Presentation on theme: "Evaluation of Reconstruction Techniques"— Presentation transcript:
1Evaluation of Reconstruction Techniques A MATLAB Toolbox for Parallel Imaging using Multiple Phased Array CoilsSwati D. Rane, Jim X. JiMagnetic Resonance Systems Laboratory, Department of Electrical Engineering, Texas A&M UniversityParallel Magnetic Resonance ImagingCoil Sensitivity Function:Artifact Power :In simulation, coil maps are generated witha linear array of receivers with Gaussian profilesor a non-linear array of receivers with 2D Gaussianprofiles.Biot- Savart’s Law*Parallel Magnetic Resonance Imaging (MRI) uses an array of receivers/ transceivers to accelerate imaging speed, by reducing the phase encodings. The image is reconstructed using different methods such as SENSE , PILS , SMASH , GRAPPA , SPACE RIP , SEA  and their variations, utilizing complimentary information from all the channels.Coil sensitivity is estimated byUse of reference scans and divide by a body coil imageUse of extra calibration lines and Sum-Of-Squares techniqueUsing singular value decomposition‘g’ factor for SENSE:Need of a Toolbox for Parallel MRI.x = point by point multiplicationS = sensitivity encoding matrixΨ = noise correlation matrixQuality of the reconstructed image by depends on:Receiver coil array configuration and coil localizationk-space coverageParallel Imaging technique used for reconstructionResolution:Filtering for noise reduction byPolynomial filteringWindowingMedian filteringWavelet denoisingUse of different phantoms to check degradationOptimality of the reconstruction can be evaluated on thebasis of:Signal-to-Noise Ratio (SNR)Artifact PowerResolution‘g’ factor (for SENSE) or numerical conditionsComputational complexityImage Reconstruction:SENSE: 1D SENSE, Regularized SENSE, 2D SENSE*PILS:Fig.3: Resolution phantomsSMASH: Basic SMASH, AUTO-SMASHThere is a need of a toolTo help select the optimal method for a given imagingenvironmentTo provide a platform for developing new algorithmsTo facilitate the learning/ testing of parallel imagingmethodsConclusionSPACE RIP: Variable density sampling and reconstructionGRAPPA: Multiple block implementationA software tool has been developed in MATLAB to analyzeparallel imaging methods on the basis of SNR, resolution,artifact power and computational complexity.* Yet to be doneEvaluation of Reconstruction TechniquesThe MATLAB ToolboxThe toolbox can be used as a learning or testing tool andas a platform for developing new imaging methods.Signal-to-Noise Ratio (SNR):Sensitivity EstimationFilteringData Input- Simulated dataAcquired dataImproved ReconstructionIterative SOSReconstructionRegularized SENSEAUTO-SMASHPerformance Analysis- SNRArtifact Power‘g’ factor calculation- ResolutionComputationsSENSEHarmonics- fittingGaussian fittingSMASHSPACE RIPGRAPPAPILSReferencesMethod 1: Pruessmann K., et al., MRM, 42: , Nov.1999. Grisworld M., et al., MRM, 44: , OctROI Sodickson D., et al., MRM, 38: , 1997.Fig.2: SNR Calculation: Selection of regionof interest (ROI) and noise(RON) Grisworld M., et al., MRM, 47: , June 2002. Kyriakos W., et al., MRM, 44: , AugRON Wright S., et al., Proc. Of 2nd Joint EMBS/BEMS Conference,OctMethod 2: Kellman P., et al., IEEE Proc., Intl. Symposium On BiomedicalImaging, July 2002. Walsh D., et al., MRM, 43: , SeptMethod 3 ( with two acquisitions):Fig.1: Block Diagram of the developed toolbox Hsuan-Lin F., et al., MRM, 51: , 2004. Jakob P., et al., MAGMA, 7:42:54, 1998.Data Input:Simulated coil sensitivities and k-space dataAcquired/ real data collected from the MR scanner Firbank M., et al., Phys.Med.Biol, 44:N261-N264, 1999.S1 = mean signal intensity in the ROI of the one imageSD1-2 = std. deviation in the ROI of the subtraction image Weiger M., et al., MAGMA, 14:1-19, March 2002.