TY - GEN
T1 - MR-based attenuation correction for whole-body PET/MR system
AU - Hu, Z.
AU - Renisch, S.
AU - Schweizer, B.
AU - Blaffert, T.
AU - Ojha, N.
AU - Guo, T.
AU - Tang, J.
AU - Tung, C.
AU - Kaste, J.
AU - Schulz, V.
AU - Torres, I.
AU - Shao, L.
PY - 2010/12/1
Y1 - 2010/12/1
N2 - Philips has introduced the world's first whole body sequential PET/MR system. We present the current status of MR-based attenuation correction (MRAC) technique. MRAC consists of MR image acquisition, segmentation, truncation compensation (TC), -value assignment, as well as correction for patient table and RF coils. These components have been described last year; this paper focuses on updates of the two most critical steps of MRAC: segmentation and TC. The segmentation algorithm attempts to distinguish 3 biological classes: air, lungs, and soft tissue. It combines an intensity-based region-growing technique with lung-model adaptation. For TC, the following three-step approach to correct for truncation in the MR-based attenuation maps has been developed and investigated: (A) Areas in the attenuation map which are possibly truncated are identified. (B) For these areas, an estimate of the outer patient contour is extracted from a registered PET image which is reconstructed without attenuation correction. (C) Truncation correction areas as extracted from the PET contours are added to attenuation map. The segmentation algorithm was applied to a number of datasets from a large pool of volunteers from multiple MR systems. The algorithm yields expected results except for susceptibility and motion artifacts. While the truncation compensation algorithm works for most cases, the robustness needs to be further improved.
AB - Philips has introduced the world's first whole body sequential PET/MR system. We present the current status of MR-based attenuation correction (MRAC) technique. MRAC consists of MR image acquisition, segmentation, truncation compensation (TC), -value assignment, as well as correction for patient table and RF coils. These components have been described last year; this paper focuses on updates of the two most critical steps of MRAC: segmentation and TC. The segmentation algorithm attempts to distinguish 3 biological classes: air, lungs, and soft tissue. It combines an intensity-based region-growing technique with lung-model adaptation. For TC, the following three-step approach to correct for truncation in the MR-based attenuation maps has been developed and investigated: (A) Areas in the attenuation map which are possibly truncated are identified. (B) For these areas, an estimate of the outer patient contour is extracted from a registered PET image which is reconstructed without attenuation correction. (C) Truncation correction areas as extracted from the PET contours are added to attenuation map. The segmentation algorithm was applied to a number of datasets from a large pool of volunteers from multiple MR systems. The algorithm yields expected results except for susceptibility and motion artifacts. While the truncation compensation algorithm works for most cases, the robustness needs to be further improved.
UR - http://www.scopus.com/inward/record.url?scp=79960333147&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=79960333147&partnerID=8YFLogxK
U2 - 10.1109/NSSMIC.2010.5874153
DO - 10.1109/NSSMIC.2010.5874153
M3 - Conference contribution
AN - SCOPUS:79960333147
SN - 9781424491063
T3 - IEEE Nuclear Science Symposium Conference Record
SP - 2119
EP - 2122
BT - IEEE Nuclear Science Symposuim and Medical Imaging Conference, NSS/MIC 2010
T2 - 2010 IEEE Nuclear Science Symposium, Medical Imaging Conference, NSS/MIC 2010 and 17th International Workshop on Room-Temperature Semiconductor X-ray and Gamma-ray Detectors, RTSD 2010
Y2 - 30 October 2010 through 6 November 2010
ER -