2
Wen-Yih Isaac Tseng, MD, PhD
December 1, 2005
3T MRI at the National Taiwan University Hospital (NTUH) was installed in November 2003
under the auspices of the Ministry of Education (MOE) and NTUH. The mission of this
research-dedicated whole-body MR scanner is to develop and apply novel MRI techniques to
understand and treat the disease of body and mind. This MRI system is not only used by the
Project for Promoting Academic Excellence of Universities sponsored by MOE, it also opens to
researchers in the hospital and campus. In the past two years, approximately twenty-five funded
research projects were carried out at this system by twenty principal investigators, mostly on the
clinical research in neurological and cardiovascular disease. The achievement is attributed to the
unique role NTUH assigned to this 3T MRI system; it is hospital-based yet dedicated to research
use. This unique role sets the first example in Taiwan and is scarcely seen worldwide.
Although it is a whole-body MR imager, the system is designed not merely for routine clinical
use. High gradient strength, 40 mT/m, and high slew rate, 200 mT/m/s, enable the system to
perform advanced neuroimaging such as functional MRI (fMRI), perfusion and diffusion MRI
that requires ultrafast data acquisition speed. In addition, the system is equipped with several
state-of-the-art features that are vital to high-quality neuroimaging research. First, the system has
a parallel imaging technology including eight-channel array head coil, 1 MB/sec receiver
bandwidth for each channel, and comprehensive reconstruction algorithms in both image space
and k space. This feature can effectively reduce notorious susceptibility artifacts such as image
distortion and signal drop out in echo planar imaging. Second, the system has a navigator echo
technology that monitors 3D motion of the head and corrects for the motion by adjusting the
position and orientation of the slice-select gradients in real-time fashion. This feature is
particularly useful in high resolution fMRI in which motion of less than tens of microns is
required. Third, the system has a function to perform t-test statistics and render BOLD activation
maps on-line. This function provides a way to control the quality of fMRI data and ensure the
success of each fMRI study.
Besides fMRI, NTUH MRI lab has a core imaging technique called diffusion spectrum imaging
(DSI). DSI is the most advanced diffusion MRI technique in the world designed to resolve
3
complex axonal fiber tracts. Our lab has established this technique through validation,
optimization, quantification, visualization and finally application. Implementation of DSI
technique in the clinical setting is tremendously difficult. It entails transferring an experimental
technology to a robust technology that meets the clinical standards. Currently, we have
successfully applied DSI to patients with various brain diseases. The total time for data
acquisition, transfer, reconstruction and tract display is about two hours. This technique is
beginning to show its potential in pre-surgical planning, prediction of post-surgical outcome and
monitoring of disease progress. Combining fMRI and DSI further allows us to study gray matter
function and white matter structure, which is potentially important to the understanding of
pathophysiology of many neurological or psychiatric diseases.
Since the installment of MRI system in November 2003, we have furnished the system with
various facilities up to the standard of a world-class fMRI research lab. Listed below is an
overview of the dates and facilities installed in the past two years. Currently this MRI system is
capable of performing fMRI study with various programmed paradigms involving visual,
auditory, motor and sensory stimulations. It is also equipped with physiological monitor system,
synchronization system, response recorder, eye tracking system, and custom-made eye gargles.
The infrastructure of such scale ensures reliability of data produced from this MRI system.
3 Tesla MRI Laboratory Time Line
Dates Events
2003/07/03 ~ 07/18 shielding installation
2003/07/21 magnet installation
2003/07/18 ~ 08/03 exam room construction
2003/08/04 ~ 08/22 system start-up test
2003/08/25 system function approval
2003/09/17 ~ 09/30 system training
2003/10/1 vision system installation
2003/10/14 final system approval
2003/10/31 open house
2004/05 15 x 10 cm Surface coil
2004/07 Custom-built lenses & gargles
2004/08 TTL synchronization circuit
2004/09 Eye tracking system
2005/07 Physiological monitor system
2005/10 Auditory stimulation system
2005/10 Heat stimulation system
In summary, we spent two years to furnish and fine tune the 3T MRI system. Now, the system is
in its best shape and is ready to fly in full speed. We have set up a core team and consultation
network to provide technical and knowledge supports to users. Most importantly, a group of
researchers have gone through the learning curve and became competent users. I strongly believe
that with the ground works been laid out, there will be a big harvest in the coming years. Finally,
I am deeply indebted to Professor Lee, Yuan-Teh and Professor Hsu, Su-Ming. Without their
strong support this project would not have advanced so successfully.
5
15 x 10 cm Surface coil
In Vivo physiological monitor system
PATHWAY Pain & Sensory Evaluation System
PST Serial Response Box
Nordic NeuroLab Sync Box
Eye Tracking Camera System
Nordic NeuroLab Audio System
6
Assist with forming imaging protocol and result interpretation.
2. Scanning operation and safety training course
MRI scanning service by request.
3. SPM instruction and fMRI data analysis
Assistance and short training course for SPM fMRI image processing and analysis.
4. DSI/DTI data reconstruction and analysis
DSI/DTI image acquisition and tractography reconstruction and visualization.
5. System maintenance
6. Web-based management system
7
Functional MRI and diffusion MRI training course, NTUH
The purpose of the course was to familiarize trainees with the principles of MRI, hands-on
experience in MRI experiments and image data analysis methods.
October 31, 2003
Open house of NTUH 3T MRI. Chancellor Li of NTUH
and colleagues from Neurology, Psychiatric, Rehabilitation
and Radiology departments gathered in the MRI scan room
and listened to Dr. Tseng’s introduction to the facilities and
potential applications of 3T MRI.
February 7, 2004
2004 International Symposium on Integrated Neuroscience and Technology, NTUH
The speakers came from the USA, EU and Taiwan, specialized on multi-modality integrated
research in neuroscience including MEG, EEG, ERP, TMS and MRI. Group meetings and Lab
visits were scheduled during their stay in Taipei to facilitate international collaboration.
June 7, 2004
Press conference, NTU
Our lab announced a novel diffusion MRI technique to unravel the white matter tracts in human
brain.
8
Siemens MRI IDEA programming training course, Carey, North Carolina, USA
Three graduate students Li-Wei Kuo, Hsu-Hsia Peng and Wen-Yang Chiang received a one-week
training course of sequence programming.
August 28 ~ 31, 2005
Co-sponsored with the Program for Promoting Academic Excellence of Universities, we invited
distinguished scientists specialized in high-field microscopic MRI from USA, EU and Israel.
September 23, 2005
The users were six Principal Investigators funded by
NSC in 2005. The research topics included fMRI
study on schizophrenia, autism, oppressive
compulsive disorder, comprehension auditory
set-up.
October 5, 2005
Professor Van Wedeen from Harvard Medical School visited Taiwan and had lab meeting with
students and video conference with research collaborators in Japan.
