Online real‐time reconstruction of adaptive TSENSE with commodity CPU/GPU hardware

Abstract

Adaptive temporal sensitivity encoding (TSENSE) has been suggested as a robust parallel imaging method suitable for MR guidance of interventional procedures. However, in practice, the reconstruction of adaptive TSENSE images obtained with large coil arrays leads to long reconstruction times and latencies and thus hampers its use for applications such as MRguided thermotherapy or cardiovascular catheterization. Here, we demonstrate a real-time reconstruction pipeline for adaptive TSENSE with low image latencies and high frame rates on affordable commodity personal computer hardware. For typical image sizes used in interventional imaging (128 ؋ 96, 16 channels, sensitivity encoding (SENSE) factor 2-4), the pipeline is able to reconstruct adaptive TSENSE images with image latencies below 90 ms at frame rates of Real-time MRI for therapy control is used to provide retroactive feedback information for interventional applications such as catheter guidance or MRI-controlled tissue ablation. Such interventions require several minutes of continuous imaging and are preferably carried out under free-breathing conditions, which presents the following challenges to MR acquisition and reconstruction.The data acquisition has to be rapid enough to resolve physiologic motion such as respiratory or cardiac motion and the signal changes from the interventional process itself. In addition, the employed data acquisition and reconstruction schemes must not introduce long image latencies because this degrades the value of the image data for feedback control.One of the significant developments in this field was the introduction of parallel imaging methods such as simultaneous acquisition of spatial harmonics (SMASH) (1), sensitivity encoding (SENSE) (2), or generalized auto-calibrating partially parallel acquisitions (GRAPPA) (3), which allows, on one hand, acceleration of the data acquisition and, on the other hand, shortening of readout echo-trains and thus reduces signal dropout and image distortions in regions of large susceptibility variations.Apart from variable-density methods, such as GRAPPA, which incorporate self-calibration by nature of the sampling pattern, these parallel imaging methods require coil sensitivity data for the image reconstruction, which depends on the coil position. Because of the long duration of interventional imaging procedures, displacements of the receiver coils are common and lead either to increased image artifacts over time or, in the cases in which the receiver coils are embedded in the interventional device itself, the complete unsuitability of these approaches. Temporal sensitivity encoding (TSENSE) (4) is an approach that overcomes this problem by collecting highresolution calibration data at a lower temporal resolution than that used for image encoding. However, the additional computational overhead leads to reconstruction times that often exceed the MR acquisition time, in particular if large coil arrays are used, and are therefore in practice hard to exploit for real-time therapy guidance.Gutt...

Fast reduction of undersampling artifacts in radial MR angiography with 3D total variation on graphics hardware

“…Therefore, the application of TV based artifact removal is no longer the time limiting step in the imaging chain. The goal of future work is to connect dedicated GPU computation hardware directly to the MR scanner, as was already described in the context of parallel imaging [32]. This will allow interactive TV artifact elimination already during data acquisition.…”

Non-invasive thermotherapy of abdominal organs

“…Although modern MRI scanners offer the possibility of acquisition acceleration using parallel imaging techniques, the TSENSE reconstruction has been investigated in the field of thermometry [1]. This method was of particular interest since it provides a dynamic update of calibration data sets, which is of particular interest for flexible array coils positioned around the abdomen.…”

Medical image processing on the GPU – Past, present and future

“…Hansen et al (2008) described reconstruction of regular SENSE on the GPU. This work was then extended to data collected with temporal SENSE (Roujul et al, 2009) and radial sampling with SENSE . In compressed sensing (Donoho, 2006;Lustig et al, 2007), the idea is to sample the data in a sparse basis.…”