Zhuopin Sun1, Steven Meikle2,3, and Fernando Calamante1,3,4
1School of Biomedical Engineering, The University of Sydney, Sydney, Australia, 2Faculty of Medicine and Health, The University of Sydney, Sydney, Australia, 3Brain and Mind Centre, The University of Sydney, Sydney, Australia, 4Sydney Imaging, The University of Sydney, Sydney, Australia
Synopsis
Recent advances in hybrid PET-MRI systems enable
simultaneous acquisition of PET and MR data. PET is used to visualize and
measure biochemically-specific metabolic processes, but has limited spatial
resolution and signal-to-noise ratio. Combining diffusion MRI (dMRI) and PET
data, which provide highly complementary information (e.g. structural
connectivity and molecular information), has rarely been exploited previously in
image postprocessing. The proposed CONNectome-based Non-Local
Means (CONN-NLM) exploits synergies between dMRI-derived structural
connectivity and PET intensity information to denoise PET data. This method is
based on the rationale that structurally-connected voxels and voxels with similar
intensity should be highly weighted when smoothing noise.
Introduction
Recent advances in hybrid PET-MRI systems enable
simultaneous acquisition of PET and MR data, providing new opportunities in
research and clinical applications.1 PET is used to visualize and measure biochemically-specific metabolic
processes, but has limited spatial resolution and signal-to-noise ratio.2 Combining diffusion MRI (dMRI) and PET data, which provide highly complementary
information (e.g. structural connectivity and molecular information), has rarely
been exploited in image postprocessing. The proposed CONNectome-based Non-Local
Means (CONN-NLM) filter exploits synergies between dMRI-derived
structural connectivity and PET intensity information to denoise PET data. Motivated
by anatomical-NLM in PET/CT 3 and connectome-based cortical smoothing in EEG, 4 this method is based on the rationale
that structurally-connected voxels and voxels with similar intensity should be highly
weighted when computing the weighted-average. We developed a realistic PET-MRI
simulation framework to test CONN-NLM. Methods
Simulation framework: a novel framework is
proposed (fig. 1A) to simulate realistic PET-MRI data for testing/validating
CONN-NLM. dMRI is simulated using Fiberfox, 5,6 and processed using MRtrix3.7 Based on dMRI structural connectivity
matrix, pairs of highly-connected regions are constructed by dilating fibre
endpoints (to simulate high-intensity connected lesions) and combined with segmented
T1 as input for analytical PET simulation using ASIM; 8 after adding realistic Poisson noise, the resulting sinograms are reconstructed with STIR FBP3DRP 9 to produce the noisy PET images for
optimisation/testing.
CONN-NLM: A novel filtering method which
combines dMRI structural connectivity and PET intensity similarity information
is proposed (fig. 2). The original non-local means (NLM) filter 11
smooths voxel i by the weighted-average for all voxels j in a search
window; here, j is any voxel in the AAL 10 parcellation:
$$NL_{i,j}=\exp\biggl(-\frac{(x_i-x_j)^2}{h^2} \biggr)$$
$$h^2=C\times\sigma_{PET}^2$$
where x denotes intensity value, h the smoothing
strength (proportional to PET noise level, and constant C controls the smoothing
strength): voxels with similar intensity are weighted highest. A hybrid
connectivity measurement A is also introduced as: 4
$$A={\lambda}A_{dist}+A_{loc}$$
$${\lambda}=\frac{\lambda_{dist}}{\lambda_{loc}}=B\times\sigma_{TDI}^2$$
Adist is derived from connectivity to other AAL nodes
and Alocal from the spatial adjacency of voxels within the
same node. λ regulates the distant/local connectivity
balance and is set proportional to track-density image (TDI) 12 variance, to account for the
connectome sensitivity to tracking quality: worse-quality data (reflected as lower
TDI contrast) leads to less reliable connectome, hence distant connectivity is
weighted less. Combining intensity similarity and connectivity:
$$W_{i,j}=NL_{i,j}A=\exp\biggl(-\frac{(x_i-x_j)^2}{h^2} \biggr)({\lambda}A_{dist}+A_{loc})$$
the normalised weighted-average at voxel i becomes:
$$x_{CONN\_NLM_{i}}=\sum_{j}x_{i}W_{i,j}/\sum_{j}W_{i,j}$$Results
Figure 3
shows that CONN-NLM has the capacity to improve the overall PET image quality
in gray matter while enhancing lesion contrast-to-noise ratio. CONN-NLM
demonstrated further improvement relative to NLM without connectivity
information. Figure 4 shows the effect of filtering strength on performance is
consistent across noise levels and lesion patterns. An optimal filter parameter
C can be estimated by inspecting the ratio of h2 to
the overall variance before filtering. Figure 5 explores the effect of distant/local
connectivity ratio, suggesting that the proposed hybrid connectivity model
improves overall performance. Discussion
CONN-NLM has
unique advantages of providing more informative and accurate PET smoothing by
adding complementary structural connectivity information from diffusion MRI. For
example, clinical research has established correlations between properties of
structural tract and tau accumulation measured by PET.13 The CONN-NLM method represents a new avenue to exploit synergies between
diffusion MRI and PET.
