Search

Background: Cardiac MRI has become the gold‐standard imaging technique for assessing cardiovascular morphology and function. In spite of this, its slow data acquisition process presents imaging challenges due to the motion from heartbeats, respiration, and blood flow. In recent studies, deep learning (DL) algorithms have shown promising results for the task of image reconstruction. However, there have been instances where they have introduced artifacts that may be misinterpreted as pathologies or may obscure the detection of pathologies. Therefore, it is important to obtain a metric, such as the uncertainty of the network output, that identifies such artifacts. However, this can be quite challenging for large‐scale image reconstruction problems such as dynamic multi‐coil non‐Cartesian MRI. Purpose: To efficiently quantify uncertainties of a physics‐informed DL‐based image reconstruction method for a large‐scale accelerated 2D multi‐coil dynamic radial MRI reconstruction problem, and demonstrate the benefits of physics‐informed DL over model‐agnostic DL in reducing uncertainties while at the same time improving image quality. Methods: We extended a recently proposed physics‐informed 2D U‐Net that learns spatio‐temporal slices (named XT‐YT U‐Net), and employed it for the task of uncertainty quantification (UQ) by using Monte Carlo dropout and a Gaussian negative log‐likelihood loss function. Our data comprised 2D dynamic MR images acquired with a radial balanced steady‐state free precession sequence. The XT‐YT U‐Net, which allows for training with a limited amount of data, was trained and validated on a dataset of 15 healthy volunteers, and further tested on data from four patients. An extensive comparison between physics‐informed and model‐agnostic neural networks (NNs) concerning the obtained image quality and uncertainty estimates was performed. Further, we employed calibration plots to assess the quality of the UQ. Results: The inclusion of the MR‐physics model of data acquisition as a building block in the NN architecture led to higher image quality (NRMSE: −33±8.2%$-33 \pm 8.2 \%$, PSNR: 6.3±1.3%$6.3 \pm 1.3 \%$, and SSIM: 1.9±0.96%$1.9 \pm 0.96 \%$), lower uncertainties (−46±8.7%$-46 \pm 8.7 \%$), and, based on the calibration plots, an improved UQ compared to its model‐agnostic counterpart. Furthermore, the UQ information can be used to differentiate between anatomical structures (e.g., coronary arteries, ventricle boundaries) and artifacts. Conclusions: Using an XT‐YT U‐Net, we were able to quantify uncertainties of a physics‐informed NN for a high‐dimensional and computationally demanding 2D multi‐coil dynamic MR imaging problem. In addition to improving the image quality, embedding the acquisition model in the network architecture decreased the reconstruction uncertainties as well as quantitatively improved the UQ. The UQ provides additional information to assess the performance of different network approaches. [ABSTRACT FROM AUTHOR]

Objectives: To determine whether noninvasive assessment of pulmonary artery flow (Qp) by cardiac magnetic resonance (CMR) would predict pulmonary vascular resistance (PVR) in patients with congenital heart disease characterized by an unrestricted left-to-right shunt. Background: Patients with an unrestricted left-to-right shunt who are at risk of obstructive pulmonary vascular disease require PVR evaluation preoperatively. CMR cardiac catheter (XMR) combines noninvasive measurement of Qp by phase contrast imaging with invasive pressure measurement to accurately determine the PVR. Methods: Patients referred for clinical assessment of the PVR were included. The XMR was used to determine the PVR. The noninvasive parameters, Qp and left-to-right shunt (Qp/Qs), were compared with the PVR using univariate regression models. Results: The XMR was undertaken in 26 patients (median age 0.87 years)—ventricular septal defect 46.2%, atrioventricular septal defect 42.3%. Mean aortic flow was 2.24 ± 0.59 l/min/m2, and mean Qp was 6.25 ± 2.78 l/min/m2. Mean Qp/Qs was 2.77 ± 1.02. Mean pulmonary artery pressure was 34.8 ± 10.9 mm Hg. Mean/median PVR was 5.5/3.0 Woods Units (WU)/m2 (range 1.7 to 31.4 WU/m2). The PVR was related to both Qp and Qp/Qs in an inverse exponential fashion by the univariate regression equations PVR = exp(2.53 − 0.20[Qp]) and PVR = exp(2.75 − 0.52[Qp/Qs]). Receiver-operator characteristic (ROC) analysis was used to determine cutoff values for Qp and Qp/Qs above which the PVR could be regarded as clinically acceptable. A Qp of ≥6.05 l/min/m2 predicted a PVR of ≤3.5 WU/m2 with sensitivity 72%, specificity 100%, and area under the ROC curve 0.90 (p = 0.002). A Qp/Qs of ≥2.5/1 predicted a PVR of ≤3.5 WU/m2 with sensitivity 83%, specificity 100%, and area under the curve ROC 0.94 (p < 0.001). Conclusions: Measurement of Qp or left-to-right shunt noninvasively by CMR has potential to predict the PVR in patients with an unrestricted left-to-right shunt and could potentially determine operability without having to undertake invasive testing. [Copyright &y& Elsevier]

