Attenuation & Motion Correction Strategies for PET Using PET/MRI
Youngho Seo1

1Department of Radiology and Biomedical Imaging, University of California, San Francisco

Synopsis

Attenuation correction (AC) for PET reconstruction using PET/MRI is not trivial. MRI-based segmentation, atlas registration, and time-of-flight derived attenuation coefficients are most commonly applied and studied approaches for AC. Motion management of PET reconstruction, aided by the temporal resolution of MR-based signal, is promising, but still requires extensive investigation and robust method development.

Target Audience

Scientists and physician-scientists who are new to or have been using PET/MRI

Outcome/Objectives

The audience will have a chance to review the current technologies and technology development areas of attenuation and motion correction methods for PET using PET/MRI.

Purpose

1) To review the currently studied methodologies for attenuation and motion management for PET reconstruction using PET/MRI, and 2) To identify areas for improvement of attenuation and motion correction methodologies

Methods

PET attenuation correction requires an attenuation map that is basically an anatomical image with voxel values that are equal to linear attenuation coefficients of 511 keV photons. For PET or PET/CT imaging, the attenuation map is derived from transmission scans using either radioactive sources or x-ray CT. The most physically accurate attenuation map is based on transmission scan using a radioactive source that emits the same 511 keV photons such as Ge-68 (1). X-ray CT transmission scan also has shown its accuracy in deriving the attenuation map for PET reconstruction, despite the polychromatic nature of x-ray photon energies from conventional x-ray tubes (2).

In PET/MRI, particularly in current commercial implementations, there is no radioactive source or x-ray CT. Thus, attenuation map derivation should rely on other available information either from MRI or other PET-related data (3-5). The most basic method is to derive water and fat from Dixon MRI. Water, fat, air, and lungs then are segmented, and pre-defined linear attenuation coefficients are assigned to each tissue class. Unlike water and fat, air and lungs are derived by defining anatomical boundaries since their locations are distinct from other tissue/material classes. This method has a great advantage of convenience; however, there have been concerns for its accuracy that affects quantification of PET radiopharmaceutical uptake. In addition to the fixed pre-defined linear attenuation coefficients which are not true values and ignore both intra- and inter-patient variations, excluding highly attenuating material such as cortical bone creates a higher level of inaccuracy in volumes that contain a large amount of bone. In order to circumvent this issue, a CT-based atlas approach has been used instead of MRI-derived attenuation map (6). For example, the volume that contains the brain is surrounded by the skull, and the attenuation correction excluding bone severely underestimates radiopharmaceutical uptake values in PET reconstruction (4). A population-average CT-based atlas for attenuation map creation that involves nonrigid registration of the CT atlas to patient-specific MR images could be applied for attenuation correction in this case.

Since the compensation of bone in attenuation correction of PET reconstruction is key to improving the quantitative accuracy of PET radiopharmaceutical uptake measurement, other MRI-based methods that can visualize cortical bone better than conventional T1 or T2 weighted MR images have emerged as a promising solution. Both ultrashort echo time (UTE) and zero echo time (ZTE) MRI sequences have been implemented in adding bone to MRI-based attenuation map, and have shown a great promise in improving the accuracy of attenuation correction (7-9). While UTE/ZTE methods are fairly mature for the volume that contains the brain, extension of UTE/ZTE type of bone compensation to other parts of the body is still an area that requires more extensive evaluation. In addition to using UTE/ZTE methods, the CT atlas method also has its promise to add bone into the attenuation map in other body parts than the brain.

More recently, researchers found a way to derive attenuation sinogram purely based on PET data when time-of-flight (TOF) information of 511 keV annihilating photons is recorded (10-14). In PET/MRI, this method has a clear advantage over other methods since there needs no general assumption of tissue homogeneity when the pre-defined attenuation coefficients are assigned by tissue classification. This algorithm however requires a TOF-capable PET/MRI system, and still needs much fine-tuning to function robustly and appropriately in the clinical setting.

Regarding motion management, there has been a healthy hype that good temporal resolution of MR signal could help significantly motion-corrected PET reconstruction. Unfortunately, this hype has not been fully realized primarily because the durations of MRI sequence do not completely overlap with those of PET acquisition. However, traditional methods of motion management such as navigator and bellows are still very useful, and are applied in PET reconstruction. Time-resolved 4D MR-based attenuation correction for thorax and abdomen where respiration creates geometric blurring in PET signal collection also requires appropriate motion-field derivation with an assumption of consistent breathing.

Results

Figure 1 shows the workflow of how MRI-based or atlas-based attenuation map is generated for the head using PET/MRI. In our analysis, both atlas-based attenuation correction (AC) and ZTE-based AC perform well in comparison to the reference standard of CT-based AC (Figure 2). However, ZTE-based AC shows less spatial variability than the atlas-based AC.

Our investigation of extending the ZTE (or UTE) based AC method to other body volumes that contain a large amount of bone such as the pelvis indicated the importance of the bone compensation in attenuation map generation. As shown in Figure 3, without the bone compensation, as in the brain, the lesions in the pelvis suffer significant underestimation of PET uptake values.

In case for TOF-capable PET/MRI, attenuation correction could be performed from the PET TOF information. As shown in Figure 4 which was produced by our colleagues at Massachusetts General Hospital, in addition to the TOF information, MR images as anatomical reference could be also used as an anatomic prior to improve the accuracy of the attenuation correction.