9
Associate Professor
Taiwan University College of Medicine
System Manager
Graduate Students
Jun-Cheng Weng
University
University
University
10
University
University
University
University
PhD Candidate, Institute of Biomedical Engineering, National Taiwan
University
11
Hospital
University College of Medicine
Hospital
University College of Medicine
Li-Chieh Chen
Su-Chun Huang
12
Fang-Chen Yeh
Administration Assistants
Ya-Chi Wu
Yi-Hwa Fang
Past Graduated Students
Chin-Po Lin, PhD
Jyh-Ray Chen, BE
Hsiou-Ping Lee, BE
Yuan-hung Wu, BE
Department of Neurology
Chen CC, MD
Lu CJ, MD
Yang CC, MD
Department of Surgery
Tsai, MD, PhD
Tseng HM, MD
Yu HY, MD
Lee, MD, PhD
Department of Ophthalmology
Department of Rehabilitation
Lin LI, PhD
Department of Psychology
Chen CC, PhD
Yao C, PhD
1 Optimization of Diffusion Spectrum Magnetic Resonance Imaging for Clinical Scanner
2 Using Track Similarity to Determine Optimum Sequence Parameters for Diffusion
Spectrum Imaging
4 Diffusion Spectrum Imaging of Body-Center-Cubic Sampling Lattice in Q-Space
5 Magnetic Resonance Diffusion Diffractogram in the Assessment of Microstructure
Sizes of Rat Corpus Callosum during Brain Maturation
6 Assessment of Myocardial Perfusion Reserve in Patients with Ischemic Heart Disease
on 3 Tesla MRI
First-pass Contrast-Enhanced MRI
9 Simultaneous Monitoring of Temperature and Magnetization Transfer for HIFU
Treatment
16
J-C. Weng 1 , V. J. Wedeen
2 , J-H. Chen
University, Taipei, Taiwan, 2 MGH Martinos Center for Biomedical Imaging, Harvard Medical
School, Charlestown, MA, United States, 3 Center for Optoelectronic Biomedicine, National
Taiwan University College of Medicine, Taipei, Taiwan
Abstract
Diffusion spectrum imaging (DSI) has been proposed to define tract orientations of neural
fibers [1]. The technique typically requires large gradient pulses that are not attainable in clinical
scanners. By using longer and weaker gradient pulses in the clinical scanner, DSI images would
be subject to truncation artifacts and produce unstable results. The goal of this study is to
determine the optimum parameters of DSI in clinical scanners. We proposed a method to
calculate the reproducibility of the DSI vectors in the areas with different complexity. Optimum
parameters, including cutoff b value, sampling number in q-space, and number of excitation
(nex) were determined. Using the optimum parameters in an echo planar imaging (EPI)
sequence, we obtained the whole human brain scan with spatial resolution of 2.3 x 2.3 x 3 mm 3
in 20 minutes. Our results showed that DSI performed in the clinical scanners produced
reasonable tract orientations in both simple and complex white matter regions.
Introduction
DSI technique probes probability density function (PDF) of the water molecular
diffusion by sampling diffusion weighted images (DWI) in 3-dimensional space of
spatial modulation q. The technique typically requires large gradient pulses that are not
attainable in clinical scanners. By using longer and weaker gradient pulses in the clinical
scanner, DSI images would be subject to truncation artifacts and produce unstable
results. To apply this technique to clinical study, it is necessary to investigate the
feasibility of DSI in clinical scanners. Therefore, the goal of this study is to determine
the optimum parameters, including cutoff value of diffusion sensitivity b, sampling
number in q-space, and number of excitation (nex), and obtain DSI with optimum
parameters in a clinical scanner. Specifically, the effects of different cutoff b values,
different sampling density in the q-space with the same cutoff b value combined with
different nex’s to attain the same scan time were studied.
17
Methods
Since it has been validated that DSI can detect fiber crossing, which is inaccessible to
conventional diffusion tensor imaging (DTI) [2]. The reproducibility of the DSI primary
and secondary vectors can be used as an index to determine the optimum parameters.
We assume that there is no head motion in the subject during twice continuous scans.
The deviation angles in the areas with different diffusion complexity between two scans
can be measured. Then we used the standard deviation of the deviation angle
distribution to quantify the bias produced by the noise or artifact when using different
parameters. The diffusion complexity [3] is an index that can distinguish the anisotropy
from Gaussiantity of the water molecular diffusion between different tissues. When the
distances of DSI primary and secondary vectors differ significantly, the primary vector
may correspond to the fiber orientation and the secondary vector may be produced by
noise or artifact. We compare each of them with themselves to get the deviation angle
distribution. If the DSI primary and secondary vectors have similar distances, both
vectors may correspond to individual fiber orientations. We used cross comparison to
determine the deviation angle distribution.
The data were acquired using 1.5T Sonata system (Siemens, Erlangen, Germany). In
the study of the optimization of cutoff b-value, we used a spin echo diffusion weighted
EPI sequence with 203 diffusion-encodings to obtain DSI of human brain. The images
were acquired with TR/TE = 500/140~160 ms, and number of excitation (nex) =2. The
cutoff diffusion sensitivities (bmax) in DSI were incremented from 3000 to 10000
mTm -1
. 8 sets of images with spatial resolution of 4.2 x 4.2 x 5 mm 3 were obtained in
about 54 minutes.
In the studies of the optimization of number of sampling in q-space and nex, the
images were acquired with TR/TE = 500/145 ms, the cutoff diffusion sensitivities
(bmax) = 4000 mTm -1
. The images of DSI were acquired with 203, 515, 925, 2109
diffusion-encodings. These encodings are comprised of isotropic 3D grid points in the
q-space contained within a spherical volume of 3.7, 5, 6, 8 radial increments. For fair
comparison, the nex = 10, 4, 2, 1 were used, respectively. The total scan time for DSI
with spatial resolution of 2.3 x 2.3 x 3 mm 3
was about 67 minutes.
18
Results
In the result (Fig. 1) of optimization of cutoff b value, the minimum standard
deviation of the deviation angles can be found to be 16.8 o in the low diffusion
complexity area, i.e. white matter, in the cutoff b value = 4000 s/mm 2 . The optimum
cutoff b-value can’t be estimated in the high complexity area, i.e. gray matter, may
caused by the partial volume effect duo to slice thickness.
In the study of the optimization of sampling number in q-space, we used the sampling
points of 2109 as reference standard, and compare DSI primary and secondary vectors
of each sampling points, 203, 515, 925, with 2109 in gray and white matter regions (Fig.
2). Our results showed that there is no significant difference given that the sampling
intervals met the Nyquist criteria. Considering the total scan time, the 203 and 515
sampling points are better choices than 925.
In the study of the influence using different nex, we used the maximum nex as
reference standard. We compared DSI primary and secondary vectors of nine nex
increments, namely, 1 to 9, with those of nex=10 in gray and white matter regions (Fig.
3). As expected, the deviation angle became smaller as the nex increased.
Using the optimum parameters in an echo planar imaging (EPI) sequence, we
obtained the whole human brain scan with spatial resolution of 2.3 x 2.3 x 3 mm 3 in 20
minutes (Fig. 4). The results showed reasonable tract orientations in both simple and
complex white matters.
2 .
Fig.3 The influence of nex.
19
Conclusions
We proposed a method to establish the optimum parameters of DSI. The effects of
different parameters of imaging were also discussed. Using the optimum parameters in an
echo planar imaging (EPI) sequence, the whole human brain scan with high spatial
resolution can be obtained in 20 minutes. Our results showed that DSI was feasible in
clinical scanners, and might be potentially useful in neural science research and clinical
application. References
[1] Van J. Wedeen, ISMRM, Denvor, USA, 2000, 82.
[2] Ching-Po Lin, et al., NeuroImage 2003, 19: 482–495. [3] Li-Wei Kuo, et al., ISMRM, Toronto, Canada, 2003, 592.
Fig.4 DSI result.
Using Track Similarity to Determine Optimum Sequence Parameters for Diffusion Spectrum Imaging
L-W. Kuo
University, Taipei, Taiwan, 2 MGH Martinos Center for Biomedical Imaging, Harvard
Medical School, Charlestown, MA, United States, 3 Center for Optoelectronic Biomedicine,
National Taiwan University College of Medicine, Taipei, Taiwan
Abstract
lower maximal diffusion-encoding sensitivity (bmax) are needed to reduce scan time and
keep the hardware stable. However, aliasing or truncation artifact may become prominent
when 3D Fourier transform of the q-space spectral data is performed. The purpose of this
study was to determine the optimum sampling scheme for clinical DSI. We used 925
encoding gradients with bmax = 9000 s/mm 2
as the gold standard. Similarity of
tractography generated from different sampling schemes to the tractography generated
from the gold standard was compared. Using 515 diffusion-encoding gradients and bmax =
6000 s/mm 2 was considered to be the optimum condition for clinical DSI.
Introduction
It has been shown that diffusion spectrum imaging (DSI) has the capacity to resolve
complex fiber orientations [1,2], and that tractography based on DSI data can differentiate
crossing neural fiber tracts [2,3]. In the clinical setting, however, the maximal
diffusion-encoding sensitivity (bmax) and diffusion-encoding gradients are often reduced
to keep the hardware stable and scan time acceptable. In this case, 3D Fourier transform of
the spectral signal in the q-space is liable to aliasing and truncation artifact, resulting in
inaccurate tractography. The purpose of this study, therefore, was to determine the
optimum number of diffusion-encoding gradients and the value of bmax for clinical DSI.
Tractography results generated from different reduced sampling schemes were compared
with the tractography generated from the gold standard. The degree of similarity in
tractography was quantified in terms of a similarity index.
