Highly accelerated free-breathing real-time myocardial tagging for exercise cardiovascular magnetic resonance

Figures

• 

  Fig. 1. Highly accelerated real-time tagging at 3 T. Steady-state free-precession imaging has a higher signal-to-noise ratio than gradient-echo and is thus more robust to high acceleration rates. Tag lines were perpendicular to phase-encoding to preserve harmonics. Off-resonance artifacts were reduced by using fat-saturation (S1) and steady-state ramp-up (S3) pulses. Only five pulses were used to minimize the delay between tag preparation (S2) and the first imaging frame

• 

  Fig. 2. Tagging imaging reconstruction workflow. a A resolution enhancement generative adversarial inline network (REGAIN) architecture was previously trained to deblur cine images. REGAIN was developed using breath-hold electrocardiogram-segmented segmented cine for training. b Reconstruction consisted of vendor-provided compressed sensing (CS) reconstruction per beat to create images from sixfold under-sampled k-space data. Resulting blurred images (with twofold reduced phase encoding spatial resolution) were used as input to REGAIN to deblur low-resolution CS-reconstructed images. The REGAIN recon was used to reconstruct tagging data without any modification. 12-fold accelerated k-space data were reconstructed in-scanner in real-time, and images were available upon completion of scan

• 

  Fig. 3. Prospective studies. A total of 95 subjects were prospectively recruited to evaluate our proposed technique. In the first study, CMR imaging was performed exclusively at rest in 27 patients and 19 healthy subjects. Patients were selected from those undergoing clinical CMR and willing to undergo an additional 5–10 min of scanning. Real-time tagging images were collected in all subjects. To serve as comparison, ECG-gated tagging images were also collected in all patients and a subset of healthy subjects (14 of 19). To assess repeatability, an extra set of real-time tagging images was collected in a subset of patients (6 of 27) and all healthy subjects. In the Ex-CMR study, real-time tagging images were collected in all subjects before and after supine bike exercise. ECG electrocardiogram, Ex-CMR exercise cardiac magnetic resonance

• 

  Fig. 4. CMR and exercise CMR imaging protocols. A Our technique was first evaluated at rest, which consisted of breath-hold, electrocardiogram (ECG)-segmented tagging followed by free-breathing real-time tagging. The real-time acquisition was repeated once to assess repeatability. B Feasibility of real-time tagging imaging with physiology stress was evaluated in subjects prospectively recruited for exercise CMR. Real-time tagging images were collected pre- and post-exercise. Subjects were exercised in the supine position using a cycle ergometer. Work rate was started at Ω0 = 10W, 15W, 20W or 25W, and was increased by ∆Ω0 = + Ω0 every 2 min while subjects maintained a constant pedaling speed of 75 rpm. After 10 min or exhaustion, subjects were immediately transferred for post-exercise stress imaging. This consisted of a repetition of the real-time tagging sequence

• 

  Fig. 5. Breath-hold ECG-segmented and free-breathing real-time tagging imaging. Steady-state free-precession tagging images were collected in patients (a–c) and a healthy subject (d) at 3 T. Tag lines with 8 mm thickness were oriented 90° perpendicular to the phase encoding, which resulted in horizontal and vertical lines. Three slices at basal, mid, and apical levels were acquired. Imaging parameters for both ECG-segmented and real-time acquisitions were: matrix size = 176 × 176, acquired spatial resolution = 4.0 × 2.0 mm2, reconstructed spatial resolution = 2.0 × 2.0 mm2, slice thickness = 8 mm, slice distance factor = 150%, echo and repetition time = 1.1/2.62 ms, temporal resolution = 28.8 ms and flip angle = 30°. Images are shown at end-systole. ECG electrocardiogram

• 

  Fig. 6. Evaluation of mid-wall global circumferential strain (GCS). Tagging images were collected at rest with breath-hold ECG-segmented and with repeated free-breathing real-time tagging sequences. Measurements of mid-wall GCS were obtained from tagging images in the short-axis view at basal, mid and apical levels and were averaged across slices. A The average strain obtained with the two different sequences was compared. Linear regression with forced zero intercept had a slope = 0.87 (0.83, 0.91), and Pearson correlation coefficient r was 0.71. Bland–Altman analysis showed a mean difference of − 2.3% (− 7.4, 2.8). B The average strain obtained from repeated real-time tagging sequence was compared to evaluate intra-subject repeatability of mid-wall GCS measures. Intraclass correlation coefficient was 0.86 (0.71, 0.94). The Bland–Altman analysis showed a mean difference of 0.2% (− 2.3, 2.8). In Bland–Altman plots, limits of agreement are shown as dotted black lines, and biases as solid red lines

• 

  Fig. 7. Clinical case illustrating microvascular dysfunction. A 54-year-old female presenting with persistent chest pain without angina symptoms or ischemic ECG changes on the stress test. Coronary angiography revealed a right dominant system with catheter-induced vasospasm in the right coronary artery and evidence of microvascular dysfunction. Following a 10-min exercise CMR protocol with a 15-W resistance increase every 2 min, her heart rate increased from 56 to 130 bpm. Real-time tagging CMR images were acquired pre- and post-exercise in short-axis and 2-chamber views. Tagging line quality for all acquisitions in both views was excellent (score = 1), characterized by well-defined tagline patterns with clear differentiation between dark and bright lines. Post-exercise tag lines were markedly deformed. All images shown near end-systole. ECG electrocardiogram

• 

  Fig. 8. Clinical case of heart failure. A 29-year-old male with a history of heart failure and an ejection fraction of 33%. After a 10-min exercise CMR protocol with a 10-W resistance increase every 2 min, his heart rate increased from 80 to 116 bpm. Real-time tagging images were acquired pre- and post-exercise in short-axis and 2-chamber views. Tagging line quality for all acquisitions in both views was good (score = 2), defined as a well-defined pattern but with partial signal intensity in saturated lines. Tagging lines demonstrated a moderate increase in myocardial deformation post-exercise. CMR cardiac magnetic resonance

• 

  Fig. 9. Subjective evaluation of image quality. Short-axis images from 95 subjects at rest and 49 subjects post-exercise were evaluated. For 2-chamber images, scans from a total of 92 subjects at rest and 44 subjects pre-exercise were assessed. Therefore, there were a combined total of 280 scans, including 2-chamber and short-axis images. Two readers separately rated data on a 4-point Likert scale based on tagline quality (1-excellent; 2-good; 3-moderate; 4-poor) and artifact level (1-none; 2-minimal; 3-moderate; 4-significant). a The tagging line quality score for 2-chamber and short-axis images at rest was 1.9 ± 0.8 and 1.6 ± 0.7. Post-exercise, the 2-chamber score increased to 2.5 ± 0.8, and the score for short-axis images increased to 1.9 ± 0.7. The artifact level for both 2-chamber and short-axis images pre-exercise was 1.4 ± 0.7 and 1.2 ± 0.5. Artifact level rose to 1.5 ± 0.8 and 1.3 ± 0.6 for 2-chamber and short-axis images, respectively

• 

  Fig. 10. Quantification of myocardial deformation in Ex-CMR real-time tagging imaging. a A 55-year-old male with a history of heart failure with preserved ejection fraction. Real-time tagging images were acquired pre- and post-exercise in short-axis views. Quantification of myocardial deformation shows a minimal response. b A 63-year-old male with a history of heart failure with reduced ejection fraction. Quantification of showed an augmented myocardial deformation