Reproducibility of left ventricular volumetric parameters derived from fully automatically prescribed image planes in CMR: a direct comparison to manual planning at 1.5T and 3T

Presenting Author(s)

• 

  Benjamin Böttcher, MD

  Resident / Physician
  Institute for Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock
  Rostock, Mecklenburg-Vorpommern, Germany

Primary Author(s)

• 

  Benjamin Böttcher, MD

  Resident / Physician
  Institute for Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock
  Rostock, Mecklenburg-Vorpommern, Germany

Co-Author(s)

• KD

  Karolin Deyerberg, MD

  Doctoral candidate / Student
  Institute of Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock
  Rostock, Mecklenburg-Vorpommern, Germany
• LW

  Lena-Maria Watzke, MD

  Doctoral candidate / Student
  Institute of Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock
  Rostock, Mecklenburg-Vorpommern, Germany
• AK

  Ann-Christin Klemenz, PhD

  Postdoctoral Researcher
  Institute of Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock, Germany
• MM

  Mathias Manzke

  Scientist
  Institute of Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock, Germany
• MG

  Margarita Gorodezky, PhD

  Clinical Scientist
  GE HealthCare
  Munich, Bayern, Germany
• GD

  Gaspar Delso, PhD

  Senior scientist
  GE Healthcare
  Barcelona, Catalonia, Spain
• AD

  Antonia Dalmer, MD

  Physician
  Institute of Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock, Germany
• RL

  Roberto Lorbeer, PhD

  Scientist
  Ludwig-Maximilians-University of Munich, Germany
• MW

  Marc-André Weber, MD, MSc

  Institute Director
  Institute of Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock, Germany
• FM

  Felix G. Meinel, MD

  Vize Chair
  Institute of Diagnostic and Interventional Radiology, Pediatric Radiology and Neuroradiology / University Medical Center Rostock, Germany

Background: Cardiac magnetic resonance imaging (CMR) is the gold standard for evaluating cardiac morphology and function. Standardized planes are necessary to obtain meaningful quantitative values that can be compared to established normal reference values and used for follow-up. However, manual planning is time-consuming and error-prone. This prospective single-center cohort study evaluated the impact of a fully automated plane positioning tool on the reproducibility of left ventricular (LV) volumetric parameters in CMR.

Methods: 52 healthy volunteers (26 males, median age 44.5 years) were divided into two sub-cohorts undergoing two CMR scans. The 1.5T sub-cohort (n = 32) was scanned twice on a 1.5T scanner with an interval of one to five weeks, the 3T sub-cohort (n = 20) was scanned first at 1.5T and subsequently on a 3T scanner with an interval of one to two hours. All scans were performed using automated and manual planning with identical protocols acquiring short-axis (SAX), 2-chamber (2CH), 3-chamber (3CH) and 4-chamber (4CH) image series. Subjective quality of plane position was rated fully blinded by two radiologists on a 5-point scale. Volumetric analysis was performed fully automatically on SAX images without manual corrections in order to avoid any potential biases arising from manual segmentation. Following LV parameters were retrieved: end-systolic volume (ESV), end-diastolic volume (EDV), stroke volume (SV), and ejection fraction (EF). Wilcoxon matched-pairs signed rank test was used for statistical assessment, along with intraclass correlation coefficient (ICC) and Bland-Altman analysis.

Results: Subjective quality of image planes showed high consistency with good to excellent ratings across all plane orientations for both planning methods in both sub-cohorts (1.5T: manual: 4.5 vs. 4.6/5, p = 0.028; automated: 4.6 vs. 4.7/5, p = 0.235; 3T: manual: 4.6 vs. 4.6/5, p = 0.446; automated: 4.5 vs. 4.5/5, p = 0.487). However, in some cases automated planning avoided suboptimal plane prescription (Figure 1). Reproducibility of EDV, ESV and SV were good to excellent in both sub-cohorts (Table 1). However, EF demonstrated only poor reproducibility in the 1.5T sub-cohort and fair reproducibility in the 3T sub-cohort. This occurred in both planning approaches, suggesting a methodological cause. Overall, the reproducibility of LV parameters was better in the 3T sub-cohort than in the 1.5T sub-cohort, which was most likely caused by the difference in time intervals between scans. ICC values for automated planning were slightly higher than those for manual planning across both sub-cohorts. These trends were also observed in the Bland-Altman analysis (Figure 2).