21
Materials and Methods
DSI data were obtained from one healthy volunteer with a 3T MRI system (Trio,
Siemens, Erlangen, Germany). A twice-refocused balanced echo diffusion EPI sequence
was used to acquire MR diffusion images. Images of DSI were acquired with 925
diffusion-encodings comprising isotropic 3D grid points over the q-space. DSI data were
acquired with bmax = 9000 s/mm 2 , and TR/TE = 2900/150 ms. In order to implement the
tractography, isotropic voxels were obtained by setting in-plane resolution and slice
thickness to be 2.7 mm. Fifteen horizontal slices encompassing the middle portion of the
brain were acquired. The scan time was less than one hour. DSI analysis was based on the
relationship that the echo signal S(q) and the diffusion probability density function P(r)
were a Fourier pair, i.e., S(q) = FT{P(r)} [4]. The integration of P(r) r 2 along each radial
direction was used to calculate the orientation density function (ODF). The main
orientations of the water diffusion were then determined by the local maximum vectors of
the ODF [5]. The fiber tracking was based on an automated algorithm that was adapted for
DSI data. In the selected seed points, the first three largest ODF vectors of each voxel were
used to track. All fiber orientations of the nearest voxels were used to decide the
proceeding orientation for the next step. The most coincident orientation which was less
than 22° was chosen. The proceeding length of each step was 0.5 voxel. The tracking
would stop if there was no coincident orientation in the nearest voxels.
Two different approaches of reduced sampling were studied. The datasets were
generated by sub-sampling the original dataset. The first approach consisted of
sub-sampled datasets with the same bmax but different reduced samples, the second
approach consisted of subsets with the same q but different reduced bmax. Tractography
with the same seed points was reconstructed from each sampling scheme. The results were
compared with the standard tractography generated from the original dataset. To define the
track similarity, each track was separated into small segments based on the tracking steps.
A unit vector was assigned to each segment according to the orientation of the proceeding
orientation. The similarity index of the unit vector v1 at each step position p1 with respect
to its corresponding vector v2 at p2 in the standard tractography was defined as the inner
product of v1 and v2 divided by the distance between their step positions. The track
similarity was obtained by averaging the similarity indices over all the segments in a track.
If the length of the two compared tracks differed by more than 20 steps, the similarity was
22
set to 0. Track similarity for each seed point was calculated first and the mean similarity
over the total tracks was obtained. Results
The sampling points in q-space of the original experimental data was shown in Fig. 1, a,
the sub-sampled schemes were also shown in b, c, d, e, f and g. Tractography from the
original dataset was used as the gold standard (Fig. 2). Six tractography results were
generated from the six sub-sampled datasets. The track numbers generated from each
dataset were 670, 679, 595, 699, 689 and 660 for b, c, d, e. f and g, respectively, and the
mean similarities with respect to the standard tractography were 9.00, 7.78, 4.68, 7.86, 7.16
and 3.47 (x 10 8 ), respectively. Comparison between schemes b and e, as well as schemes d
and g indicated that schemes with higher bmax were relatively better than the schemes with
reduced bmax (p = 0.009 and 0.002). However, the difference was not significantly
different between schemes c and f (p = 0.12). Among the schemes using a fixed bmax =
9000 s/mm2, similarity increased with the gradient number (Fig. 3). Among the schemes
with a fixed q and reduced bmax, both e and f had better track similarity than scheme g,
but there was no significant difference between schemes e and f (p=0.066).
Figure 1. The gold standard is the 925 grid sampling points over the q-space with bmax = 9000 s/mm2 (a). Schemes b, c and d are the datasets with the same bmax as a, and their sampling numbers in q-space are 691, 515 and 203, respectively. Schemes e, f and g are the interpolated datasets with the same q as a. Their sampling numbers are 691, 515 and 203, and bmax are 7250, 6250 and 3250 s/mm2, respectively.
Figure 2. Tractography from the standard DSI dataset with 925 diffusion-encoding gradients and bmax = 9000 s/mm2 (a in Fig. 2). A total of667 tracks were generated from 253 seed points. Tract orientation was encoded by 3D color, red in left-right, green in anterior-posterior and blue in top-bottom.
23
Conclusions
In this work, DSI tractography was used to determine optimum sequence
parameters for DSI acquisition in the clinical setting. Using 203
diffusion-encoding gradients with bmax = either 3000 or 9000 s/mm 2 , the track
similarity is approximately 2 folds lower than that reconstructed from 691
encoding gradients. Using 515 diffusion-encoding gradients with bmax = 6000
s/mm 2 , the track similarity is only about 10-20 % worse than that reconstructed
from 691 encoding gradients. Taking the balance between the scan time,
hardware limitation and accuracy of tractography, 515 encoding gradients with
bmax of 6000 s/mm 2 is considered to be the optimum condition for clinical DSI.
Reference
[2] Lin et al., NeuroImage. 19:482-95, 2003.
[3] Hagmann et al., ISMRM2004, p62
[4] Kuo et al. ISMRM p1286. [5] Callaghan: Principles of nuclear magnetic resonance microscopy. Oxford Science Publication, 1991.
Figure 3. Plot of track similarity vs. gradient numbers using two different sampling approaches: one using fixed bmax and the other using fixed q.
24
L-W. Kuo 1 , V. J. Wedeen
2 , J-C. Weng
University, Taipei, Taiwan, 2 MGH Martinos Center for Biomedical Imaging, Harvard
Medical School, Charlestown, MA, United States, 3 Center for Optoelectronic Biomedicine,
National Taiwan University College of Medicine, Taipei, Taiwan Abstract
Knowing the location of a brain tumor and the status of the surrounding white matter
tracts helps neurosurgeons predict or even avoid unnecessary complications secondary to
post-op tract injury. In this work, we applied diffusion spectrum imaging to patients with
brain tumors and demonstrated its capacity to provide detailed information about the
relationships between tumors and the affected tracts.
Introduction
Knowing the location of a brain tumor and the status of the surrounding white matter
tracts helps neurosurgeons predict or even avoid unnecessary complications secondary to
post-op tract injury [1]. Recent advance in diffusion tractography techniques has shown its
potential to meet this need. Diffusion tensor imaging (DTI) tractography in patients with
brain tumors has been reported [2]. The tracts beside the tumors are often compressed and
distorted. The complex geometry of fiber tracts would compromise the accuracy of DTI in
defining local fiber orientation. Previously we proved that diffusion spectrum imaging
(DSI) could determine local fiber orientations accurately, especially in regions of complex
fiber structures [3]. Reconstruction of tractography from DSI data from normal subjects
clearly showed distinct axonal fiber tracts at the criss-cross regions [4,5]. In this work, we
applied DSI to patients with brain tumors and demonstrated its capacity to provide detailed
information about the relationships between tumors and the affected tracts.
Materials and Methods
Patients were scanned with a 3T MRI system (Trio, Siemens, Erlangen, Germany)
before surgery. An echo planar imaging (EPI) diffusion sequence with twice-refocused
balanced echo was used to acquire diffusion-weighted images. Isotropic spatial resolution
25
was obtained by making both in-plane and through-plane resolution to be 2.3 mm. The DSI
experiment was performed by applying 203 diffusion gradient vectors, each corresponded
to one of the isotropic 3D grid points in the q-space. The maximum diffusion sensitivity
bmax = 4000 s/mm 2 , and TR/TE = 6500/150 ms. Approximately 45 to 50 transaxial slices
were acquired encompassing the whole brain. The experiment completed in 30 min.
DSI analysis was based on the relationship that the echo signal S(q) and the diffusion
probability density function P(r) were a Fourier pair, i.e., S(q) = FT{P(r)} [6]. The
orientation density function (ODF) was determined by computing the second moment of
P(r) along each radial direction. The main orientation of diffusion probability was then
determined by the local maximum vectors of ODF [7]. Tractography was based on a simple
algorithm that was adapted for DSI data. The first three DSI vectors of each voxel over the
whole brain were used as the seeds. All fiber orientations of the nearest voxels were used to
decide the proceeding orientation for the next step; the most coincident orientation less
than 22° was chosen. A new starting point was then obtained to repeat tracking procedure.
