Fig 1.
Summary of the main steps involved in the image reconstruction.
Starting from a free-running acquisition (A), the reference SI projections are concatenated to obtain a matrix that is used as input to SIMBA (B). After applying SIMBA, the resulting binning consists of a set of disjoint clusters. As a first common step, the four most populated clusters are selected and reconstructed using a non-uniform 3D gridded reconstruction (C). The original SIMBA image consists of taking the non-uniform 3D gridded reconstruction of the most populated cluster. (D1) A motion-resolved reconstruction (XD-SIMBA) is obtained by performing a compressed-sensing reconstruction and regularization over the clustering dimension. (D2a) Additionally, we integrate a non-rigid motion estimation in the iterative reconstruction framework (D2b) to regularize over co-registered clusters and obtain a motion-resolved reconstruction with inter-cluster motion compensation (XD-MC-SIMBA).
Fig 2.
Example of the different images reconstructed from the four selected SIMBA clusters using a simple gridded reconstruction (SIMBA), a CS reconstruction for which the SIMBA clustering is a dimension of sparsity (XD-SIMBA), and the same CS reconstruction extended with inter-cluster motion compensation (XD-MC-SIMBA).
The data shown are from a patient with tetralogy of Fallot (M, 36 years). The indication for the ferumoxytol-enhanced scan was a transannular patch repair.
Fig 3.
Example (also shown in Fig 2) of a patient after repair of tetralogy of Fallot and zoomed-in view of the aortic valve.
The coaptation area of the aortic leaflet (arrow) is more clearly visible in XD-MC-SIMBA and FRF. In this example, XD-MC-SIMBA has a much higher assigned image quality score, even compared to FRF. When looking at the coronary reformat, the distal portions of the left anterior descending (LAD) and left circumflex (LCX) coronary arteries are only depicted in XD-MC-SIMBA and FRF, and the proximal right coronary artery (RCA) is also better delineated in XD-MC-SIMBA, even compared to FRF. Abbreviations: RCA, right coronary artery; LM, left main coronary artery; LAD, left anterior descending.
Fig 4.
Quantitative image analysis metrics for the 3D gridded reconstruction (SIMBA) in red, the motion-resolved reconstruction (XD-SIMBA) in purple, the motion-resolved reconstruction with inter-cluster motion compensation (XD-MC-SIMBA) in blue, and a resting phase from the 5D free-running reconstruction (FRF) in green.
Analysis of the LAD and RCA coronary arteries comprises the total visible vessel length, and the sharpness of the first proximal 4 cm. All results are shown using box plots, where the “x” indicates the mean values and the solid line the median values. Differences that are statistically significant are indicated by an asterisk (*) for p<0.0125. When looking at the relative percentage changes in image quality scores (IQS), overall XD-SIMBA results in a decrease in image quality score compared to SIMBA. Conversely, XD-MC-SIMBA improves the image quality compared to both SIMBA and XD-SIMBA. For FRF, there is also an improvement compared to both SIMBA and XD-SIMBA, but a slight worsening in score compared to XD-MC-SIMBA. Abbreviations: LM+LAD, left main+left anterior descending; RCA, right coronary artery.
Table 1.
Summary of all metrics values (mean standard deviation) for the different reconstructions, with the corresponding statistical analysis (p-values) using one-way analysis of variance (ANOVA).
Table 2.
Count of visible ostia, visible proximal and distal portions of the RCA and LAD for all analyzed cases.
Reported scores are from all 24 cases.
Fig 5.
23-year-old male patient post Fontan procedure, with a right atrial isomerism.
In the coronal view, the RCA is clearly visible (yellow arrow), and sharper for the XD-SIMBA and XD-MC-SIMBA reconstructions. Moreover, the papillary muscles (red arrow) are better delineated for XD-MC-SIMBA. This is observed also in the axial view (yellow arrow). When looking at the coronary reformats, the more distal segments of the RCA, LAD and LCX can be visualized in the XD-MC-SIMBA reconstruction (red arrows). Also for FRF there is a very good depiction of the cardiac anatomy, while only the coronary vessels seem less visible, compared to XD-MC-SIMBA. Abbreviations: RCA, right coronary artery; LAD, left anterior descending; LCX: left circumflex.
Fig 6.
2-year-old female patient with Kawasaki disease.
The yellow bar (first image on the left in the sagittal view) indicates the scale after the zoom around the heart (heart diameter <10cm). In the axial view, the right atrial wall is sharper for XD-SIMBA and XD-MC-SIMBA, compared to SIMBA (yellow arrow), and so are the papillary muscles in the left ventricle (red arrow). In the sagittal view, the yellow arrow indicates the LAD, while the green arrow indicates the left ventricular wall, which is sharper in FRF. In the coronal views, the papillary muscles (yellow arrow) are sharper compared in XD-MC-SIMBA, compared to both SIMBA and XD-SIMBA. For FRF, the image quality is good in terms of delineation of big anatomical features (e.g. papillary muscles, septal wall), but small features such as the coronary vessels are blurrier. Abbreviations: RCA, right coronary artery; LM, left main coronary artery; LAD, left anterior descending.