Meeting Banner
Abstract #1473

Assessment of Compressed-Sensing Reconstruction Fidelity for Depicting Ventilation Defects in Hyperpolarized He3 MRI Using H1 Image-Masked Segmentation

Kun Qing1, Nicholas J. Tustison2, Talissa A. Altes2, Jaime F. Mata2, G. Wilson Miller2, Eduard E. De Lange2, William A. Tobias3, Gordon D. Cates Jr. 3, James R. Brookeman2, John P. Mugler, III, 12

1Biomedical Engineering, University of Virginia, Charlottesville, VA, United States; 2Radiology and Medical Imaging, University of Virginia, Charlottesville, VA, United States; 3Physics, University of Virginia, Charlottesville, VA, United States

Previous studies have demonstrated the potential utility of the combined acquisition of helium-3 (He3) and proton (H1) 3D image sets within one breath-hold, accelerated using the compressed-sensing (CS) technique. The purpose of this study was to use an automated segmentation method to compare and quantify the ventilation defects found in fully-sampled versus CS-reconstructed undersampled He3 image sets acquired in the same subjects but during different breath-holds. Relatively high similarities were found between the segmentation results. Much of the difference appears to be due to real variation of ventilation defects between breath holds rather than artifacts related to the CS acquisition.

Abstract PDF Poster PDF

Keywords

absolute accelerated acquisition acquisitions actual advanced agreement ants appeared appears applied approved arrows artifacts assessment asthmatics audience automated axial bias biomedical board breath built caused chest class collisional combined compressed coronal corrected custom datasets defects depicting described diaphragm dice direct eliminate engineering equilibrium example exchange experimental fidelity full fully function generated good gradient helium hold holds implemented included incorporating informed inhomogeneity institutional intensities intensity john label labeled labels lung magnetization major mapping maps masked masks matrix measured medical metric miller much normalization normally optically others overlaid package patterns percentage physician physics pixel polarized poor poorly precession protocol proton pulmonary pulse quantifying quantitative quantitatively rather real reconstructed reconstruction reconstructions register registration related relatively reported representing respective review reviewing rubidium sampled sampling scripts segment segmentation segmentations segmented sensing sets settings similarity simulated solutions spatial steady structural subjects system table target template templates tools transform underlying unrelated variation ventilated ventilation versus whole worse yellow