“Two-for-one”: 3D cardiac and pulmonary MR imaging from a single acquisition using bSTAR

Background: Morphological and functional assessment of cardiac and pulmonary systems currently requires separate exams. Due to the highly coupled nature of these organs, a combined evaluation is desirable. Recent advances in MRI facilitate 4D imaging of the whole heart using the Free-running framework¹ or for the whole lung using ultra-short TE approaches such as bSTAR². However, long image reconstruction times (~several hours) limit clinical adoption. We propose a single acquisition with 3D radial free-breathing bSFFP dual echo acquisition (bSTAR) paired with two image reconstructions in one pipeline, one focused on the heart and the other on the lungs, deployed inline on the scanner with low-latency.

Methods: Seven healthy volunteers and four patients with cardiac or lung diseases were imaged on a 0.55T MRI scanner (MAGNETOM Free.Max, Siemens Healthcare, Germany). We used the bSTAR sequence implemented in Pulseq³ (TE1/TE2/TR=0.17/2.07/2.32ms, FA=40°, 1.2mm² resolution, FOV=450mm², scan time=9 min 40s) and a stack of 2D Cartesian bSFFP cine sequence breath held, 4min 30s, slice thickness = 8 mm, slice gap= 2 mm, resolution=1.5 mm² / 30ms, FOV=360x270 mm²).

Fig. 1 shows the dual-image reconstruction pipeline in Gadgetron⁴, during the scan, raw data is pre-processed and coil sensitivity map are estimated. After the scan, cardiac and pulmonary image reconstructions are performed in parallel. For the heart, stable-binned cardiac-resolved images were reconstructed at 1.4 mm³ with 30 ms temporal resolution using an iterative reconstruction with spatial and temporal total variation constraints. For the lungs, respiratory resolved images were reconstructed at 1.2 mm³ in 8 phases using Motion Compensated Low Rank algorithm (MoCoLoR)⁵.

Left ventricular end-diastolic, end-systolic volumes and ejection fraction (EF) were measured and compared to 2D cine images using a paired Student’s t-test. We tested lung image sharpness by fitting a sigmoid to a line profile between the liver-lung interface and compared gridded reconstruction with MoCoLoR.

Results: Cardiac and pulmonary images (Fig. 2) were reconstructed inline in 2 min 2s and 8 min 6s, respectively. The dual image reconstruction enables the 3D assessment of the left ventricle function and pulmonary morphology (Fig. 3). Volumetric and EF measurements showed no significant difference between bSTAR and 2D cines with a mean bias of 3.0 mL for EDV, 2.2mL for ESV and -0.2% for EF. The MoCoLoR reconstruction produced a significantly (p=5.0e-4) sharper liver-lung interface (slope s=4.01 cm^-1) compared to the gridding image reconstruction (s=1.80 cm^-1).

Conclusion: We demonstrated a new method to assess both cardiac and lung structure and function in a single scan. Whole heart bSTAR imaging provided results comparable to standard 2D volumetry. Pulmonary bSTAR imaging gave high quality image and may enable evaluation of the cardio-pulmonary interplay. The inline low latency reconstruction ~8 minutes for both 4D images, can facilitate clinical deployment.

Figure 1: Cardio-pulmonary image reconstruction pipeline: during the scan, raw data are pre-processed and coil sensitivity map is estimated. At the end of the acquisition, two image reconstructions run in parallel: Cardiac motion resolved and respiratory motion-compensated image reconstruction using TV regularization and respiratory resolved images using Motion compensated Low Rank.
Figure 2: Cardiac and respiratory resolved images in a healthy volunteer, lymphangioleiomyomatosis patient and a patient with cardiomyopathy. Red line profiles show the temporal resolution of cardiac images (30ms) or pulmonary images (N= 8 bins). Orange arrows show the pulmonary (nodules) or cardiac (aortic regurgitation) defects.
Figure 3: Reformatted bSTAR in 4-chamber and short axis view along with 3D left ventricle (LV) segmentation and 3D lung maximum intensity projection (MIP) in a health volunteer, patient with cardiomyopathy and a patient with pulmonary diseases.