PET-MR Attenuation Correction using 3D Multi-Echo Dixon-Based UTE Sequence

• Published in: The Journal of nuclear medicine (1978) Vol. 59; p. 652
• Main Authors: Han, Paul, Horng, Debra, Lai, Xiaochun, Fakhri, Georges El, Ouyang, Jinsong, Ma, Chao
• Format: Journal Article
• Language: English
• Published: New York Society of Nuclear Medicine 01.05.2018
• Subjects: 3-D technology; Accuracy; Attenuation coefficients; Background radiation; Bone density; Chemical equilibrium; Computed tomography; Containers; Error correction & detection; Fast Fourier transformations; Fourier transforms; Gelatin; Ground truth; Health care; Image acquisition; Image reconstruction; Inhomogeneity; Medical imaging; Muscles; Organic chemistry; Positron emission; Positron emission tomography; Proton density (concentration); Representations; Scanners; Soft tissues; Tibia; Tissue engineering; Tomography; Germany
• ISSN: 0161-5505; 1535-5667

Objectives: Accurate estimation of PET attenuation coefficients (AC) is a challenging problem in PET/MR1. Current MR-based AC method widely used in the PET/MR scanners segments an MR image into different tissue classes (e.g., air, lung, fat, and soft-tissue) and then assigns a single AC value for each class. This method does not account for intra- and/or inter-subject variations2-5, and can lead to unacceptable biases especially in areas such as bone6. Thus, accurate representation of bone AC with a continuous variation is necessary for reliable PET/MR imaging7-8. In this work, we propose a new MR acquisition method that integrates UTE and multi-echo Dixon in a single acquisition and a new quantification method to represent continuous variation of water, fat and bone AC using MR. The performance of the proposed method is evaluated via a phantom study. Methods A phantom was made by placing a pig tibia containing muscle, fat, and bone inside a plastic container. The container was then filled with 5% gelatin containing radioactive 18F to resemble background soft-tissue. The phantom was first imaged using a whole-body PET-MR scanner (mMR, Siemens Healthcare, Erlangen, Germany). The PET list-mode acquisition was performed for 10 min. MR acquisition was performed using the proposed 3D UTE/multi-echo Dixon sequence (Fig.1A) with the following imaging parameters: resolution = 1.875×1.875×1.875 mm3, TE = 70, 2110, 2810, 3550, 4250, 4990, 5690 μs, and acquisition time = ~1 min. UTE/multi-echo Dixon images were reconstructed using non-uniform fast Fourier transform (NUFFT)9 and SENSE10 reconstruction. The phantom was then imaged using a CT scanner (Biograph 64, Siemens Healthcare, Erlangen, Germany) with tube peak voltage = 120 kVp and tube current time product 30 = mAs. The CT images were reconstructed with in-plane resolution = 0.56×0.56 mm2 and slice-thickness = 1 mm. Water, fat, and bone signals were estimated using the signal model shown in Fig.1B. I(TEp) denote the image from the pth echo time with p = 1,⋯, 7, ρB, ρW, and ρF each denote the proton density signals of bone, water and fat, respectively, R2[asterisk] denote the single representative relaxation rate of both water and fat, Δffat denote the chemical shift of main-peak fat with respect to water, Δfoffset denote frequency offset to account for different fat components, and ΔfB0 denotes the frequency offset due to magnetic field inhomogeneity (B0). The pseudo-CT map was generated using the model shown in Fig.1C. μmapsoftdenotes the AC map of soft tissue only, FF denotes fat fraction, LACwater denotes the linear AC (LAC) of water (0.100 cm-1 at 511 keV5), LACfat denotes the LAC of fat (0.086 cm-1 at 511 keV5), μmapbonedenotes the AC map containing bone, BF denotes bone fraction, and LACbone denotes the linear AC (LAC) of bone (0.172 cm-1 at 511 keV5). The AC maps obtained by the proposed method were compared with those obtained by the two-point Dixon method implemented by the manufacturer and those from the CT scan as the ground truth using Pearson’s linear correlation. Results Noticeable swap of fat and water signals were observed from the water and fat images obtained from the scanner’s method (Fig.1D). Water, fat and bone maps were successfully obtained using the proposed method (Fig.1D). Continuous bone AC was successfully obtained using the proposed method over the bone region (Fig.1E). The pseudo-CT map generated from the proposed method showed visually similar representation over fat (yellow arrowhead) and bone regions (red arrows), showing higher correlation (r=0.90) with the CT map compared to the pseudo-CT map generated from the scanner’s method (r=0.85). Conclusion The proposed method can be used to obtain continuous water, fat and bone AC using MR and potentially can be used to generate subject-specific AC maps in PET/MR applications.