Hongxi Zhang1, Hua Li2, Zhipeng Shen1, Yi Zhang3, and Dan Wu3
1Children's Hospital, Zhejiang University School of Medicine, Hangzhou, China, 2Nemours AI duPont Hospital for Children, Wilmington, DE, United States, 3Key Laboratory for Biomedical Engineering of Ministry of Education, Department of Biomedical Engineering, College of Biomedical Engineering & Instrument Science, Zhejiang University, Hangzhou, China
Synopsis
Diffusion-time dependent diffusion MRI has shown
potential in probing tumor microstructure. This study investigated the
feasibility of time-dependent dMRI to map brain tumor microstructure in a
pediatric population at 3T. Oscillating and pulsed gradient dMRI was performed
to access a series of diffusion times and b-values, and the data were fitted
with the IMPULSED model to estimate cell diameter,
intracellular fraction, and diffusivity metrics. In a pilot study of 17
pediatric brain tumor patients, all high-grade tumors showed higher
intracellular fraction and cellularity than the low-grade ones, while the cell
diameter showed differentiation among different types of high-grade
tumors.
Introduction
Recent progress on diffusion-time
(td) dependent diffusion
MRI 1,2 offers new insights into tissue
microstructure. Particularly, the technique has unique advantages in
characterizing tumor microstructure, as reported in several preclinical studies 3-6. Despite the challenges to achieve
short td on clinical
scanners, recently, the feasibility of td-dependent
dMRI in clinical tumor applications 7-11 has also been explored. In this
study, we investigated the utility of td-dependent
dMRI based microstructural mapping in pediatric brain tumor patients, using
pulsed and oscillating gradients on a clinical 3T system. Methods
Seventeen pediatric
brain tumor patients (aged 8 months to 14 years old, 8 males and 9 females)
were enrolled in the study with IRB approval and parental consent. An oscillating
gradient dMRI (OG-dMRI) sequence with trapezoid-cosine gradients was implemented
on a 3T Phillips Achieva scanner (maximum gradient =80 mT/m). The OG-dMRI was
performed at oscillating frequencies of 50Hz (b=0.37 ms/µm2), 33Hz (b=0.5
ms/µm2), and 17Hz (b=0.5/1/2 ms/µm2), 4 diffusion
directions, TE/TR = 168/3000 ms, FOV=180×180 mm, in-plane resolution = 1.4×1.4
mm, 3 slices with a slice thickness of 8 mm. Pulse gradient dMRI (PG-dMRI) was
performed with σ/Δ = 60/82.8 ms (b=0.5/1/2 ms/µm2), and the other
parameters were kept the same as OG-dMRI. The total protocol was under 5min.
Routine T1-weighted images before and after Gd-enhancement, T2-weighted, FLAIR,
and routine DWI (b=0.6 ms/µm2) were also obtained.
The PG- and OG-dMRI
data were fitted using the imaging microstructural parameters using limited
spectrally edited diffusion (IMPULSED) method 3, assuming a two-compartment model
with impermeable cells. The fitting was performed using least square curve
fitting in MATLAB, and the fitting was repeated 100 times with randomized
initialization to avoid local minimum. Cell diameter (d), intracellular fraction (fin)
and extracellular diffusivity (Dex)
were estimated, while the intracellular diffusivity (Din) was fixed at 1 µm2/ms since the fitting
is insensitive to the choice of Din
7. The cellularity was measured as fin/d. The regions of interest (ROIs) were
defined in the solid region of the tumor based on b0 images.Results
Based on the
pathological examination, among the 17 patients, seven had low-grade glioma
(WHO I and II), two had high-grade glioma (WHO IV), three had medulloblastoma
(WHO IV), two had atypical teratoid/rhabdoid tumor (AT/RT, WHO IV), along with
the other types of tumors (n=1 per category). Figure 1 demonstrated that microstructural
mapping results in four types of pediatric tumors, along with the apparent
diffusivity coefficient (ADC) maps from routine DWI and Gd-enhanced T1w images.
It is evident that the fin
and cellularity were much higher in high-grade glioma, medulloblastoma, and
AT/RT, compared with that in low-grade glioma, while the Gd-enhanced images
could not distinguish high-grade from low-grade tumors in these four cases. The
low and high-grade tumors can be separated based on fin (5.4 folds difference) and cellularity (6.3 folds difference)
(Figure 2). The ADC values from individual PGSE and OGSE (33Hz) scans also
showed significant differences between low and high-grades with relatively
smaller effective sizes (2.1 and 1.7 fold differences, respectively).
