S. Mazdak Abulnaga1,2, Esra Abaci Turk3, Jie Luo4, Justin Solomon1,2, Lawrence L. Wald5,6,7, Elfar Adalsteinsson1,7, Carolina Bibbo8, Julian N. Robinson8, William H. Barth, Jr.9, Drucilla J. Roberts10, P. Ellen Grant3, and Polina Golland1,2
1Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, 2Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA, United States, 3Fetal-Neonatal Neuroimaging and Developmental Science Center, Boston Children's Hospital, Harvard Medical School, Boston, MA, United States, 4School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China, 5Radiology, Harvard Medical School, Boston, MA, United States, 6Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States, 7Institute for Medical Engineering and Science, Massachusetts Institute of Technology, Cambridge, MA, United States, 8Maternal and Fetal Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, United States, 9Maternal and Fetal Medicine, Obstetrics & Gynecology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States, 10Obstetrics and Perinatal Pathology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States
Synopsis
We demonstrate a volumetric mesh-based mapping of the
placenta to a canonical template that resembles the better-known ex vivo shape. Placental shape presents significant
challenges for visualization of the associated signals. No standard framework exists
for visualizing the organ in vivo. Our
approach is to flatten a volumetric mesh that captures subject-specific
placental shape while penalizing local distortion to maintain anatomical
fidelity. The resulting algorithm produces an invertible transformation to the
canonical template. To demonstrate the promise of the proposed approach, we present
visualization of BOLD MRI intensity and oxygenation measures after mapping them
to a flattened placenta template.
Purpose
Monitoring placental function in vivo promises to support pregnancy assessment and to improve
care outcomes. Blood oxygen level dependent (BOLD) magnetic resonance imaging
(MRI) with maternal hyperoxia has been shown recently to provide
characterization of placental function in
vivo.1 We demonstrate an approach for mapping the placenta to
a canonical template to enable intuitive visualization of the placenta embedded
in the 3D MRI scan. The advantage of such visualization was first demonstrated
via surface-based flattening of nested level sets of the placental shape.2
In contrast, we propose a continuous volumetric mapping that simulates a
physical deformation of the organ. This approach enables us to explicitly
control local volumetric distortion in the resulting flatted placenta. Our formulation
naturally accepts a wide range of penalty functions that encourage desired geometric
properties of the mapping.Methods
Subjects: This IRB approved study enrolled seven women with twin
pregnancy (gestational age: 29-34 weeks). Acquisition: MRI
BOLD scans were acquired on a 3T Skyra scanner (Siemens Healthcare, Erlangen,
Germany) using single-shot gradient echo with 3mm isotropic voxels,
TR=5.8-8s, TE=32-38ms, FA=90°. The maternal
oxygenation paradigm included 10min of room air (21% O2), followed
by 10min of 100% O2, followed by another 10min of room air. To
assess the placenta after birth, histopathology was performed by an experienced
placental pathologist blinded to the MRI findings. A subjective score from 1
(mild pathology) to 4 (severe pathology) was given for each placental region.1
Processing: The placenta was manually segmented in the first volume of
each BOLD MRI series. Time-to-plateau (TTP) of the BOLD signal was evaluated at
every voxel within the placental segmentation to quantify the delay in oxygen saturation
from the start of maternal hyperoxia.1 Algorithm: We first generate
a volumetric tetrahedral mesh that accurately represents the segmentation of
the placenta.3 We then map the mesh to a canonical coordinate system
by minimizing a cost function that comprises a data term, which penalizes
deviations of the surface vertices of the mesh from the ellipsoidal template,
and a regularization term, which penalizes the local volume distortion metric.4-6
We employ gradient descent with line search to estimate the optimal deformation
of the mesh.7 Bijectivity is enforced at every step of the
optimization procedure.6 Finally, image values associated with
specific locations within the placenta are transferred into the canonical
coordinate system for visualization (Figure 1). Results and Discussion
Figures
2 and 3 present the mapping results for two subjects. For each subject, we display the image intensity
in the first volume of the BOLD MRI series and the TTP values estimated from
the entire series. Both types of information are shown in the original image
space and in the canonical template space to enable the comparison of
visualization effectiveness. Figure 2 shows a healthy twin pregnancy with both
placenta regions associated with each twin assigned a pathology score of 2
(mild pathology). Figure 3 shows a twin pregnancy where the placenta region
associated with one twin was healthy (score of 1, no pathology) and the
placenta region associated with another twin contained a pathology (score of 4,
severe pathology). We note that TTP values are higher (i.e., longer delay) and more
heterogeneous in the pathological regions of the placenta, which is consistent with
the findings of the previous study.1 Our visualization method brings
forth the spatial pattern of the delay. We also observe differences in the BOLD
image texture between the two regions of the placenta, which is nearly
impossible to appreciate in the original volume.