The proceeding length for each step was 0.5 voxel length. The tracking stopped if there was
no coincident orientation in the nearest voxels.
In order to visualize the relationships between tumors and fiber tracts, tumor contours
were traced manually and 3D surface of the tumors was rendered. All tracks over the whole
brain or selected tracts passing through a specific region were shown and fused with the
tumor outlines.
Results
A total of thirteen patients were studied up to the preparation of this abstract. Here we
demonstrate tractography in two patients. The first patient had a tumor in right occipital
lobe (Fig. 1). The tumor displaced right inferior longitudinal fasciculus downward and
corona radiate anteriorly. The callosal fibers in the splenium were displaced upward. The
second patient had a tumor in left frontal lobe. The tumor displaced the callosal fibers
medially. The left superior longitudinal fasciculus was stretched, displaced medially and
upward by the tumor.
Fig. 1: a) T2-weighted images show a tumor in the right occipital lobe (yellow arrow).
b) Tractography viewing from the tumor side. The tumor (red arrow) displaces the
26
inferior longitudinal fasciculus downward (black arrow) and corona radiate anteriorly.
c) Tractography viewing from the top. Callosal fibers in the splenium are displaced
upward (yellow arrow).
Conclusions
We have applied the DSI tractography to patients with brain tumors and visualized the
change of the fiber tracts by the tumors. Further works on the clinical benefit of
tractography in the pre-surgical planning are currently undertaken. Besides tractography,
Fig. 1: a) T2-weighted images show a tumor in the right occipital lobe (yellow arrow). b) Tractography viewing from the tumor side. The tumor (red arrow) displaces the inferior longitudinal fasciculus downward (black arrow) and corona radiate anteriorly. c) Tractography viewing from the top. Callosal fibers in the splenium are displaced upward (yellow arrow).
Fig. 2: a) T2-wighted images show a tumor in left frontal lobe. b) Tractography viewing from the tumor side. The tumor (in red) displaces the callosal fibers medially. The left superior longitudinal fasciculus is stretched and displaced medially and upward by the tumor (yellow arrows)
27
functional MRI can also be acquired to study the relationships between the affected fiber
tracts and the function of the connected cortical regions.
Reference
[2] Mori et al., Ann. Neurol. 45, 265-269, 2002.
[3] Lin et al., NeuroImage. 19:482-95, 2003.
[4] Hagmann et al., ISMRM2004, p623.
[5] Kuo et al., ISMRM2004, p1286.
[6] Callaghan PT: Principles of nuclear magnetic resonance microscopy. Oxford Science
Publication 1991. [7] Wedeen et al., ISMRM2000, p82.
28
Diffusion Spectrum Imaging of Body-Center-Cubic Sampling Lattice in Q-Space
W-Y. Chiang1, L-W. Kuo2, M-H. Perng1, W-Y. I. Tseng3 1Precision Positioning and Control Lab, Department of Power Mechanical Engineering, National
Tsing Hua University, Hsinchu, Taiwan, Taiwan, 2Interdisciplinary MRI/MRS Lab, Department of Electrical Engineering, National Taiwan
University, Taipei, Taiwan, Taiwan, 3Center for Optoelectronic
Biomedicine, National Taiwan University College of Medicine, Taipei, Taiwan, Taiwan Abstract
In order to shorten the data acquisition time of diffusion spectrum imaging (DSI) [1], a new
homogeneous sampling scheme in q-space is proposed that the body-center cubic (BCC) sampling
lattice, which was previously experimented in CT or MRI [2], is used instead of the Cartesian one.
For spherical band-limited 3D signal, BCC sampling lattice is one of the most efficient sampling
schemes, and the sampling efficiency of BCC is 30% higher than the conventional Cartesian one
[2]. Because there are similar features in the diffusion spectrum of DSI, using BCC sampling
scheme may reduce sampling number up to 30% theoretically, which means 30% reduction of data
acquisition time. Once the q-space data of BCC lattice is acquired, probability density function
(PDF) is reconstructed directly by BCC inverse discrete Fourier transform (IDFT) without
interpolation. In addition, BCC sampling method saves data storage volume for 30% as well.
Introduction
Although fiber crossing problem was solved by DSI [1,3], the data acquisition time of it
is too long (compared with DTI) and many problems such as SAR problem may arise as
well. For instance, the optimum condition for clinical DSI is set to be 515
diffusion-encoding gradients and bmax = 6000 s/mm2 [4], and this requires data
acquisition time up to more than one hour, which is still time consuming. Traditionally,
data acquisition time is saved by reducing the sampling number in q-space, but reduced
number of sampling data may introduce aliasing effect, loss of PDF detail, and SNR drop,
etc. These problems may result in fiber orientation error.
When BCC sampling lattice in q-space is used, the aliasing effect of PDF can be
reduced because the repeated patterns is shifted to the corner of FOV (field of view) (Fig.2),
and the detail of PDF may be preserved by its higher sampling efficiency (Fig.1).
Therefore, if spherical windowing function is used, the aliased pattern of PDF in the corner
29
of FOV can be totally removed whereas the aliased signal in the Cartesian one still remains
(Fig.2).
Materials and Methods
DSI data were obtained from one healthy volunteer with a 3T MRI system (Trio,
Siemens, Erlangen, Germany). A twice-refocused balanced echo diffusion EPI sequence
was used to acquire MR diffusion images which were acquired with 515
diffusion-encodings comprising isotropic 3D grid points over the q-space. DSI data were
acquired with bmax = 6000 s/mm 2 , and TR/TE = 2900/150 ms, and isotropic voxels were
obtained by setting in-plane resolution and slice thickness to be 2.7 mm. 40 horizontal
slices encompassing the middle portion of the brain were acquired. Data of 515 points
sampled by Cartesian lattice in q-space served as the gold standard. Two sets of reduced
number of sampling points with different sampling lattice, i.e. BCC and Cartesian, were
assigned to be Set1 (experimental set) and Set2 (control set), respectively. The number of
data points between Set1 and Set2 was set as close as possible. The DSI data with reduced
sampling number of the Set1 and Set2 were obtained by sub-sampling / interpolation from
the same gold standard. The data points of all the experiment sets stated above were
sampled within sphere of radius of max equals 6000 s/mm 2 . Hamming filter was performed
before reconstructions of PDFs for all experiment sets, and the PDFs were windowed by a
spherical windowing function in order to reduce the folded signal caused by aliasing effect
into FOV—our experimental results showed that these processes were part of the key to
obtain more accurate orientation density function (ODF) by small number of DSI data
points. As a preliminary study, the comparison of quality of DSI tractography has not been
performed. Currently, orientations of fibers in each single volume were focused because
the accuracy of DSI tractography was closely related with the accuracy of ODFs, i.e. fiber
orientations. It was assumed that if the accuracy of ODFs was maintained, the quality of
DSI tractography was maintained as well. In order to compare ODFs, the first 3 maximum
orientations of Set1 and Set2 were compared with gold standard by angles between them;
this angular difference between orientations was adopted by this study as the criterion of
similarity comparison.
30
Results
Figure 2 demonstrates the aliased pattern of a PDF taken from a voxel in the 1st slice,
coordinate: (77, 77). When Set1 was set to be BCC sampling with 339 points, and Set2 was
Cartesian one with 341 points (34% reduction of sampling points), the similarity analysis
(angle between gold standard and Set1/Set2) showed in Fig.3. The mean differences of
angles between gold standard and Set1/Set2 are 5.4°/8.1° and 99.07%/92.06% volume
within 20° when all the volume of diffusion anisotropy higher than 0.035 is considered.
Fig.1 Comparison of sampling efficiency between Cartesian (top) and BCC (bottom) sampling lattice.
Fig.2 PDF vs. Sampling Lattice of Q-Space: (Type/Points) (a) Cartesian, 515 (b) BCC, 259 (c) Cartesian, 251
Fig.3 Similarity vs. Min. DA Fig.4 DSI Map: Axial view of human brain—Reconstructed by BCC lattice with 339 samples in q-space.
