TY - GEN
T1 - MR based PET attenuation correction (MRAC) and anatomical localization ofhuman brain using an optimized UTE-mDixon pulse sequence
AU - Hu, Lingzhi
AU - Stehning, Christian
AU - Nguyen, Nghi
AU - Sher, Andrew
AU - Martinezrios, Claudia
AU - Hu, Zhiqiang
AU - Shao, Lingxiong
N1 - Publisher Copyright:
© 2014 IEEE.
PY - 2016/3/10
Y1 - 2016/3/10
N2 - MR based Attenuation Correction (MRAC) is essential for PET quantitation and image quality assurance in PET/MR. Ultra-short TE (UTE) sequence is promising in generating positive contrast for cortical bone but its further adoption is limited by prohibitively long scan time, lack of soft tissue contrast, and potential ambiguity in a tissue classification due to MR imaging artifacts. In this investigation, we aimed to develop a new MRAC method that consists an optimized under-sampled UTE-mDixon sequence and an iterative voxel-based tissue classification algorithm to generate 4-compartment μ-map, including water, fat, bone and air cavity. In vivo UTE-mDixon images were acquired on 12 human subjects and the developed segmentation method was employed for tissue classification. Diagnostic quality of MR images and the tissue classification accuracy for MRAC was evaluated by three radiologists independently. As a unique advantage over other MRAC sequences, the under-sampled UTE-mDixon of whole brain with retrospective trajectory delay calibration can be finished within less than 3 minutes providing both high quality water/fat separation images for anatomical localization and UTE images for bone segmentation. Robust tissue classification was achieved in all subjects as evaluated by radiologists. The developed MR scan methodology together with tissue classification algorithm may provide a one-scan solution for attenuation correction and anatomical localization in PET/MR.
AB - MR based Attenuation Correction (MRAC) is essential for PET quantitation and image quality assurance in PET/MR. Ultra-short TE (UTE) sequence is promising in generating positive contrast for cortical bone but its further adoption is limited by prohibitively long scan time, lack of soft tissue contrast, and potential ambiguity in a tissue classification due to MR imaging artifacts. In this investigation, we aimed to develop a new MRAC method that consists an optimized under-sampled UTE-mDixon sequence and an iterative voxel-based tissue classification algorithm to generate 4-compartment μ-map, including water, fat, bone and air cavity. In vivo UTE-mDixon images were acquired on 12 human subjects and the developed segmentation method was employed for tissue classification. Diagnostic quality of MR images and the tissue classification accuracy for MRAC was evaluated by three radiologists independently. As a unique advantage over other MRAC sequences, the under-sampled UTE-mDixon of whole brain with retrospective trajectory delay calibration can be finished within less than 3 minutes providing both high quality water/fat separation images for anatomical localization and UTE images for bone segmentation. Robust tissue classification was achieved in all subjects as evaluated by radiologists. The developed MR scan methodology together with tissue classification algorithm may provide a one-scan solution for attenuation correction and anatomical localization in PET/MR.
UR - http://www.scopus.com/inward/record.url?scp=84965079425&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84965079425&partnerID=8YFLogxK
U2 - 10.1109/NSSMIC.2014.7431028
DO - 10.1109/NSSMIC.2014.7431028
M3 - Conference contribution
AN - SCOPUS:84965079425
T3 - 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2014
BT - 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2014
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - IEEE Nuclear Science Symposium and Medical Imaging Conference, NSS/MIC 2014
Y2 - 8 November 2014 through 15 November 2014
ER -