Conclusion: Fully automated plane positioning for CMR reliably provides high-quality image planes and maintains the high reproducibility of LV volumetric assessment. This innovative technology can simplify CMR scans while maintaining the comparability of follow-up scans.

Left Ventricular Volumetric Results for 1.5T and 3T Sub-Cohort

Left ventricular volumetric results 1.5T sub-cohort
	Manual prescription				Automated prescription
	Scan 1	Scan 2	p-value	ICC	Scan 1	Scan 2	p-value	ICC
LV EDV (ml)	161 (114; 245)	158 (112; 225)	0.304	0.962	163 (121; 249)	160 (116; 226)	0.239	0.938
LV ESV (ml)	64 (42; 105)	68 (39; 112)	0.155	0.815	67 (50; 96)	68 (38; 104)	0.940	0.766
LV SV (ml)	93 (72; 150)	88 (70; 148)	0.059	0.849	96 (70; 152)	93 (68; 145)	0.064	0.894
LV EF (%)	61 (46; 69)	57 (44; 66)	0.100	0.323	60 (53; 69)	59 (44; 69)	0.601	0.213
LV Mass (g)	100 (61; 162)	100 (59; 146)	0.089	0.978	104 (68; 159)	98 (62; 147)	0.654	0.984
Left ventricular volumetric results 3T sub-cohort
	Manual prescription				Automated prescription
	Scan 1	Scan 2	p-value	ICC	Scan 1	Scan 2	p-value	ICC
LV EDV (ml)	152 (101; 190)	140 (95; 185)	0.101	0.972	146 (97; 201)	142 (102; 208)	0.576	0.981
LV ESV (ml)	62 (36; 96)	57 (33; 87)	0.263	0.915	62 (36; 100)	57 (37; 113)	0.502	0.957
LV SV (ml)	88 (58; 118)	84 (62; 113)	0.601	0.893	86 (61; 121)	86 (65; 117)	0.433	0.918
LV EF (%)	59 (50; 65)	59 (51; 67)	0.526	0.597	59 (47; 65)	59 (46; 66)	0.502	0.742
LV Mass (g)	91 (60; 141)	92 (63; 139)	0.654	0.994	92 (63; 144)	91 (62; 142)	0.601	0.992

Table 1: Left ventricular volumetric parameters of the 1.5T sub-cohort (n = 32) and 3T sub-cohort (n = 20) derived from short-axis image stacks. Values presented in medians with corresponding minimum and maximum values. Statistical significance values are presented as p-values calculated with the Wilcoxon matched pairs signed rank test (level of significance p ≤ 0.05). In addition, the ICC was specified. LV – left ventricle; EDV – end-diastolic volume; ESV – end-systolic volume; SV – stroke volume; EF- ejection fraction.Figure 1: Accelerated cine sequences in 4CH view of a healthy 59 years old female. a,b) The first scan was performed at 1.5T. Both prescription approaches provided excellent quality (5/5) of the 4CH image plane. c,d) The second scan was performed at 3T. Manual prescription (c) provided initially a non-diagnostic quality (1/5) of the 4CH view, requiring a repetition of the cine sequence acquisition or, if not reacquired, an impaired comparability between both scans. Automated prescription (d) demonstrated a high reproducibility in this case. 4CH – four chamber view.
Figure 2: Bland-Altman plots of the 1.5T and 3T sub-cohort for left ventricular volumetric parameters derived from short-axis image stacks. The matching colors thick dashed lines represent the mean bias, the matching colors dotted lines the 95% limits of agreement. LV – left ventricle; EDV – end-diastolic volume; ESV – end-systolic volume; SV – stroke volume; EF – ejection fraction.