31
Discussion
Because maximum directions of isotropic PDF is more random than anisotropic one, the
similarities (accuracy) and minimum diffusion anisotropy (DA) [5] are heavily related
(Fig.3) which means the statistical accuracy is higher when volume with lower DA is
ignored. Experimental results showed that the sampling number of q-space may reduce up
to 34% (515→339 points) with acceptable error (5.4° by BCC lattice) , and even if the
same reduced number of sampling points was used, the accuracy improved 33.33%
(8.1°→5.4°) from BCC sampling lattice to Cartesian one.
162 directions is used in the ODF calculation of this thesis, and it is believed that the
greater difference between BCC sampling lattice and Cartesian one would be observed if
angular resolution of ODF was higher. Moreover, comparison should be done by new DSI
sequence of BCC sampling lattice, and acquire DSI data directly to compare the DSI
accuracy between different sampling lattice on the same phantom by the future work.
Reference
[1] Wedeen et al., ISMRM2000, p82.
[2]Theussl, T., Moller, T, Groller, M.E., “Optimal Regular Volume Sampling,”
Proceedings of the IEEE visualization Conference, San Diego, CA, USA, p91-98, 2001. [3]
Lin et al., NeuroImage. 19:482-95, 2003.
[4] Kuo et al., ISMRM2005, p391.
[5]Kuo et al., ISMRM2003, p592.
32
J-C. Weng
University, Taipei, Taiwan, 2 Center for Optoelectronic Biomedicine,
National Taiwan University, Taipei, Taiwan
Abstract
Magnetic resonance (MR) diffusion diffractogram has the potential to probe size of the
microstructure. However, realization of this method in the biological tissue is challenging.
Previously we have established a q-space diffraction technique to measure microscopic
length scales in a phantom. In this work we extended the same technique to a more
complex biological system, the rat corpus callosum at different ages. We found that T1 and
T2 values of the corpus callosum decreased monotonically from day 10 to day 14, and kept
rather constant afterwards. In contrast, the microstructure size of the corpus callosum,
defined as the reciprocal of the q position at the first node in the diffractogram, decreased
monotonically from day 10 to day 84. Slower change in the microstructure size implies that
the change in transmembrane water permeability during brain maturation might be slower
than the change in myelinization as indicated by T1 and T2. Further investigation of this
technique in monitoring the integrity of the cell membrane in diseased brain is warranted.
Introduction
Magnetic resonance (MR) diffusion diffractogram has the potential to probe size of the
microstructure [1]. However, realization of this method in the biological tissue is
challenging. The difficulty arises partly from the fact that cells are not perfectly aligned
and that water molecules exchange between intracellular and extracellular compartments.
To the best of our knowledge, no report was made to reveal coherent diffraction patterns
from an intact biological organ. Only one report demonstrated the coherence features in
erythrocyte suspensions [2]. Previously we have established a q-space diffraction
technique to measure microscopic length scales in a phantom [3]. In this work we extended
the same technique to a more complex biological system, the rat corpus callosum at
33
different ages. Axons in the corpus callosum pack themselves in the middle portion and
align regularly. Myelin sheath of the axon forms a natural barrier to hinder transmembrane
water exchange. These features make it an ideal model to realize diffusion diffraction
experiment in an intact organ. To test this model with our technique, rats at different ages
were studied to demonstrate the evolution of microstructure size of the corpus callosum
derived from the diffraction experiment during the process of brain maturation.
Materials and Methods
Seven male Wistar rats with the age at day 7, 10, 14, 21, 28, 56, and 84 were studied, the
body weights being 22, 28, 35, 48, 73, 330, and 400 grams, respectively. T1, T2 and
diffusion diffraction measurements were performed in each rat. To test the reproducibility,
additional five rats at day 84 were recruited for the diffusion diffraction measurement. The
data were acquired on a 3T MRI Biospec system (Bruker, Germany). T1WI of rat brains in
sagittal and coronal planes were obtained to localize an axial plane parallel to the middle
portion of the corpus callosum. After the imaging plane was defined, a spin echo diffusion
weighted sequence was performed to obtain q-space diffractograms. Images were acquired
with TR/TE = 1500/67 ms, FOV = 22 mm, slice thickness = 1 mm, matrix size = 16, and
number of excitations = 8. The diffusion gradients g perpendicular to the imaging plane
were applied with the gradient duration δ = 4 ms and diffusion time = 60 ms. Magnitudes
of the diffusion gradients were incremented from 0 to 950 mTm -1
, reaching the maximal
2 . After a series of diffusion-weighted images
were acquired, a region of interest (ROI) covering the middle portion of the corpus
callosum was selected, and the echo intensities of the pixels in the ROI were averaged. The
measured echo intensities in each spectral-series were normalized by the value of the
largest peak in the series. The T1 and T2 values of the corpus callosum were also measured
using TrueFISP and MSME sequences [4].
Results and Discussions
At day 7 the coherent diffraction pattern was not observed. It started to appear at day 10
and became obvious as the age increased (Fig. 1). The diffraction pattern was compatible
with the envelope squared sinc curves observed previously in our phantom study [3]. From
34
the curves we determined the reciprocal of the q-distance between the center and the first
node to indicate the microstructure size of the corpus callosum. During the process of brain
maturation, T1 and T2 values of the corpus callosum decreased monotonically from day 10
to day 14, and kept rather constant afterwards (Fig. 2a & 2b). In contrast, the
microstructure size of the corpus callosum decreased monotonically from 70 µm at day 10
to 30 µm at day 84 (Fig. 2c & 2d). The microstructure size at day 84 was 30.67µm ±3.79
µm (N= 5, Fig. 2c & 2d).
Fig.1 NMR diffusion diffraction patterns observed in rat corpus callosum at different ages.
Fig.2 Changes in T1 (a), T2 (b), q-value (c), and microstructure size (d) of the rat corpus callosum during the process of brain maturation.
35
The microstructure size measured by the diffraction pattern is greater than the actual
size of an axon, approximately 1 µm. The mismatch can be explained in part by the fact
that in addition to the intracellular water, there is significant contribution from the
extracellular water to the measured signal. The effect of transmembrane water exchange
on the diffractogram can be observed during the process of brain maturation. As shown in
Figure 1, as the brain matures, increasing integrity of the myelin sheath hinders the water
exchange, leading to progressive stretching of the diffraction pattern to the right, thus a
decrease in the measured size (Fig. 2c & 2d). Slower change in the microstructure size
implies that the change in transmembrane water permeability during brain maturation
might be slower than the change in myelinization as indicated by T1 and T2.
Conclusions
We have measured the microstructure size of the rat corpus callosum from the coherent
features of the q-space diffraction experiment. We have demonstrated progressive change
in the microstructure size during brain maturation. The change might correspond to the
change in transmembrane water permeability. Further investigation of this technique in
monitoring the integrity of the cell membrane in diseased brain is warranted.
References
[1] P. T. Callaghan et al., J.Phys.E: Sci.Instrum. 21,820. 1988.
[2]A. M. Torres et al., J. Magn. Reson. 138,135. 1999.
[3] J. Weng et al., ISMRM No. 589. 2003.
[4] P. Schmitt et al., Magn. Reson. Med. 51,661. 2004.
36
Assessment of Myocardial Perfusion Reserve in Patients with Ischemic Heart Disease on 3 Tesla MRI
M-Y. M. Su
1 , K-C. Yang
2 , C-C. Wu
2 , R-Y. Tseng
3 , Y-W. Wu
2 , W-C. Chu
Taiwan, 2 Department of Medicine, National Taiwan University Hospital, Taipei, Taipei,
Taiwan, 3 Center for Optoelectronic Biomedicine, National Taiwan University, Taipei,
Taipei, Taiwan Introduction
Clinical protocols of 3 Tesla MRI systems in the assessment of ischemic heart disease
are currently under active development. The diagnostic accuracy of myocardial
ischemia using 3 Tesla MRI has not been validated yet. Therefore, the purpose of this
study is to study the feasibility of the first-pass contrast-enhanced myocardial perfusion
technique applied on a 3 Tesla MR scanner, and to determine the diagnostic accuracy of
myocardial perfusion reserve index derived from the image data.
Materials and Methods
Study population
The study population consisted of 5 patients with typical angina pectoris and 12
age-matched healthy volunteers. All subjects underwent both rest and stress first-pass
contrast-enhanced MR studies on a 3 Tesla MRI scanner (Trio, Siemens, Germany).
