↑
NC ↑
FFD
0.74 ± 0.05
0.66 ± 0.09
0.75 ± 0.04
0.66 ± 0.08
DARTEL
0.75 ± 0.04
0.65 ± 0.07
0.76 ± 0.03
0.66 ± 0.07
UGICON
0.71 ± 0.10
0.61 ± 0.15
0.73 ± 0.08
0.63 ± 0.13
Ours
0.77 ± 0.02
0.70 ± 0.04
0.77 ± 0.02
0.69 ± 0.05
TABLE III: FOR and NC in clinical data evaluation. STD was computed across cardiac phases for each case and then averaged across cases. Bold : best; underlined : second best.
Fig. 4: Visualization results for the first representative clinical case. The cardiac phases close to the reference phase (75%) were omitted from the visualization. The regions enclosed by blue and yellow boxes are magnified for visual clarity. The box positions may vary across cardiac phases and are fixed across the original multiphase reconstruction and all tested methods. The display window is [-250, 450] HU.
Fig. 5: Visualization results for the second representative clinical case. The cardiac phases close to the reference phase (80%) were omitted from the visualization. The regions enclosed by blue and yellow boxes are magnified for visual clarity. The box positions may vary across cardiac phases and are fixed across the original multiphase reconstruction and all tested methods. The display window is [-250, 450] HU.
Fig. 6: Visualization results for CMGR intermediate frames in clinical data evaluation. The regions enclosed by blue and yellow boxes are magnified for visual clarity. The box positions are fixed for each view. The display window is [-250, 450] HU.
Fig. 7: Visualization results for non-RCA cardiac structures in clinical data evaluation. The regions enclosed by blue and green boxes are magnified for visual clarity. The box positions may vary across cardiac phases and are fixed across the original multiphase reconstruction and CMGR. The display window is [-250, 450] HU.
Rank Score ↑
Radiologists
Researchers
RCA
Chamber & Aorta
RCA
Chamber & Aorta
Method
Motion
Geometry
Motion
Structure
Motion
Geometry
Motion
Structure
DARTEL
2.00 ± 0.45
1.98 ± 0.42
2.08 ± 0.63
2.08 ± 0.56
1.75 ± 0.46
1.99 ± 0.38
1.69 ± 0.65
2.17 ± 0.64
UGICON
1.24 ± 0.59
1.14 ± 0.40
1.38 ± 0.60
1.40 ± 0.69
1.35 ± 0.60
1.09 ± 0.29
1.64 ± 0.79
1.21 ± 0.55
Ours
2.76 ± 0.55
2.88 ± 0.38
2.54 ± 0.75
2.52 ± 0.75
2.91 ± 0.33
2.92 ± 0.32
2.67 ± 0.52
2.61 ± 0.51
TABLE IV: Rank scores in the observer study. STD was computed across cases and observers. Bold : best; underlined : second best.
P-value
RCA
Chamber & Aorta
Statistical Test
Motion
Geometry
Motion
Structure
Friedman
<0.001
<0.001
<0.001
<0.001
Wilcoxon(Ours > DARTEL)
<0.001
<0.001
0.019
0.011
Wilcoxon(Ours > UGICON)
<0.001
<0.001
<0.001
<0.001
TABLE V: P-values in statistical tests of the observer study. Observer–case rank scores were used. Statistical tests were performed separately for radiologists and researchers, and larger p-values were reported.
