MR-based attenuation correction for PET using an ultrashort echo time (UTE) sequence

Vincent Keereman*, Stefaan Vandenberghe, Y. De Deene, Robert Luypaert, Tom Broux, Ignace Lemahieu

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference proceeding contribution

11 Citations (Scopus)


One of the limitations of a multimodality PET/MR scanner is the lack of an attenuation map to correct the PET image for attenuation. Some methods have been developed for deriving the attenuation map from a magnetic resonance image. However, the very low signal intensity of cortical bone in images acquired with conventional MR sequences makes it difficult to distinguish this tissue type from air, while they have strongly different attenuation coefficients. The necessity of detecting cortical bone is investigated by reconstructing simulated data with different attenuation maps. It is shown that the asignment of a soft tissue attenuation coefficient to voxels containing bone causes SUV quantitation errors of up to 20%. When the attenuation coefficient of bone is used a lesion in the spine is even rendered invisible.The feasibility of visualizing cortical bone with an ultrashort echo time (UTE) sequence and of segmenting MR images acquired with such a sequence into bone, soft tissue and air is investigated. the X2 and proton density (relative to the same volume of water) of this tissue type is derived from relaxometry experiments on 5 samples of cortical bone. Both results (T2 = 1.51ms, PD = 29% of the same volume of water) indicate that the visualization of cortical bone with MRI is possible if the signal can be acquired very quickly after the excitation pulse. The head of a pig is used as a phantom for imaging, because the tissue composition resembles very well that of the human head. The MR UTE images are segmented by a simple thresholding technique into bone, soft tissue and air. The resulting segmentation is compared to the segmented CT of the phantom. For all tissue classes 85 % of pixels are assigned to the correct pixel class. Most errors are found between bone and soft tissue, and to a lesser extent between bone and air. The effect of the segmentation errors on the resulting attenuation-corrected PET image still has to be evaluated.

Original languageEnglish
Title of host publicationIEEE Nuclear Science Symposium conference record
EditorsPaul Sellin
Place of PublicationPiscataway, NJ
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
Number of pages6
ISBN (Print)9781424427154
Publication statusPublished - 2008
Externally publishedYes
Event2008 IEEE Nuclear Science Symposium Conference Record, NSS/MIC 2008 - Dresden, Germany
Duration: 19 Oct 200825 Oct 2008


Other2008 IEEE Nuclear Science Symposium Conference Record, NSS/MIC 2008

Fingerprint Dive into the research topics of 'MR-based attenuation correction for PET using an ultrashort echo time (UTE) sequence'. Together they form a unique fingerprint.

Cite this