Within 72 hours after MR exam, patients received coronary angiography and were
proven to have lumen stenosis of >75% in various vessels; two in 1 vessel, one in 2
vessels, and two in 3 vessels.
Image acquisition
Three short-axis planes at basal, mid left ventricular (LV) and apical levels were
acquired using SR-TrueFISP pulse sequence (TR/TE/FA=160ms/0.98ms/10°,spatial
resolution=2mm, temporal resolution=RR interval and the total number of time
frames=80). Right after the scanning started, Gd-DTPA (0.05mmol/kg) was injected via
left antecubital vein at a rate of 4 ml/sec. The stress study was performed in the same
way, except that vasodilator (dipyridamole, 140µg/kg -1
/min -1
) was infused
intravenously via right antecubital vein for 4 min and the image acquisition began at 7
37
Image analysis
LV myocardium and cavity were segmented semi-automatically and myocardium was
divided into 16 equiangular segments according to the definition of coronary artery
territories (1). Baseline signals measured before contrast enhanced were used to correct
for the depth dependent signal variation resulted from the receiver surface coils.
Myocardial perfusion was quantified by measuring the maximum upslope of the first pass
signal intensity curve from the LV myocardium, normalized relative to that from the LV
cavity. Myocardial perfusion reserve (MPR) index was calculated by dividing the results
at maximal vasodilation by the results at rest (2). Color mapping of MPR in bull’s eye
view was used for visual comparison (Fig. 1).
Statistic analysis
Group differences were tested by one-way ANOVA and Wilcoxon nonparametric
tests. Significance was defined as P<0.05. Receiver operating characteristic (ROC)
analyses were performed to evaluate the diagnostic sensitivity, specificity, and
accuracy of this method.
Ischemic segments (0.90±0.32) showed significant difference in MPR index comparing
with remote non-ischemic segments (1.57±0.47; p=0.004) and normal subjects (1.95±0.46;
p<0.001). There was no significant difference between remote and normal myocardium
(Fig. 2). In the ROC analysis, cutoff value of 1.3 was chosen and the sensitivity, specificity,
and diagnostic accuracy were 90%, 91% and 97%, respectively.
Conclusions
We derived MPR index with the first-pass contrast-enhanced technique on a 3 Tesla MR
scanner, and showed significant difference between normal and ischemic segments. These
results are consistent with those reported using 1.5T MR scanners (3). This study promises
the use of the first-pass perfusion study on 3T Tesla MR scanners to assess ischemic heart
disease.
38
References
[2] Wilke N. et al. Radiology 1997;204:373-384
[3] Nidal AS et al. Circulation 2000;101:1379-1383
Fig 1. Color mapping of myocardial perfusion in a patient with LAD stenosis (top row) and a normal subject (bottom row).
Fig 2. MPR in ischemic segments are significantly different from MPR in remote and normal controls, whereas there is no significant difference between remote and normal controls.
39
M-Y. M. Su 1 , H-Y. Yu
2 , W. Chu
1 , S-C. Huang
2 Department of Surgery, National Taiwan University Hospital, Taipei, Taiwan,
3 Department of Medical Imaging, National Taiwan University Hospital, Taipei, Taiwan,
4 Center for Optoelectronic Biomedicine, National Taiwan University College of Medicine,
Taipei, Taiwan
First-pass contrast-enhanced MRI is a potential tool to assess myocardial perfusion.
Current analysis methods mainly focus on the evaluation over the full thickness of the left
ventricular segments. This full-thickness approach, however, is inadequate to resolve
subendocardial perfusion defect, leading to miss diagnosis of focal perfusion abnormality.
The purpose of this study, therefore, was to develop a pixel-based perfusion analysis
method. The study was performed at 3T MR system. Algorithms of motion correction were
developed to co-register a series of images in dynamic contrast enhancement. With this
technique, subendocardial perfusion defects were detected and corresponded accurately to
regions of subendocardial infarction, documented by delayed contrast-enhanced MRI.
Materials and Methods
contrast-enhanced MRI (N=4) were recruited. All patients received first-pass
contrast-enhanced MR studies at a 3T MR scanner (Trio, Siemens, Germany) in both
rest and dipyridamoleœinduced stress conditions.
Image acquisition
Three short-axis planes at basal, mid left ventricular (LV) and apical levels were
acquired using SR-Turbo FLASH pulse sequence, TR/TE/FA = 160ms/0.98ms/10°,
spatial resolution = 2 mm, temporal resolution = 1 R-R interval and the total number of
time frames = 80. Right after the scanning started, Gd-DTPA of 0.025 mmol/kg dosage
was injected via left antecubital vein at a rate of 4 ml/sec. The stress study was performed
approximately 10 min after the rest study. Vasodilator, dipyridamole, 140 µg/kg -1
/min -1
40
was infused intravenously via right antecubital vein for 4 min and the image acquisition
began at the 7th min when the maximal vasodilation was achieved.
Image analysis
acquired before contrast enhancement were used to correct the depth-dependent signal
variation due to surface coils. For image co-registration, least square approximation was
applied to correct for in-plane translation and rotation at each time frame. Full-thickness
measurement and pixel-based measurement were performed with a semi-quantitative
method. Specifically, maximal upslope was calculated by the first derivative of the SI
curve during the initial up-rise of the first pass [1]. For full thickness measurement, LV
myocardium was subdivided into 16 equiangular segments according to the definition of
coronary artery territories [2]. Rendering of full-thickness analysis in bull‘s eye view (Fig.
1a and 1d) and pixel-by-pixel mapping of maximal upslope (Fig. 1b and 1e) were
generated for visual comparison.
Results
In this study, low dose contrast medium was administered to maintain the linear
relationship between the dose of contrast medium and the measured signal intensity in the
LV cavity. In such low dose regime, direct visualization of perfusion defect was difficult
even in stress studies. Subendocardial perfusion defects were not detected at the regions
of subendocardial infarction in the full-thickness analysis (Fig. 1a and 1d). In the
pixel-based perfusion mapping, reduced perfusion in the subendocardium was detected,
corresponding accurately to the infarction areas shown in the delayed contrast-enhanced
MRI (Fig. 1c and 1f).
Conclusion
In this study, we proposed a pixel-based perfusion analysis method to detect
subendocardial perfusion defect that could be missed in the full-thickness analysis. Our
works also demonstrated that standard 3T MR system can provide adequate contrast to
noise ratio for pixel-based perfusion analysis. The accuracy of this method, however, is
highly affected by severe through-plane motion and beat-to-beat variation of the R-R
41
interval. In conclusion, pixel-based perfusion analysis is a feasible method; it may be
potentially useful to detect subendocardial ischemia in syndrome X [3].
References
[1] Al-Saadi N, Nagel E, Gross M, et al. Noninvasive detection of myocardial ischemia
from perfusion reserve based on cardiovascular magnetic resonance.
Circulation 2000; 101:1379-1383.
[2] Cerqueira MD, Weissman NJ, Dislsizian V, et al. Standardized myocardial
segmentation and nomenclature for tomographic imaging of the heart: a statement for
healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical
Cardiology of the American Heart Association. Circulation 2002; 105:539-
542.
[3] Panting JR, Gatehouse PD, Yang GZ, et al. Abnormal subendocardial perfusion in
cardiac syndrome X detected by cardiovascular magnetic resonance imaging. N
Engl J Med 2002; 346(20):1948-1953.
Figure 1. Full-thickness analysis of perfusion in two patients, upper and lower rows, respectively, shows rather homogeneous distribution (a, d). Pixel-based perfusion mapping of the same patients (b, e) shows subendocardial perfusion defects, accurately corresponding to infarction areas in the delayed contrast-enhanced MRI (c,f).
42
M-Y. M. Su
1 , K-C. Yang
2 , C-C. Wu
2 , W. Chu
College of Medicine, Taipei, Taiwan
Introduction
Percutaneous coronary intervention (PCI) has been used to restore myocardial blood
flow in the stenotic coronary arteries. Conventional methods such as coronary
angiography and nuclear tracer imaging are limited in the assessment of the magnitude
and timing of recovered myocardial perfusion after PCI. In this study, we performed
first-pass contrast-enhanced myocardial perfusion MRI (CEMR) at 3T MR system to
determine the timing and magnitude of perfusion recovery by evaluating the change in
myocardial perfusion reserve (MPR) index in a serial follow-up.
