Eo-Jin Hwang1, Hyun-Seok Choi1, Yoon-Ho Nam1, and Joon-Yong Jung1
1Department of Radiology, St.Mary Hospital, Seoul, Korea, Republic of
Synopsis
The purpose of this study was
to quantitatively evaluate ADCs from two different DWI sequences – echo planar
spin echo (EPSE) and read-out segmented echo planar (RESOLVE) imaging. The mean
ADCs of the whole brain, GM, WM and CSF were compared based on masking
templates generated from T2-weighted images. Our results demonstrated that
although difference between the two ADCs was statistically significant over the whole brain regions, ADCs from msDWI and ssDWI were highly correlated (R2 =
0.989). The relationship between the two ADCs should be
considered to utilize it as an effective quantitative metric.Purpose
The purpose of
this study was to quantitatively evaluate diffusion-weighted signals and ADC
values from two different DWI sequences – echo planar spin echo (EPSE) and read-out
segmented echo planar (RESOLVE) imaging - and to perform a comparative analysis
of the two.
Methods
and Materials
Image
acquisition: The participants included 30 normal subjects with
no medical history of neurological diseases (mean age = 64.4, standard deviation
= 14.2, 15 males and 15 females). MR imaging was performed on a 3T MR system
(Siemens). A single-shot diffusion weighted image (ssDWI) was acquired using an
echo-planar spin echo (EPSE) sequence (TR = 6800ms, TE = 100ms, voxel dimension = 1.15 x 1.15 x 6 mm
3), and a high resolution multi-shot DWI (msDWI) was obtained from a
readout-segmented echo planner imaging (RESOLVE) sequence (TR = 4900ms, TE =
68ms, voxel dimension = 1.15 x 1.15 x 6 mm
3, number of readout segment per image = 11, echo-spacing = 300µs) at b-values of 0 and 1000s/mm
2. The axial T2-weighted
image was also acquired for image registration and segmentation (TR = 4900ms,
TE = 68ms, voxel dimension = 0.47x 0.47 x 6mm
3).
Image
processing: All pre-processing and statistical analysis steps were
performed using a Statistical Parametric Mapping version 8 (SPM8) program and
MATLAB software. The T2-weighted images were co-registered to msDWI at b = 0
s/mm
2 and were segmented into gray matter (GM), white matter (WM)
and cerebrospinal fluid (CSF) to generate masking templates. The calculated ADC
maps, msDWI and ssDWI of b0 and 1000 s/mm
2 were segmented into GM,
WM and CSF using the masking templates, and mean ADC values from voxels
containing more than 90% of GM, WM, CSF and combination of the three were
estimated.
Statistical analysis: A correlation analysis was performed
between the two mean ADC values. The ADC maps of all subjects were spatially
normalized with the resized voxel dimensions of 1.5 x 1.5 x 3mm
3 and
were smoothed for a voxel-based statistical analysis. The full-width at half
maximum (FWHM) of the Gaussian smoothing kernel was chosen to be 4 x 4 x 8 mm
3.
A paired t-test was performed between the two ADC maps of each subject to
identify voxels with significant differences. A false discovery rate (FDR)
correction was performed for those voxels representing significant differences,
and the p-value less than 0.05 was
chosen.
Results
Figure 1 shows the mean ADCs of the whole brain, GM only, WM only and CSF only,
respectively. The mean ADC values from ssDWIs were greater in all components.
The ADC difference between single-shot and multi-shot DWIs was smallest in CSF
and was largest in WM. Figure 2illustrates the brain regions, in which the
differences between ADC values from msDWI and ssDWI were statistically
significant (
p < 0.05, FDR
corrected). As the color indicates, the ADC values from msDWI and ssDWI were
statistically different over the whole brain regions. Figure 3 illustrates the
correlation analysis results between ADCs from msDWI and ssDWI. Our results
showed that the two ADCs were highly correlated (R
2 = 0.989), with
the correlation coefficient of 0.897 and a residual constant of 65.368 given
that ADC from msDWI was a dependent variable.
Discussion and Conclusion
Although a number of studies compared image quality and detectability of
lesions between ssDWI and msDWI [1-3], no study has been conducted to
quantitatively analyze ADCs from the two sequences. As an initial study, we
compared ADCs from normal subjects in different components of the brain and
discovered that although they were significantly different over the whole brain
regions, the two ADCs were highly correlated (R2 = 0.989). A correlation coefficient
and a residual constant were also estimated to find a relationship between
the two ADCs in normal brains. Although msDWI has a superior signal-to-noise ratio than ssDWI does,
it suffers from a long acquisition time, which might not be suitable for certain patients. Evaluating diseased tissues such as stoke infarcts [4] should enable us to compare the feasibility of ADCs calculated from each sequence and possibly suggest a complimentary role of ssDWI to msDWI for characterizing certain types of diseases.
Acknowledgements
No acknowledgement found.References
[1] Tokoro et
al, Eur J Radiol. 2014 83(1):1728-1733
[2] Li et al,
ANR Am J Roentgenol. 2015 Jul 205(1):70-76
[3] Friedli et
al, Magn Reson Imaging. 2015 Jul 33(6):701-8 [4] Morelli et al, Acta Radiol. 2013 Apr 1;54(3):299-306