- Poster presentation
- Open Access
Accelerated three-dimensional free-breathing first pass cardiac perfusion at 1.5T
© Akcakaya et al; licensee BioMed Central Ltd. 2013
- Published: 30 January 2013
- Subjective Image Quality
- Image Quality Score
- Local Imaging Feature
- Image Reconstruction Technique
- Random Undersampling
Perfusion CMR provides an important diagnostic tool for the detection of myocardial ischemia in patients with known or suspected CAD . Conventionally, 2D multislice imaging with 2-4 short-axis slices is often used for evaluation of left ventricular perfusion. 3D CMR perfusion has been proposed for its superior coverage and higher SNR . However, accelerated imaging is required for adequate spatio-temporal resolution. Recent studies in accelerated CMR have shown that utilizing local information in the imaging volume significantly improves reconstruction quality [3, 4]. However, such highly accelerated k-t based techniques used for 3D perfusion  require breath-hold acquisitions, which may be difficult following physical stress or for some patients. In this study, we sought to develop a compressed-sensing (CS) based image reconstruction technique utilizing local imaging features for highly-accelerated 3D free-breathing perfusion CMR.
9 patients (46.0±14.8 years; 6 males) underwent 3D perfusion CMR on a 1.5T Philips Achieva magnet. Images were acquired using a saturation-recovery GRE sequence (TR/TE/α=2.1/1.2 ms/20o). 7.5-fold random undersampling, with fully-sampled central k-space (11 × 3 in ky-kz) was used to acquire 8 slices at 2.3 × 2.3 × 10 mm3 resolution within a 250 ms imaging window.
In the proposed reconstruction technique, each dynamic was first reconstructed individually with total variation (TV) regularization, which were then used for utilizing local features as follows: 10×10 image patches were hypothesized to be slowly-varying across 5 consecutive heartbeats. The locality in each volume was used to capture how various anatomies move differently, whereas the locality in time was used to reduce temporal blurring artifacts due to free-breathing. Principal component (pc) bases were derived for each of these 10 × 10 × 1 × 5 overlapping image volumes. Then the distinct x-pc sparsity of these volumes was used for thresholding the image in an iterative B1-weighted technique. Comparison images were generated using dynamic-by-dynamic (DBD) TV reconstruction, reconstruction using a pc basis for the whole volume (x-pc), and zerofilling of the data. Images were then scored by a blinded reader (1=poor, 4=excellent).
The proposed framework allows for 7.5-fold accelerated free-breathing 3D perfusion reconstruction at 1.5T. Further clinical studies are needed to assess its diagnostic value in patients with suspected CAD.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.