Extension of the MR field-of-view with HUGE for MR-based attenuation correction in integrated PET/MR
Maike E. Lindemann1, Jan Ole Blumhagen2, and Harald H. Quick1,3

1High Field and Hybrid MR Imaging, University Hospital Essen, Essen, Germany, 2Siemens Healthcare GmbH, Erlangen, Germany, 3Erwin L. Hahn Institute for Magnetic Resonance Imaging, University Duisburg-Essen, Essen, Germany


In quantitative PET-imaging, it is essential to correct the attenuation of photons in tissue. In combined PET/MR-imaging the attenuation correction (AC) is based on MR-data and subsequent tissue class segmentation. The MR-FOV is limited due to B0-inhomogeneities and gradient nonlinearities. Therefore, the AC-map is truncated and reconstructed PET-data are biased. HUGE (B0-Homogenization using gradient enhancement), which determines an optimal readout gradient to compensate gradient nonlinearities, is evaluated in phantom experiments and applied to MR-imaging of volunteers. The extension of the MR-FOV for MR-based AC showed an improvement of PET-quantification in integrated PET/MR-imaging by reducing the truncated areas of the AC-map.


To achieve quantitative PET imaging, it is essential to correct the attenuation of photons in human tissue. In combined PET/MR hybrid imaging the attenuation correction (AC) is based on MR data and subsequent tissue class segmentation [1]. In MR imaging, the field-of-view (FOV) is limited to a diameter of typically 50 cm in x-y direction due to B0 inhomogeneities and gradient nonlinearities. Therefore, the AC map (umap) might be truncated or geometrically distorted at the edges of the FOV, and hence the reconstructed PET data may be biased. In this work, a method called HUGE [2] (B0-Homogenization using gradient enhancement), which determines an optimal readout gradient in x-direction to locally compensate the gradient nonlinearities, is evaluated in phantom experiments and applied to MR-imaging of volunteers.

Materials and Methods

All measurements were performed on an integrated whole-body PET/MR system (Biograph mMR, Siemens Healthcare, Erlangen, Germany). For quantitative PET imaging a NEMA IQ phantom (PTW, Freiburg, Germany) was used. To simulate patient arms, two cylindrical structure phantoms (diameter 12.5 cm, length 65 cm) were built. All phantoms were filled with MR-signal-producing fluid and placed on a low-attenuating Styrofoam block to ensure exact and reproducible repositioning (Fig.1). The MR-based AC map was measured with a standard Dixon VIBE sequence. A modified HASTE sequence with continuous table movement was acquired to obtain MR signal from the extended FOV (HUGE) [3]. The HUGE MR data is used to complete the truncated Dixon-AC map by filling the missing parts with a linear attenuation coefficient of 0.1 cm-1. The phantom setup was scanned by a dual-source CT scanner (SOMATOM Definition Flash, Siemens Healthcare) to provide a CT-based umap of all phantom components, serving as reference standard. For converting CT data with an energy level of 140 keV to the PET energy level of 511 keV, a bilinear function was used [4]. All AC maps are shown in Fig. 2. The PET measurements with 18F were accomplished by NU 2-2007 standard with a sphere-background-ratio 8:1, total activity 51.32 MBq and 12 min per bed position [5]. All PET reconstructions were performed with e7 tools (Siemens Molecular Imaging, Knoxville, USA). Five volunteers (mean age 44 y ± 19 y, mean BMI 21.4 ± 2.5) were imaged with Dixon VIBE and HUGE sequences. To evaluate the quantitative effect in phantom measurements of limited and extended AC maps on PET data compared to CT-based AC, ROIs were analyzed in the reconstructed PET images (Fig. 3). Fusion images of Dixon VIBE umap and HUGE MR data of five volunteers were generated (Fig. 4).


The Dixon VIBE umap shows truncations at the edges of the FOV (arms, Fig. 2A), while the CT-based AC, serving as standard of reference, images the entire phantom setup with fluid filling and phantom housing and is not hampered by truncation artifacts (Fig. 2B). In Fig. 2C, the Dixon VIBE-based umap extended by HUGE shows less distortion at the edges of the FOV. The calculated SNR of activity concentration in the hot spheres using different umaps for attenuation correction benefits from extended AC correction, but there is still an underestimation in PET signal when compared to the CT reference (Fig. 3B). Nevertheless, relative differences in counts up to 5.2% in HUGE corrected images in comparison to limited Dixon VIBE AC show an improvement in MR-based AC. The remaining difference to the CT-based activity values can be explained by the fact, that the MR-based AC techniques Dixon and HUGE do not display and consider the PET signal attenuating phantom housing as the CT-based umap does. Volunteer scans (Fig. 4) also show the improvement in extended AC maps. While segmented AC map of the Dixon VIBE sequence shows signal truncations along the arms, MR-based HUGE data and fusion images depict added body volume along the arms by applying HUGE.

Discussion and Conclusion

In phantom measurements the extension of the MR FOV using HUGE for MR-based attenuation correction showed an improvement of PET quantification in integrated PET/MR imaging by reducing the otherwise truncated areas of the phantom umap in MR-based Dixon imaging. The phantom results were in good agreement with the CT reference scan, with the difference that in the CT-based umap also the phantom housing and its attenuating effect is additionally considered. The volunteer measurements demonstrate that the proposed method of using HUGE and image segmentation to complete the MR-based umap can reduce the distortion at off-center position and therefore has the potential to improve MR-based AC in PET/MR hybrid imaging.


No acknowledgement found.


1) Martinez-Möller A, et al. Tissue classification as a potential approach for attenuation correction in whole-body PET/MRI: evaluation with PET/CT data. J Nucl Med. 2009; 50(4):520-526

2) Blumhagen JO, et al. MR-based field-of-view extension in MR/PET: B0 homogenization using gradient enhancement (HUGE). Magn Reson Med. 2013; 70(4):1047-1057

3) Blumhagen JO, et al. Field of view extension and truncation correction for MR-based human attenuation correction in simultaneous MR/PET imaging. Med Phys. 2014; 41(2):022303. doi:10.1118/1.4861097

4) Carney JP, et al. Method for transforming CT images for attenuation correction in PET/CT imaging. Med Phys. 2006; 33(4):976-983

5) Ziegler S, et al. NEMA image quality phantom measurements and attenuation correction in integrated PET/MR hybrid imaging. EJNMMI Phys. 2015; 2(1):18.doi:10.1186/s40658-015-0122-3


Figure1: Phantom setup consisting of a NEMA IQ phantom (1) on a Styrofoam positioning block (2) with two arm structure phantoms (3) and additional MR signal adjustment bottles (4).

Figure2: Combined AC maps consisting of the CT-based hardware umap of the patient table and the phantom umap in axial (left), coronal (middle) and sagittal (right) orientations. The phantom umap is based on the Dixon VIBE sequence (A), CT-based (B) and the extended Dixon VIBE using HUGE information (C).

Figure3: PET image with 6 spheres aligned radially around the lung insert, 4 spheres are injected with activity (hot) and 2 are filled with water (cold) (A). (B) Calculated SNR of tracer activity in hot spheres using different umaps for attenuation correction: Dixon (blue), CT (green) and HUGE (red) AC.

Figure4: Segmented AC maps of the Dixon VIBE sequence showing signal truncations along the arms; MR-based HUGE data from the right volunteer side (for demonstration purposes) and fusion images showing the added body volume along the arm by applying HUGE on five volunteers.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)