These results suggest many
potential applications. Enhanced visualization of BOLD signal differences
across the placental volume promises to facilitate segmentation of the placenta
into functionally homogeneous regions. The mapping provides spatial context and
can assist clinicians in localizing regional pathology. Future studies will
assess the utility of the flattened views for detecting placental pathology,
thereby improving monitoring of fetal health. The standardized visualization
framework will facilitate comparisons across time and across subjects, improving
placental health monitoring and statistical analysis.
Conclusion
We demonstrated a method for volumetric mapping of the
placenta to a canonical template that resembles the ex vivo shape of the organ. The mapping is based on deformations of
a volumetric mesh, which represents the in
vivo placental shape, to a canonical template while minimizing local
volumetric distortion. The proposed method improves visualization of image
texture and of derived measures of placental function for visual assessment and
subsequent analysis. Our results highlight spatial variations of oxygen
delivery within the placenta, showing promise for clinical use.Acknowledgements
NIH NIBIB NAC P41EB015902, NIH NICHD U01HD087211.References
1. Luo, J., Turk, E. A., Bibbo, C., Gagoski, B.,
Roberts, D. J., Vangel, M., Tempany-Afdhal, C. M., Barnewolt, C., Estroff, J.,
Palanisamy, A., Barth, W. H., Zera, C., Malpica, N., Golland, P.,
Adalsteinsson, E., Robinson, J. N., & Grant, P. E. In Vivo Quantification
of Placental Insufficiency by BOLD MRI: A Human Study. Scientific
Reports. 2017;7(1):3713.
2. Miao, H., Mistelbauer, G., Karimov, A., Alansary,
A., Davidson, A., Lloyd, D. F., Damodaram, M., Story, L., Hutter, J., Hajnal,
J. V., Rutherford, M., Preim, B., Kainz, B., & Gröller, M. E. Placenta
Maps: In Utero Placental Health Assessment of the Human Fetus. IEEE
Transactions on Visualization and Computer Graphics 2017;23(6):1612-1623.
3. Fang, Q., & Boas, D. A. Tetrahedral Mesh Generation
from Volumetric Binary and Grayscale Images. IEEE International Symposium on Biomedical
Imaging: From Nano to Micro. 2009;ISBI ’09:1142-1145.
4. Sheffer, A., Praun, E., & Rose, K. Mesh Parameterization
Methods and Their Applications. Foundations and Trends® in Computer Graphics
and Vision. 2007;2(2):105-171.
5. Rabinovich, M., Poranne, R., Panozzo, D., &
Sorkine-Hornung, O. Scalable Locally Injective Mappings. ACM Transactions on
Graphics (TOG). 2017;36(2):16.
6. Smith, J., & Schaefer, S. Bijective
parameterization with free boundaries. ACM Transactions on Graphics (TOG).
2015;34(4):70.
7. Wright, S. J., & Nocedal, J. Numerical
optimization. Springer Science. 1999;35(67-68):7.