When looking into the
specific tumor types, we noticed that the medulloblastoma showed a larger cell
size compared to the other types of high-grade tumors (Figure 3). By examining
the dMRI signals at different td
and b-values (Figure 4), it is observed that the td-dependence was higher in high-grade tumors than the
low-grade ones. In addition, we found a negative relationship between ADC
(PG-dMRI) and fin based on
the scatter plots of individual voxels (Figure 4), which may explain the
diagnostic performance of both metrics. Discussion and Conclusion
This pilot study demonstrated that
td-dependent
dMRI based microstructural mapping of brain tumor was feasible on clinical
systems, and it is capable of tumor grading for typical types of pediatric
tumors. The microstructural information, such as cell size and cellularity, may
provide valuable information about the tumor pathology, e.g., the low fin in low-grade glioma (lower
than normal tissue) indicates loose and less cellular micro-environment with
cyst formations. The added information may also be useful in differentiating different
tumor types, e.g., the medulloblastoma (a type of embryonic tumor) showed a larger
cell size compared to other high-grade tumors. Histological evidence and more
samples are needed to support these findings.
Another limitation lies
in that the IMPULSED model, like the other existing models 4,5, assumed a simple two-compartment configuration,
and permeability was not considered. Therefore, the model may not be
appropriate for normal brain tissue, resulting in underestimation of fin and unstable fitting of
cell size in non-tumor regions. Nevertheless, the model requires
relatively small q-t space samples
and provided useful microstructural characterization of the tumor. Given the
relatively short scan time, the current imaging protocol can be readily incorporated into clinical practice.Acknowledgements
This work was supported by the Natural Science Foundation of China (61801424, 81971606, and 91859201) and the Ministry of Science and Technology of the People’s Republic of China (2018YFE0114600).References
1. Stepisnik
J. Time-Dependent Self-Diffusion by Nmr Spin-Echo. Physica B
1993;183(4):343-350.
2. Gore JC, Xu JZ, Colvin DC, Yankeelov
TE, Parsons EC, Does MD. Characterization of tissue structure at varying length
scales using temporal diffusion spectroscopy. NMR in biomedicine
2010;23(7):745-756.
3. Jiang XY, Li H, Xie JP, Zhao P, Gore
JC, Xu JZ. Quantification of cell size using temporal diffusion spectroscopy.
Magn Reson Med 2016;75(3):1076-1085.
4. Reynaud O, Winters KV, Hoang DM,
Wadghiri YZ, Novikov DS, Kim SG. Pulsed and oscillating gradient MRI for
assessment of cell size and extracellular space (POMACE) in mouse gliomas. NMR
in biomedicine 2016;29(10):1350-1363.
5. Panagiotaki E, Walker-Samuel S, Siow B,
Johnson SP, Rajkumar V, Pedley RB, Lythgoe MF, Alexander DC. Noninvasive
Quantification of Solid Tumor Microstructure Using VERDICT MRI. Cancer Res
2014;74(7):1902-1912.
6. Colvin DC, Yankeelov TE, Does MD, Yue
Z, Quarles C, Gore JC. New insights into tumor microstructure using temporal
diffusion spectroscopy. Cancer Res 2008;68(14):5941-5947.
7. Xu J, Jiang X, Li H, Arlinghaus LR,
McKinley ET, Devan SP, Hardy BM, Xie J, Kang H, Chakravarthy AB, Gore JC.
Magnetic resonance imaging of mean cell size in human breast tumors. arXiv
2019;arXiv:1905.07818.
8. Lemberskiy G, Rosenkrantz AB, Veraart
J, Taneja SS, Novikov DS, Fieremans E. Time-Dependent Diffusion in Prostate
Cancer. Invest Radiol 2017;52(7):405-411.
9. Lima M, Kataoka M, Honda M, Ohno AK,
Ota R, Ohashi A, Urushibata Y, Feiweier T, Toi M, Togashi K. Time-Varying
Diffusion Patterns in Breast Cancer Linked to Prognostic Factors. 2019;
Montreal, QC, Canada.
10. Lemberskiy G, Baete S, Novikov D,
Fieremans E, Zan E, Hu K, Kim S. Diffusion time-dependence of diffusivity and
kurtosis in locally advanced head and neck squamous cell carcinoma before and
after chemo-radiation therapy. 2019; Montreal, QC, Canada.
11. Reynaud O. Time-Dependent Diffusion MRI
in Cancer: Tissue Modeling and Applications. Front Phys 2017;5.