Meeting Banner
Abstract #3075

Spatial Information Based DCE-MRI Data Reconstruction and Analysis Using PCA

Dattesh D. Shanbhag1, Suresh E. Joel1, Ming-Ching Chang2, Kumar T. Rajamani1, Sandeep Narendra Gupta2, Rakesh Mullick1

1GE Global Research, Bangalore, Karanataka, India; 2GE Global Research, Niskayuna, NY, United States

In this work, we investigated a block-wise PCA based approach to reconstruct DCE-MRI data. It is based on the hypothesis that there exists overlapping temporal information within a spatial neighborhood which can be exploited to separate noise from true contrast enhancement while maintaining tissue heterogeneity. We demonstrate that PCA based reconstruction of dynamic DCE data produced smooth parametric maps while preserving the lesion conspicuity, compared to pixelated maps produced using voxel-by voxel analysis. This will improve the accuracy of DCE-MRI quantification and enhance the sensitivity of the method in clinical scenario.

Keywords

accuracy analyzed anatomy approaches appropriate approved arrival basis block bolus bottom cancer cleanup clinical component components computational computed concentration concern concomitant confidence considered consistent contrast cost curve curves datasets derived described describing developed dimensional disease driven dynamic elevated enhance enhancement ensemble entire error errors especially estimation even fidelity filter fits fitting fixed fluctuations focal functional global helps heterogeneity house implies improve improved improvement improves improving include incorporating injected introduce introduced kinetic largest lesion lesions linear location maps matrix measure median methodology micro model motion much neighborhood neighborhoods next noise noisy overall overlapping parametric patient patients phantom poor post preserving previously principal procedure processing progression prostate providing quantification quantitative radiology recognized reconstructed reconstruction recorded residual scanner section semi sensitivity separate separated separation simple since slope smooth smoothening software sophisticated spatial spectroscopic stacked statistical strategy susceptible sweep sweeps systemic table thereby tissue tool torso trans true tumor type ultimately variance variations visited visual