As previously stated, the motion field extraction from MR signal can improve significantly if the motion information can be obtained for often longer contiguous PET acquisition than a typical MRI sequence. This can be achieved by combining both navigator embedding in allowed MRI sequences and data-derived motion field from MR images. Figure 5 shows a 10-minute long uninterrupted navigator run and derived respiratory waveform and triggers obtained by our industry partner.

Discussion

The goal of achieving the true accuracy of PET AC is somewhat elusive using PET/MRI since there is no clear robust strategy. However, most of the implemented and being implemented algorithms do provide mostly adequate accuracy, which is most likely suitable for diagnostic imaging. As the data indicated, close attention to how to compensate bone in AC is essential when an image-derived attenuation map is used for reconstruction. There could be more aggressive algorithm implementations in this area, by the way. For example, when the PET TOF data are used for AC, image reconstruction becomes nontrivial since the algorithm is not yet mature, and still needs more fine-tuning. Motion is something that needs to be managed for PET no matter what, and there needs much effort to take advantage of the simultaneity of PET and MRI acquisitions and other clever ideas.

Acknowledgements

I appreciate my collaborators (Peder Larson, Jaewon Yang, Andrew Leynes, and Quanzheng Li) for sharing their data and insightful discussions.

References

1. Abella M, Alessio AM, Mankoff DA, MacDonald LR, Vaquero JJ, Desco M, Kinahan PE. Accuracy of CT-based attenuation correction in PET/CT bone imaging. Phys Med Biol. 2012;57:2477-2490.

2. Kinahan PE, Hasegawa BH, Beyer T. X-ray-based attenuation correction for positron emission tomography/computed tomography scanners. Semin Nucl Med. 2003;33:166-179.

3. Mehranian A, Arabi H, Zaidi H. Quantitative analysis of MRI-guided attenuation correction techniques in time-of-flight brain PET/MRI. Neuroimage. 2016.

4. Aznar MC, Sersar R, Saabye J, Ladefoged CN, Andersen FL, Rasmussen JH, Lofgren J, Beyer T. Whole-body PET/MRI: the effect of bone attenuation during MR-based attenuation correction in oncology imaging. Eur J Radiol. 2014;83:1177-1183.

5. Wollenweber SD, Ambwani S, Lonn AHR, et al. Comparison of 4-class and continuous fat/water methods for whole-body, MR-based PET attenuation correction. 2012 IEEE Nuclear Science Symposium Conference Record. 2012:M15-49.

6. Wollenweber SD, Ambwani S, Lonn AHR, et al. Evaluation of an atlas-based PET head attenuation correction using PET/CT & MR patient data. 2012 IEEE Nuclear Science Symposium Conference Record. 2012:M23-22.

7. Roy S, Wang WT, Carass A, Prince JL, Butman JA, Pham DL. PET attenuation correction using synthetic CT from ultrashort echo-time MR imaging. J Nucl Med. 2014;55:2071-2077.

8. Keereman V, Fierens Y, Broux T, De Deene Y, Lonneux M, Vandenberghe S. MRI-based attenuation correction for PET/MRI using ultrashort echo time sequences. Journal of nuclear medicine : official publication, Society of Nuclear Medicine. 2010;51:812-818.

9. Delso G, Wiesinger F, Sacolick LI, Kaushik SS, Shanbhag DD, Hullner M, Veit-Haibach P. Clinical evaluation of zero-echo-time MR imaging for the segmentation of the skull. J Nucl Med. 2015;56:417-422.

10. Li Y, Defrise M, Metzler SD, Matej S. Transmission-less attenuation estimation from time-of-flight PET histo-images using consistency equations. Phys Med Biol. 2015;60:6563-6583.

11. Mehranian A, Zaidi H. Emission-based estimation of lung attenuation coefficients for attenuation correction in time-of-flight PET/MR. Phys Med Biol. 2015;60:4813-4833.

12. Defrise M, Rezaei A, Nuyts J. Transmission-less attenuation correction in time-of-flight PET: analysis of a discrete iterative algorithm. Phys Med Biol. 2014;59:1073-1095.

13. Rezaei A, Defrise M, Bal G, Michel C, Conti M, Watson C, Nuyts J. Simultaneous reconstruction of activity and attenuation in time-of-flight PET. IEEE Trans Med Imaging. 2012;31:2224-2233.

14. Defrise M, Rezaei A, Nuyts J. Time-of-flight PET data determine the attenuation sinogram up to a constant. Phys Med Biol. 2012;57:885-899.

Figures

The flow chart to apply MR-based attenuation correction in PET/MRI

Voxel-by-voxel comparison between CT-based attenuation correction (AC) and either atlas-based AC or ZTE-based AC

Preliminary investigation of PET reconstruction using the current commercially-available MRAC that does not include bone and a ZTE-enhanced MRAC with bone classification (‘hybrid’ method). (left) Derived attenuation maps. (middle) Reconstructed PET images, and subtraction between the two. (right) Tumor SUVmax comparisons.

Joint estimation using a MAP method and MRI as anatomical prior. The results of estimated attenuation and attenuation bias (middle and right columns) are shown in comparison to the true activity map and attenuation sinogram. Timing resolution of 500 ps for TOF PET is assumed for the simulated raw data for reconstruction.

10-minute long navigator run and derived respiratory waveform and triggers.



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