Free-breathing isotropic 3D whole-heart imaging vs 2D cartesian imaging: A comparison of left ventricular volumetry

Presenting Author(s)

• 

  Jordan C. Taylor

  MRI Technologist
  NIH
  Laurel, Maryland, United States

Primary Author(s)

• 

  Jordan C. Taylor

  MRI Technologist
  NIH
  Laurel, Maryland, United States

Co-Author(s)

• AT

  Anastasia Tsakirellis

  Nurse Practitioner
  National Institutes of Health, United States
• RR

  Rajiv Ramasawmy, PhD

  Staff Scientist
  National Heart, Lung, and Blood Institute, National Institutes of Health
  Bethesda, Maryland, United States
• CM

  Christine Mancini, RT

  CV MRI Technologist
  National Institutes of Health, United States
• 

  Ahsan Javed, PhD

  Staff Scientist
  National Heart, Lung, and Blood Institute, National Institutes of Health
  Vienna, Virginia, United States
• 

  Adrienne E E. Campbell-Washburn, PhD

  Principal Investigator
  National Heart, Lung, and Blood Institute, National Institutes of Health
  Bethesda, Maryland, United States
• 

  Pierre Daudé, PhD

  Postdoctoral fellow
  National Heart, Lung, and Blood Institute, National Institutes of Health
  Bethesda, Maryland, United States

Background: CMR characterization of left ventricular (LV) function is typically accomplished by using a stack of short-axis (SAX) 2D cines through the LV, requiring multiple breath-holds that are difficult for sick patients [1]. Recent isotropic 3D whole-heart approaches [2] may overcome these challenges because they are free-breathing and allow greater freedom in scan plane selection post-image acquisition. To apply 3D whole-heart imaging in patients, it is important that the LV function measurements remain accurate. Therefore, we propose to validate the accuracy of LV function acquired using the isotropic 3D bSTAR (bSSFP dual echo sequence)[3] compared to reference 2D cartesian bSSFP cine.

Methods: Ten subjects (7 healthy volunteers, 3 patients) were scanned on a 0.55T MAGNETOM FreeMax system (Siemens Healthcare, Erlangen, Germany). Two patients presented with a dilated LV and reduced EF. One patient had PVCs. A standard stack of SAX 2D cartesian bSFFP cines were acquired (slice thickness=8mm, slice gap=2mm, acquisition time=4min 30s, spatial resolution=1.5 mm², temporal resolution= 28.8ms, FOV=360x270 mm²) along with the bSTAR 3D radial research sequence (acquisition time=9min 40s, spatial resolution =1.4 mm³, temporal resolution=30ms, FOV=450 mm³, compressed sensing reconstruction). Images were reformatted using 3D Slicer into a stack of SAX cines view (Figure 1). The images were processed in two ways: 1) The 3D SAX images were resliced to match the cartesian 2D with 8mm slice thickness and 2mm gap, and 2) the 3D SAX images maintained original resolution (1.4mm slice, zero gap). Reslicing and left ventricle segmentation was performed using SuiteHeart (NeoSoft LLC, Pewaukee, WI) using the automatic segmentation tool and manual correction. End-diastolic volume (EDV), end-systolic volume (ESV), and ejection fraction (EF) were analyzed and compared to the 2D cine images using Bland-Altman plots and paired t-tests

Results: Strong agreement was observed between 2D cine and 3D bSTAR (Figure 2, Table 1). There were no statistically significant differences between resliced bSTAR and 2D cines (EF bias -1.0%, p=0.56; EDV bias 5.5 mL, p = 0.05; ESV bias 4.76 mL, p=0.08). Additionally, there were no statistically significant differences between the original high-resolution bSTAR and 2D cines (EF bias -0.1%, p=1.0; EDV bias 2.8mL, p=0.49; ESV bias 2.09mL, p-value 0.56).

Conclusion: Previous studies[4] have found statistically significant bias in EF between 3D and 2D imaging. Here, the 3D bSTAR sequence proved to be accurate when compared to 2D cine, with no statistically significant difference using either matched resliced images (8mm slice, 2mm gap) or high-resolution (1.4mm slice, 0mm gap). We suspect our higher temporal resolution (30ms) has a strong influence on the concordance between cine techniques. Therefore, bSTAR may enable quantitative, free-breathing isotropic 3D cine imaging without the requirement of advanced scan planning.

Table 1. Comparison of volumes and ejection fraction between standard 2D cine and 3D bSTAR.

	EDV(ml)	ESV(ml)	EF(%)	Compared to 2D cine
2D cine	182.7+/-38.3	88.3+/-37.1	52.8+/-10.5	EDV bias	EDV LOA	ESV bias	ESV LOA	EF bias	EF LOA
bSTAR Resliced	188.2+/-38.4	93.0+/-36.8	51.8+/-9.9	5.5	-9.86 to 20.86	4.76	-10.77 to 20.29	-1	-9.16 to 7.16
bSTAR Original	185.5+/-40.7	90.4+/-41.0	52.7+/-10.1	2.8	-21.22 to 26.82	2.09	-14.56 to 18.74	-0.1	-5.75 to 5.55

Figure 1. (a.) Images of the standard 2D cine vs the 3D BSTAR in 4ch, 2ch, and SAX planes. (b.) Images depicting the difference between bSTAR resliced and bSTAR original.
Figure 2. Bland-Altman analyses of resliced bSTAR and original bSTAR vs 2D for LVEF, LVESV, and LVEDV. The solid line corresponds to the mean bias and the dotted line corresponds to the 95% upper and lower limits of agreement (LOA). Patients (PT) represented in red. Healthy Volunteers (HV) represented in blue