Background: Positron emission tomography (PET) is the non-invasive reference standard for myocardial blood flow (MBF) quantification. Hybrid PET-MR allows simultaneous PET and cardiac magnetic resonance (CMR) acquisition under identical experimental and physiological conditions. This study aimed to determine feasibility of simultaneous 13N-Ammonia PET and dynamic contrast-enhanced CMR MBF quantification in phantoms and healthy volunteers. Methods: Images were acquired using a 3T hybrid PET-MR scanner. Phantom study: MBF was simulated at different physiological perfusion rates and a protocol for simultaneous PET-MR perfusion imaging was developed. Volunteer study: five healthy volunteers underwent adenosine stress. 13N-Ammonia and gadolinium were administered simultaneously. PET list mode data was reconstructed using ordered subset expectation maximisation. CMR MBF was quantified using Fermi function-constrained deconvolution of arterial input function and myocardial signal. PET MBF was obtained using a one-tissue compartment model and image-derived input function. Results: Phantom study: PET and CMR MBF measurements demonstrated high repeatability with intraclass coefficients 0.98 and 0.99, respectively. There was high correlation between PET and CMR MBF (r = 0.98, p < 0.001) and good agreement (bias − 0.85 mL/g/min; 95% limits of agreement 0.29 to − 1.98). Volunteer study: Mean global stress MBF for CMR and PET were 2.58 ± 0.11 and 2.60 ± 0.47 mL/g/min respectively. On a per territory basis, there was moderate correlation (r = 0.63, p = 0.03) and agreement (bias − 0.34 mL/g/min; 95% limits of agreement 0.49 to − 1.18). Conclusion: Simultaneous MBF quantification using hybrid PET-MR imaging is feasible with high test repeatability and good to moderate agreement between PET and CMR. Future studies in coronary artery disease patients may allow cross-validation of techniques. [ABSTRACT FROM AUTHOR]

Abstract: Background: To compare the ventricular response to dobutamine stress between adult patients with chronic pulmonary regurgitation (PR) after repair of tetralogy of Fallot (r-TOF) and healthy volunteers using a staged dobutamine stress MR (DS-MR) protocol. Methods: Eighteen r-TOF patients (median age 31.9years, range 16.2–60.1) with severe PR and 10 healthy controls (median age 40.6years, range 23.9–51.8) completed staged DS-MR (baseline, 10 and 20μg/kg/min) with ventricular volumetry and pulmonary flow quantification. Comparative analysis involved 3-way ANOVA, t-test, regression analysis, and coefficient of variance. Results: All controls had significant increase of ejection fraction (EF) at each stress level for both ventricles (normal contractile reserve, all p<0.05). In r-TOF patients (RV-EDV 126±27ml/m2, RV-EF 55±7%, LV-EF 58±6%, PR-fraction 43±15%), low-dose DS-MR at 10μg/kg/min demonstrated normal biventricular contractile reserve as seen in volunteers. On increase from 10 to 20μg/kg/min a subgroup showed worsening ejection fraction (n=8, p<0.05), mainly due to lack of reduction or even increase of RV-ESV, while the remainder responded with further reduction of RV-ESV and RV-EDV (n=10, p<0.05) and a non-significant trend to increased EF. This different response could not be predicted at baseline. Conclusions: In r-TOF patients with chronic PR, DS-MR at 10μg/kg/min showed normal biventricular systolic response compared with controls. Increase to 20μg/kg/min provoked abnormal RV-ESV response in some r-TOF patients, suggesting presence of ventricular systolic dysfunction not evident at rest. [Copyright &y& Elsevier]