Sebastian Schaetz

A multi-GPU programming library for real-time applications

We present MGPU, a C++ programming library targeted at single-node multi-GPU systems. Such systems combine disproportionate floating point performance with high data locality and are thus well suited to implement real-time algorithms. We describe the library design, programming interface and implementation details in light of this specific problem domain. The core concepts of this work are a novel kind of container abstraction and MPI-like communication methods for intra-system communication. We further demonstrate how MGPU is used as a framework for porting existing GPU libraries to multi-device architectures. Putting our library to the test, we accelerate an iterative non-linear image reconstruction algorithm for real-time magnetic resonance imaging using multiple GPUs. We achieve a speed-up of about 1.7 using 2 GPUs and reach a final speed-up of 2.1 with 4 GPUs. These promising results lead us to conclude that multi-GPU systems are a viable solution for real-time MRI reconstruction as well as signal-processing applications in general.

Real-time flow MRI of the aorta at a resolution of 40 msec.

To evaluate a novel real‐time phase‐contrast magnetic resonance imaging (MRI) technique for the assessment of through‐plane flow in the ascending aorta.

Real-time magnetic resonance imaging of cardiac function and flow—Recent progress

Cardiac structure, function and flow are most commonly studied by ultrasound, X-ray and magnetic resonance imaging (MRI) techniques. However, cardiovascular MRI is hitherto limited to electrocardiogram (ECG)-synchronized acquisitions and therefore often results in compromised quality for patients with arrhythmias or inabilities to comply with requested protocols-especially with breath-holding. Recent advances in the development of novel real-time MRI techniques now offer dynamic imaging of the heart and major vessels with high spatial and temporal resolution, so that examinations may be performed without the need for ECG synchronization and during free breathing. This article provides an overview of technical achievements, physiological validations, preliminary patient studies and translational aspects for a future clinical scenario of cardiovascular MRI in real time.

Real-time phase-contrast flow MRI of haemodynamic changes in the ascending aorta and superior vena cava during Mueller manoeuvre.

AIM To evaluate the potential of real-time phase-contrast flow magnetic resonance imaging (MRI) at 40 ms resolution for the simultaneous determination of blood flow in the ascending aorta (AA) and superior vena cava (SVC) in response to reduced intrathoracic pressure (Mueller manoeuvre). MATERIALS AND METHODS Through-plane flow was assessed in 20 healthy young subjects using real-time phase-contrast MRI based on highly undersampled radial fast low-angle shot (FLASH) with image reconstruction by regularized non-linear inversion. Haemodynamic alterations (three repetitions per subject = 60 events) were evaluated during normal breathing (10 s), inhalation with nearly closed epiglottis (10 s), and recovery (20 s). RESULTS Relative to normal breathing and despite interindividual differences, reduced intrathoracic pressure by at least 30 mmHg significantly decreased the initial peak mean velocity (averaged across the lumen) in the AA by -24 ± 9% and increased the velocity in the SVC by +28 ± 25% (p < 0.0001, n = 23 successful events). Respective changes in flow volume per heartbeat were -25 ± 9% in the AA and +49 ± 44% in the SVC (p < 0.0001, n = 23). Flow parameters returned to baseline during sustained pressure reduction, while the heart rate was elevated by 10% (p < 0.0001) after the start (n = 24) and end (n = 17) of the manoeuvre. CONCLUSIONS Real-time flow MRI during low intrathoracic pressure non-invasively revealed quantitative haemodynamic adjustments in both the AA and SVC.

Accelerated Computing in Magnetic Resonance Imaging: Real-Time Imaging Using Nonlinear Inverse Reconstruction

Purpose To develop generic optimization strategies for image reconstruction using graphical processing units (GPUs) in magnetic resonance imaging (MRI) and to exemplarily report on our experience with a highly accelerated implementation of the nonlinear inversion (NLINV) algorithm for dynamic MRI with high frame rates. Methods The NLINV algorithm is optimized and ported to run on a multi-GPU single-node server. The algorithm is mapped to multiple GPUs by decomposing the data domain along the channel dimension. Furthermore, the algorithm is decomposed along the temporal domain by relaxing a temporal regularization constraint, allowing the algorithm to work on multiple frames in parallel. Finally, an autotuning method is presented that is capable of combining different decomposition variants to achieve optimal algorithm performance in different imaging scenarios. Results The algorithm is successfully ported to a multi-GPU system and allows online image reconstruction with high frame rates. Real-time reconstruction with low latency and frame rates up to 30 frames per second is demonstrated. Conclusion Novel parallel decomposition methods are presented which are applicable to many iterative algorithms for dynamic MRI. Using these methods to parallelize the NLINV algorithm on multiple GPUs, it is possible to achieve online image reconstruction with high frame rates.

Simultaneous flow dynamics in small and great thoracic vessels during physiological stress tests and normal breathing using real-time cardiac magnetic-resonance

Simultaneous flow dynamics in small and great thoracic vessels during physiological stress tests and normal breathing using real-time cardiac magnetic-resonance Jan M Sohns, Martin Fasshauer, Johannes T Kowallick, Andreas Schuster, Wieland Staab, Arun Joseph, Shoun Zhang, Dirk Voit, Sebastian Schaetz, Klaus-Dietmar Merboldt, Michael Steinmetz, Christina Unterberg-Buchwald, Jens Frahm, Joachim Lotz

Pulse wave velocity in real-time cardiac magnetic resonance

Background Atherosclerosis and its associated diseases are constantly increasing in developed countries. Aortic stiffness is an indicator for atherosclerosis and is associated with mortality and morbidity especially in aortic abdominal aneurysms. For further evaluation of aortic stiffness, we examined n = 13 healthy volunteers using real-time magnetic-resonance imaging (RT-MRI) with highly undersampled radial fast low-angle shot (FLASH) acquisitions, phase-sensitive image reconstructions and regularized nonlinear inversion (NLINV). We hypothesized that RT-MRI is able to determine pulse wave velocity (PWV) and flow data using just one transverse view of the ascending and descending aorta. This method could be superior to complex known MRI methods using velocity projection today. Methods We assessed PWV as surrogate parameter for aortic stiffness by velocity-encoded RT-MRI. Time lag between the ascending and descending aortic pulse wave was calculated and divided by the mean length of the aortic arch for each individual in detail. RT-MRI can determine PWV during normal breathing, physical strain and recovery from strain using the Valsalva (VM) and Mueller maneuvers (MM). During strain/maneuvers volunteers had visual feedback of intra-thoracic pressure via a mouthpiece to ensure adequate strain performance. We calculated PWV in normal breathing, at the end of strain and at the end of the recovery phase. Due to the advantage of a single heart beat-to-beat variability with