Deep learning-based left ventricular segmentation demonstrates improved performance on respiratory motion-resolved whole-heart reconstructions.

Introduction: Deep learning (DL)-based segmentation has gained popularity for routine cardiac magnetic resonance (CMR) image analysis and in particular, delineation of left ventricular (LV) borders for LV volume determination. Free-breathing, self-navigated, whole-heart CMR exams provide high-resolution, isotropic coverage of the heart for assessment of cardiac anatomy including LV volume. The combination of whole-heart free-breathing CMR and DL-based LV segmentation has the potential to streamline the acquisition and analysis of clinical CMR exams. The purpose of this study was to compare the performance of a DL-based automatic LV segmentation network trained primarily on computed tomography (CT) images in two whole-heart CMR reconstruction methods: (1) an in-line respiratory motion-corrected (Mcorr) reconstruction and (2) an off-line, compressed sensing-based, multi-volume respiratory motion-resolved (Mres) reconstruction. Given that Mres images were shown to have greater image quality in previous studies than Mcorr images, we hypothesized that the LV volumes segmented from Mres images are closer to the manual expert-traced left ventricular endocardial border than the Mcorr images. Method: This retrospective study used 15 patients who underwent clinically indicated 1.5 T CMR exams with a prototype ECG-gated 3D radial phyllotaxis balanced steady state free precession (bSSFP) sequence. For each reconstruction method, the absolute volume difference (AVD) of the automatically and manually segmented LV volumes was used as the primary quantity to investigate whether 3D DL-based LV segmentation generalized better on Mcorr or Mres 3D whole-heart images. Additionally, we assessed the 3D Dice similarity coefficient between the manual and automatic LV masks of each reconstructed 3D whole-heart image and the sharpness of the LV myocardium-blood pool interface. A two-tail paired Student's t-test (alpha = 0.05) was used to test the significance in this study. Results & Discussion: The AVD in the respiratory Mres reconstruction was lower than the AVD in the respiratory Mcorr reconstruction: 7.73 ± 6.54 ml vs. 20.0 ± 22.4 ml, respectively (n = 15, p-value = 0.03). The 3D Dice coefficient between the DL-segmented masks and the manually segmented masks was higher for Mres images than for Mcorr images: 0.90 ± 0.02 vs. 0.87 ± 0.03 respectively, with a p-value = 0.02. Sharpness on Mres images was higher than on Mcorr images: 0.15 ± 0.05 vs. 0.12 ± 0.04, respectively, with a p-value of 0.014 (n = 15). Conclusion: We conclude that the DL-based 3D automatic LV segmentation network trained on CT images and fine-tuned on MR images generalized better on Mres images than on Mcorr images for quantifying LV volumes.

Effect of Patient-Specific Coronary Flow Reserve Values on the Accuracy of MRI-Based Virtual Fractional Flow Reserve.

The purpose of this study is to investigate the effect of varying coronary flow reserve (CFR) values on the calculation of computationally-derived fractional flow reserve (FFR). CFR reflects both vessel resistance due to an epicardial stenosis, and resistance in the distal microvascular tissue. Patients may have a wide range of CFR related to the tissue substrate that is independent of epicardial stenosis levels. Most computationally based virtual FFR values such as FFRCT do not measure patient specific CFR values but use a population-average value to create hyperemic flow conditions. In this study, a coronary arterial computational geometry was constructed using magnetic resonance angiography (MRA) data acquired in a patient with moderate CAD. Coronary flow waveforms under rest and stress conditions were acquired in 13 patients with phase-contrast magnetic resonance (PCMR) to calculate CFR, and these flow waveforms and CFR values were applied as inlet flow boundary conditions to determine FFR based on computational fluid dynamics (CFD) simulations. The stress flow waveform gave a measure of the functional significance of the vessel when evaluated with the physiologically-accurate behavior with the patient-specific CFR. The resting flow waveform was then scaled by a series of CFR values determined in the 13 patients to simulate how hyperemic flow and CFR affects FFR values. We found that FFR values calculated using non-patient-specific CFR values did not accurately predict those calculated with the true hyperemic flow waveform. This indicates that both patient-specific anatomic and flow information are required to accurately non-invasively assess the functional significance of coronary lesions.

In Vivo Quantification of Regional Circumferential Green Strain in the Thoracic and Abdominal Aorta by Two-Dimensional Spiral Cine DENSE MRI.

Regional tissue mechanics play a fundamental role in the patient-specific function and remodeling of the cardiovascular system. Nevertheless, regional in vivo assessments of aortic kinematics remain lacking due to the challenge of imaging the thin aortic wall. Herein, we present a novel application of displacement encoding with stimulated echoes (DENSE) magnetic resonance imaging (MRI) to quantify the regional displacement and circumferential Green strain of the thoracic and abdominal aorta. Two-dimensional (2D) spiral cine DENSE and steady-state free procession (SSFP) cine images were acquired at 3T at either the infrarenal abdominal aorta (IAA), descending thoracic aorta (DTA), or distal aortic arch (DAA) in a pilot study of six healthy volunteers (22-59 y.o., 4 females). DENSE data were processed with multiple custom noise reduction techniques including time-smoothing, displacement vector smoothing, sectorized spatial smoothing, and reference point averaging to calculate circumferential Green strain across 16 equispaced sectors around the aorta. Each volunteer was scanned twice to evaluate interstudy repeatability. Circumferential Green strain was heterogeneously distributed in all volunteers and locations. The mean spatial heterogeneity index (standard deviation of all sector values divided by the mean strain) was 0.37 in the IAA, 0.28 in the DTA, and 0.59 in the DAA. Mean (homogenized) peak strain by DENSE for each cross section was consistent with the homogenized linearized strain estimated from SSFP cine. The mean difference in peak strain across all sectors following repeat imaging was -0.1±2.3%, with a mean absolute difference of 1.7%. Aortic cine DENSE MRI is a viable noninvasive technique for quantifying heterogeneous regional aortic wall strain and has significant potential to improve patient-specific clinical assessments of numerous aortopathies, as well as to provide the lacking spatiotemporal data required to refine patient-specific computational models of aortic growth and remodeling.