Materials and Methods
Study protocol Five patients (N=5) with angiographically-documented coronary artery
stenosis were enrolled in this study (Table 1). All subjects underwent both rest and stress
first-pass CEMR studies on a 3T MR scanner (Tiro, Siemens, Germany). All patients
received MR studies before PCI treatment and repeated MR studies by 24 hr, one week and
one month after the PCI treatment. Image acquisition Three short-axis planes at basal, mid
left ventricle (LV) and apical levels were acquired using SR-Turbo FLASH pulse sequence.
Right after the scanning started, Gd-DTPA(0.05mmole/kg) was bolus injected via left
antecubital vein at a rate of 4~6ml/sec. The stress study was performed approximately 10
min after the rest study. Vasodilator (dipyridamole, 140µg/kg/min) was infused
intravenously for 4 min and the image acquisition began at the 7th min when the maximal
vasodilatation was achieved. Image analysis LV myocardium were segmented
semi-automatically and divided into 16 equiangular segments according to the definition of
coronary artery territories (2). After the baseline signal correction, myocardial perfusion
43
was quantified by measuring the maximum upslope of first pass signal intensity curve from
the LV myocardium, normalized to that from the LV cavity. MPR index was calculated by
dividing the results at maximal vasodilatation by the results at rest. The MPR of ischemic
segments and non-ischemic segments were compared at baseline as well as at 24 hrs, 1
week and l month following PCI treatment. The paired MPR change of each segments were
also analyzed at the same time points.
Statistical Analysis
Data were presented as mean±SD or as numbers and percentages. Paired pre-and
post-PCI segment MPR index differences were tested by paired Kolmogorov-Smirnov test
at 24 hrs, 1 week and l month after treatment. The differences of MPR between ischemic
and non-ischemic segments were also analyzed using 2 sample Komogorov-Smironv test.
Statistical significance was considered if p<0.05.
Results
Among the recruited 5 patients, a total of 64 segments of the myocardium were
successfully analyzed, 21 ischemic segments and 43 non-ischemic segments (remote
segments). The MPR index of the ischemic segments (0.77±0.39) were significantly lower
than that of non-ischemic segments (1.46±0.82, p=0.003) before PCI. The differences of
MPR between these two groups disappeared at 24 hrs, 1 week and 1 month following
successful PCI procedure for the ischemic segments (Fig. 1). The MPR of the ischemic
segments significantly improved at 24 hrs after PCI. The improvement of MPR lasted at
1 week and tended to persist at 1 month following PCI
(Table 1). The MPR in non-ischemic segments increased at 24 hrs (1.99±0.96, p=0.04)
and then returned to baseline at 1 week and 1 month following PCI (Table 1).
Conclusion
Our study indicates that the myocardial perfusion in ischemic segments can be restored
as early as 24 hours after successful PCI. The improvement of myocardial perfusion can
last at least for one week and tend to persist at one month following PCI. The myocardial
perfusion in non-ischemic segments can be transiently increased at 24 hour following PCI,
which may indicate a hyperemic response in remote segments after coronary intervention.
44
References
1. Atkinson DJ, Burstein D, Edelman RR. First-pass cardiac perfusion: evaluation with
ultrafast MR imaging. Radiology 1990;174:757-762.
2. Hany TF, McKinnon GC, Leung DA et al. Optimization of contrast timing for
breath-hold three-dimensional MR angiography. J Magn Resion Imaging 1997;7:551-556.
Table 1. Pre and post-PCI MPR changes in ischemic segments and non-ischemic segments
Figure 1. MPR differences between ischemic and non-ischemic segments at baseline, 24 hrs, 1 week and 1 month following PCI.
45
H-H. Peng
1 Dept. of Electrical Engineering, National Taiwan University, Taipei, Taiwan,
2 Dept. of
Taiwan, 3 Dept. of Mechanical Engineering, National Taiwan University, Taipei, Taiwan,
4 Division of Medical Engineering Research, National Health Research Institutes, Taipei,
Taiwan, 5 Dept. of Physical Medicine and Rehabilitation, National Taiwan University
Hospital, Taipei, Taiwan, 6 Center for Optoelectronic Biomedicine, National Taiwan
University, Taipei, Taiwan
Recently the development of high intensity focused ultrasound (HIFU) technology has
offered a potentially new approach to the local ablation of cancer [1] or myoma. The
utilization of MRI for guiding HIFU beams can not only greatly increase the localized
accuracy during HIFU heating procedures but also can be used to evaluate the
HIFU-induced lesions after treatments [2]. During the past few years, the temperature
measurement procedure using phase mapping was developed that makes use of the
temperature dependence of the water proton chemical shift [3] which is so-called proton
resonance frequency (PRF) shift method. In addition, the changes in magnetization transfer
(MT) contrast of tissues after heat treatment was also evaluated in a previous study [4]. A
real time evaluating method, that includes temperature monitoring as well as MT contrast
of thermal damage during sonication, should be helpful to improve the heating efficiency
of HIFU beams and to avoid the damage of adjacent normal tissues. Our study investigated
the feasibility of estimating temperature change and MT contrast simultaneously during
HIFU heating treatments.
Materials and Methods
A special phantom setting (Fig.1) was designed to verify the feasibility of observing
temperature change and MT contrast simultaneously. The upper hot water was set up for
monitoring temperature change during scanning, while the lower part contained a gel phantom
immersed in cold water, with the gel having spots pre-treated with HIFU heating to observe
the change of MT ratio (MTR). The HIFU power and heating time of each spot was listed in
Table 1. Two sets of MT phantom were used, one using fixed HIFU power (60W) with
varying heating time, and the other with varying HIFU power at constant heating time (30s).
All MR images were acquired on a 3T clinical imager (Siemens Trio, Erlangen, Germany).
The pulse sequence used a dual gradient-echo design, with ON and OFF of the MT pulse
interleaved (Fig.2), such that the phase images from the two gradient echoes could be used to
estimate PRF shift in response to temperature change, and the first echoes from two
consecutive TR could be used to derive MTR on a pixel-by-pixel basis [4]. Imaging
parameters were TR=42, TE=4.52/20, flip angle=150 FOV=160x160 mm2, matrix
size=128x128, slice thickness=1.5 mm, off-resonance frequency of MT pulse=+500 Hz. For
comparison, the temperature in hot water was continuously measured by a thermal meter (54II
thermometer, FLUKE) placed near the bottom. To evaluate the MT contrast of spots with
different HIFU heating conditions, we selected several ROIs from MTR images, as shown in
Fig.3, and compare the difference of MTR values between HIFU-heated spots and an area
without HIFU heating (“nor” in Fig.3)
Results
The temperature changes measured by MR PRF shift method in the bottom of hot water
(ROI shown in Fig.3a) and by thermal meter were shown in Fig.4. The highly consistent
estimation values indicated that MR PRF is a precise method for continuous temperature
monitoring. The MTR contrast of HIFU-heated spots are shown in Table 1. The MTR contrast
increased from 0.0397 to 0.0809 with increasing HIFU heating time, and increased from
0.0318 to 0.0874 increasing transmitting power from 70W to 90W. Both of the monotonic
relationships are in agreement with Two sets of MT phantom were The pulse sequence the
expectation of protein denaturation via HIFU heating.
47
Monitoring temperature changes and protein denaturation during transmitting HIFU pulses is
important for evaluating treatment efficiency. In our study, we verified the feasibility of
simultaneous temperature monitoring and MTR measurement by MR imaging. The linearity of
temperature changes measured by the PRF-shift method as validated using a thermal meter
demonstrated the precision of the MR approach. Furthermore, the distinct changes of MTR
contrast illustrated its potential usefulness to evaluate the degree of protein denaturation during
HIFU transmission. We therefore conclude that MRI with simultaneous temperature
estimations and MT measurements is an effective technique that could help monitoring HIFU
treatment by reflecting local heating conditions and progress of protein denaturation. The
proposed method should be beneficial to the improvements of HIFU heating efficiency and
could decrease unnecessary heating damage to tissues nearby the targeting area.