Acknowledgements
No acknowledgement found.References
- Shah,
N. (2019). Hybrid MR-PET imaging: systems, methods and applications.
Cambridge: Royal Society of Chemistry.
- Chen,
Z., Jamadar, S. D., Li, S., Sforazzini, F., Baran, J., Ferris, N., … Egan, G.
F. (2018). From simultaneous to synergistic MR-PET brain imaging: A review of
hybrid MR-PET imaging methodologies. Human Brain Mapping, 39, 5126–5144.
- Chan, C., Fulton, R., Barnett, R., Feng, D. D., &
Meikle, S. (2014). Postreconstruction Nonlocal Means Filtering of Whole-Body
PET with an Anatomical Prior. IEEE Transactions on Medical Imaging,
33(3), 636–650.
- D. K. Hammond, B. Scherrer, and S. K. Warfield,
“Cortical graph smoothing: A novel method for exploiting dwi-derived anatomical
brain connectivity to improve EEG source estimation,” IEEE Trans. Med.
Imaging, vol. 32, no. 10, pp. 1952–1963, 2013.
- P. F. Neher, F. B. Laun, B. Stieltjes, and K. H.
Maier-Hein, “Fiberfox: Facilitating the creation of realistic white matter
software phantoms,” Magn. Reson. Med., 2014.
- Maier-Hein, Klaus, Neher, Peter, Houde,
Jean-Christophe, Caruyer, Emmanuel, Daducci, Alessandro, Dyrby, Tim, …
Descoteaux, Maxime. (2015). Tractography Challenge ISMRM 2015 Data [Data set].
- J.
D. Tournier et al., “MRtrix3: A fast, flexible and open software
framework for medical image processing and visualisation,” NeuroImage.
2019.
- B. Elston, C. Comtat, R. Harrison, and P. Kinahan,
“ASIM: An analytic PET simulator,” Monte Carlo Calc. Nucl. Med. Appl.
Diagnostic Imaging, 2012.
- K. Thielemans et al., “STIR: software for
tomographic image reconstruction release 2,” Phys. Med. Biol., vol. 57,
no. 4, pp. 867–883, Feb. 2012.
- Automated
anatomical labelling atlas 3. E.T. Rolls, C.C. Huang, C.-P. Lin, J. Feng, M.
Joliot, 2019.
- A. Buades, B. Coll and J. Morel, "A non-local
algorithm for image denoising," 2005 IEEE Computer Society Conference
on Computer Vision and Pattern Recognition (CVPR'05), San Diego, CA, USA,
2005, pp. 60-65 vol. 2.
- Calamante F, Tournier JD, Jackson GD, Connelly A.
Track-density imaging (TDI): super-resolution white matter imaging using
whole-brain track-density mapping. Neuroimage. 2010 Dec;53(4):1233-43.
- Jacobs
HIL, Hedden T, Schultz AP, Sepulcre J, Perea RD, Amariglio RE, Papp KV, Rentz
DM, Sperling RA, Johnson KA. Structural tract alterations predict downstream
tau accumulation in amyloid-positive older individuals. Nat Neurosci.
2018 Mar;21(3):424-431.