Loading...
Please wait, while we are loading the content...
Similar Documents
Fast 3 D Free-breathing Abdominal Dynamic Contrast Enhanced MRI with High Spatiotemporal Resolution
| Content Provider | Semantic Scholar |
|---|---|
| Author | Zhang, Tao Cheng, Joseph Y. Alley, Marcus T. Uecker, Martin Lustig, Michael Pauly, John M. Vasanawala, Shreyas S. |
| Copyright Year | 2013 |
| Abstract | Purpose: Dynamic Contrast Enhanced (DCE) MRI is commonly used to detect and characterize lesions. For 3D DCE MRI, a trade-off between spatial and temporal resolution is often necessary. A free-breathing DCE acquisition has high scan efficiency, but image quality can be compromised by respiratory motion. In this work, a soft-gated locally low rank parallel imaging method is proposed for free-breathing DCE MRI. The proposed method can significantly reduce motion artifacts and reconstruct highly undersampled DCE datasets with high spatiotemporal resolution (approximately 1 mm spatial resolution and 4 s frame rate). The proposed method has been validated on in vivo datasets. Theory: DCE MRI can be accelerated by low rank methods: DCE images can be reformatted into a spatiotemporal matrix (Casorati matrix), where each column represents an image at one temporal phase. The data redundancy of DCE datasets is reflected by the low rank property of this spatiotemporal matrix. The spatiotemporal matrix has even lower rank when only a local region (image block) is considered. This is referred to as the locally low rank property. In free-breathing acquisitions, data inconsistency due to respiration will create motion artifacts such as image blurring. To reduce the motion artifacts, a soft-gating approach can be used: k-space data points are assigned with a motion weighting (ranging from 0 to 1) according to the respiratory motion obtained from navigators. Data points with more motion are assigned with smaller data consistency weighting. Soft-gating, locally low rank, and parallel imaging can be combined to reconstruct highly undersampled free-breathing DCE datasets. For simplicity, a 2D acquisition is assumed, and the following variables are defined: mt as the image at time t (size: nx×ny), m as the entire DCE image series (size: nx×ny×T), yt as a matrix of acquired k-space data from all coils at time t (size: nx×ny×nc), S as the coil sensitivity (size: nx×ny×nc), F as a Fourier transform operator, Dt as the undersampling operator at time t, Cb as an operator that takes a block of m (size: bx×by×T) and reformats it into a spatiotemporal matrix (size: bxby×T), and Wt as the soft-gating function. Then the reconstruction can be formulated as: minimizem Σb||Cbm||*, subject to: ||Wt(DtFSmt -yt)|| < ε, t=1,2,...,T where ||x||* is the nuclear norm of matrix x and ε is the error. A projection onto convex sets type method was used to solve this problem. In this work, 16×16 image blocks were used. S was calculated using ESPIRiT from time-averaged data, and two sets of eigenvector maps were used in case of an overlapped field of view (FOV). The motion weighting was generated based on the respiratory motion measured by Butterfly. Methods: A 6-year-old patient was scanned on a 3T scanner using a 36-phase fat-suppressed 3D Butterfly sequence with variable density radial view ordering (VDRad) and a 32-channel cardiac coil. The acquisition parameters were: TR/TE = 3.0/1.2 ms, flip angle = 15, matrix = 320×180×78, FOV = 34×27×16 cm. The total acceleration factor per temporal phase was 14.4 and the frame rate was 4.07 s. Three reconstructions were compared: (1) frameby-frame soft-gated compressed sensing parallel imaging (softgated L1-ESPIRiT ); (2) locally low rank parallel imaging (locally low rank ESPIRiT) without soft-gating; and (3) soft-gated locally low rank ESPIRiT. Results: The measured respiratory motion and the corresponding weighting function are shown in Fig. 1. An example of the reconstructed image is shown in Fig. 2. Soft-gated L1-ESPIRiT suffered from severe image blurring due to high acceleration. Locally low rank ESPIRiT was also blurry because of respiratory motion. The proposed soft-gated locally low rank ESPIRiT method significantly reduced motion artifacts, reflected by the sharp delineation of the hepatic vein and stomach (arrows). The rapid contrast dynamics (liver, spleen, kidney, etc) were also captured and shown in Fig. 2. Together, this demonstrates the feasibility of depicting small rapidly enhancing structures in a small child with rapid hemodynamics during a free-breathing acquisition. Conclusion: A soft-gated locally low rank ESPIRiT method has been proposed and validated for fast 3D free-breathing abdominal DCE MRI. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://submissions.mirasmart.com/ismrm2014/proceedings/files/0332.pdf |
| Alternate Webpage(s) | http://www.eecs.berkeley.edu/~uecker/ismrm14/0332.pdf |
| Alternate Webpage(s) | http://wwwuser.gwdg.de/~muecker1/ismrm14/0332.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