Accurate 4D whole-heart mesh reconstruction from sparse cine MRI is critical for creating cardiac digital twins, but remains challenging due to limited 2D slice coverage and the complex coupling between cardiac shape and motion. Existing methods often rely on intermediate contour fitting and typically reconstruct static, single-phase, or partial cardiac geometries, limiting their ability to capture full-chamber dynamics. We propose a novel end-to-end framework for reconstructing temporally resolved whole-heart meshes from multi-view 2D cine MRI sequences by learning an image-to-mesh mapping. The framework incorporates a differentiable contour renderer inspired by the Beer-Lambert attenuation principle, enabling anatomy-aware supervision of 3D+t mesh deformation through contour-based projection losses. To improve temporal consistency across the cardiac cycle, we further introduce a multi-scale temporal modeling module that integrates global cycle-level dynamics with local inter-frame coherence to generate smooth and physiologically plausible mesh trajectories. The proposed method achieved a whole-heart mean absolute error of 1.68 ± 0.31 mm and a motion jitter of 0.77 ± 0.17 mm/frame3, outperforming existing methods with lower reconstruction error and substantially improved motion smoothness. It also improved 2D contour alignment across multiple cine MRI views and supported downstream proof-of-concept electrophysiological simulation. The code will be released publicly upon acceptance of the manuscript for publication.
Xiaoyue Liu, Dongcheng Cang, Xiaohan Yuan +3
Department of Biomedical Engineering, National University of Singapore, Singapore · School of Automation, Southeast University, Nanjing, China · Department of Medicine, National University of Singapore, Singapore +1
Accurate 3D+t whole-heart mesh reconstruction from cine MRI is a clinically crucial yet technically challenging task. The difficulty of this task arises from two coupled factors: inherently sparse sampling of 3D cardiac anatomy by 2D image slices and the tight coupling between cardiac shape and motion. Current cardiac image-to-mesh approaches typically reconstruct only a subset of cardiac chambers or a single phase of the cardiac cycle. In this work, we propose CineMesh4D, a novel end-to-end 4D (3D+t) pipeline that directly reconstructs patient-specific whole-heart mesh from multi-view 2D cine MRI via cross-domain mapping. Specifically, we introduce a differentiable rendering loss that enables supervision of 3D+t whole-heart mesh from multi-view sparse contours of cine MRI. Furthermore, we develop a dual-context temporal block that fuses global and local cardiac temporal information to capture high-dimensional sequential patterns. In quantitative and qualitative evaluations, CineMesh4D outperforms existing approaches in terms of reconstruction quality and motion consistency, providing a practical pathway for personalized real-time cardiac assessment. The code will be publicly released once the manuscript is accepted.
Xiaoyue Liu, Xiaohan Yuan, Mark Y Chan +2
Department of Biomedical Engineering, National University of Singapore, Singapore · School of Automation, Southeast University, Nanjing, China · Department of Medicine, National University of Singapore, Singapore +1
Reconstructing cardiac motion from CMR sequences is critical for diagnosis, prognosis, and intervention. Existing methods rely on complete CMR stacks to infer full heart motion, limiting their applicability during intervention when only sparse observations are available. We present TetHeart, the first end-to-end framework for unified 4D heart mesh recovery from both offline full-stack and intra-procedural sparse-slice observations. Our method leverages deformable tetrahedra to capture shape and motion in a coherent space shared across cardiac structures. Before a procedure, it initializes detailed, patient-specific heart meshes from high-quality full stacks, which can then be updated using whatever slices can be obtained in real-time, down to a single one during the procedure. TetHeart incorporates several key innovations: (i) an attentive slice-adaptive 2D-3D feature assembly mechanism that integrates information from arbitrary numbers of slices at any position; (ii) a distillation strategy to ensure accurate reconstruction under extreme sparsity; and (iii) a weakly supervised motion learning scheme requiring annotations only at keyframes, such as the end-diastolic and end-systolic phases. Trained and validated on three large public datasets and evaluated zero-shot on additional private interventional and public datasets without retraining, TetHeart achieves state-of-the-art accuracy and strong generalization in both pre- and intra-procedural settings. Code and dataset is available at https://github.com/Scalsol/TetHeart.
Yihong Chen, Jiancheng Yang, Deniz Sayin Mercadier +3
School of Computer and Communication Science, EPFL · ELLIS Institute Finland, Aalto University, Finland · University of North Carolina at Charlotte +1