Fig 2. The dual
measured by MR PRF shift and thermal meter
during MR scanning. High correspondence of
these two methods are noticed.
Fig. 3. (a) Phase image of hot water; (b)
MTR image of Phantom 1. (c) MTR
image of Phantom 2. The yellow lines
were the selected ROIs. “nor” stand for
area without HIFU heating.
References
[1] Cheng et. al, J Cancer Res Clin Oncol 123 :219-223, 1997.
[2] Vykhodtseva et. al, Ultrasound in Med & Biol 26(5) :871-880, 2000.
[3] Ishihara et. al, MRM 34 :814-823, 1995.
[4] Graham et. al, MRM 42 :1061-1071, 1999
Phantom 1 Phantom 2
49
Best Poster Award
Kai-Chien Yang, Assessment of Myocardial Perfusion Reserve in Patients with CAD on
3 Tesla MRI, The 35th annual congress of the Taiwan Society of Cardiology, August
20~21, 2005. Best Poster Award.
Li-Wei Kuo, Van J. Wedeen, Jun-Cheng Weng, Timothy G. Reese, Jyh-Horng Chen,
Wen-Yih Isaac Tseng. Diffusion spectrum tractography in patients with brain tumors.
Proc ISMRM 13th Ann Meeting, Miami, USA, May 7-13, 2005. Honorable Mentioned.
Li-Wei Kuo, Wen-Yih Isaac Tseng, Ching-Po Lin, Jun-Cheng Weng, Jyh-Horng Chen,
"Diffusion Spectrum MRI of Rat Epilepsy Model," Conference on Biomedical
Engineering Technology, Kaohsiung, Taiwan, 2002, poster D2-25. Best Poster Award.
2002
50
Imaging of Complex White Matter and Connectivity :

using diffusion spectrum imaging and diffusion weighted functional MRI

High Intensity Focused Ultrasound Treatment of Hepatoma: Guided by Dynamic
Contrast Enhanced MRI
and MRI

A Neurobiological Study of Schizophrenia – A study on Etiological Factors of
Schizophrenia
Mechanism and Clinical Implication of the Paramnesia: An Approach Using Functional
MRI and Tractography –

Schizophrenia —

Study of the pathological and magnetic resonance image findings of temporal lobe
epilepsy by animal model

51
Binocular suppression in amblyopia: A brain imaging study

resonance imaging study

Glass-pattern
amyotrophic lateral sclerosis (ALS)
Bone marrow-derived stem cells in the treatment of chronic myocardial ischemia: the
role of growth factor modification and in vivo assessment of cell engraftment
: (1) (2)

Application of Magnetic Resonance of Myocardial Perfusion to Evaluate the Effect of
Percutaneous Coronary Interventions in Patients with Chronic Stable Angina
—

In vivo intra-ventricular flow stream and pressure field analysis by magnetic resonance
imaging
MRI flow quantification of vertebral artery stenosis pre- and post-percutaneous
transluminal angioplasty and relations among the serum inflammation, adhesion
molecules and vertebral artery stenosis

The clinical assessment of stent implantation in coronary artery using combined medical
imaging analysis and computational fluid dynamic analysis

53
Established a panel of biomarkers for assessing RA joint erosion

Lesion Formation and Transformation during the Treatment of Experimental Liver
Tumors Using a MRI-guided High-intensity Focused Ultrasound System

Journal papers:
1. Ching-Po Lin, Van Jay Wedeen, Jyh-Horng Chen, Ching Yao, Wen-Yih I. Tseng. Validation
of diffusion spectrum magnetic resonance imaging with manganese-enhanced rat optic tracts
and ex-vivo phantoms. NeuroImage 2003; 19:482-495.
2. Wedeen VJ, Hagmann P, Tseng WY, Reese TG, Weisskoff RM. Mapping complex tissue
architecture with diffusion spectrum magnetic resonance imaging. Magnetic Resonance in
Medicine, 2005;54:1377-86.
3. Jun-Cheng Weng, Jyh-Horng Chen, Ching-Po Lin, Li-Wei Kuo, Wen-Yih I. Tseng.
Magnetic Resonance Diffusion Diffractogram in the Assessment of Microstructure Sizes of
Rat Corpus Callosum during Brain Maturation. Magnetic Resonance in Medicine; in review.
4. Mao-Yuan M. Su, Kai-Chien Yang, Chau-Chung Wu, Yen-Wen Wu, Rung-Yu Wu, Wen-Yih
I. Tseng. Assessment of Myocardial Perfusion using a 3 Tesla Cardiac Magnetic Resonance.
Journal of Magnetic Resonance Imaging, in review.
International conference papers:
1. Jyh-Ray Chen, Jung-Cheng Wong, Li-Wei Kuo, Jaw-Lin Wang, Jyh-Horng Chen, Wen-Yih
Isaac Tseng. Mapping 3D Fiber Orientation in Disc Anulus Fibrosus with Diffusion Spectrum
MRI. Proc ISMRM 13th Ann Meeting, Miami, USA, May 7-13, 2005.
2. P. Hagmann, S. Masip, C-P Lin, W-Y I. Tseng, R. Meuli, V. J. Wedeen, J-P Thiran.
Modelazation and restoration of diffusion spectra. Proc ISMRM 13th Ann Meeting, Miami, USA,
May 7-13, 2005.
3. Mao-Yuan Marine Su, Kai-Chien Yan, Chau-Chung Wu, Rung-Yu Tseng, Yen-Wen Wu,
Wen-Yih Isaac Tseng. Assessment of Myocardial Perfusion Reserve in Patients with Ischemic
Heart Disease on 3 Tesla MRI. Proc ISMRM 13th Ann Meeting, Miami, USA, May 7-13, 2005.
4. Li-Wei Kuo, Van J. Wedeen, Jun-Cheng Weng, Timothy G. Reese, Jyh-Horng Chen, Wen-Yih
Isaac Tseng. Using track similarity to determine optimum sequence parameters for diffusion
spectrum imaging. Proc ISMRM 13th Ann Meeting, Miami, USA, May 7-13, 2005.
5. Li-Wei Kuo, Van J. Wedeen, Jun-Cheng Weng, Timothy G. Reese, Jyh-Horng Chen, Wen-Yih
55
Isaac Tseng. Diffusion spectrum tractography in patients with brain tumors. Proc ISMRM 13th
Ann Meeting, Miami, USA, May 7-13, 2005.
6. Li-Wei Kuo, Van J. Wedeen, Jun-Cheng Weng, Timothy G. Reese, Jyh-Horng Chen, Wen-Yih
Isaac Tseng. Reconstruction and visualization of white matter tracts based on clinical diffusion
spectrum imaging. Proc ISMRM 13th Ann Meeting, Miami, USA, May 7-13, 2005.
7. Jun-Cheng Weng, Jyh-Horng Chen, Der-Yow Chen, Wen-Yih Isaac Tseng. Magnetic Resonance
Diffusion Diffractogram in the Assessment of Microstructure Sizes of Rat Corpus Callosum
during Brain Maturation. Proc ISMRM 13th Ann Meeting, Miami, USA, May 7-13, 2005.
8. Wen-Yih I. Tseng, L. Magnussen, R. Gilbert, T. Benner, R. Wang, T. Reese, V. Wedeen. Global
myocardial fiber and sheet architecture in diffusion spetrum imaging tractography. Proc ISMRM
13th Ann Meeting, Miami, USA, May 7-13, 2005.
9. Shiou-Ping Lee, Chung-Ming Chen, Jun-Cheng Weng, Wen-Yih Isaac Tseng. Optimum
Diffusion Encoding Steps for Fiber Tractography. Proc ISMRM 13th Ann Meeting, Miami, USA,
May 7-13, 2005.
10. Hsu-Hsia Peng, Hsiao-Wen Chung, Hsi-Yu Yu, Wen-Yih Issac Tseng. Pulmonary Windkessel
Volume and Resistance Parameters in Patients with Pulmonary Hypertension Using MR Phase
Contrast Imaging. Proc ISMRM 13th Ann Meeting, Miami, USA, May 7-13, 2005.
11. Hsu-Hsia Peng, Yen-Hon Lin, Hsien-Li Kao, Hsiao-Wen Chung, Wen-Yih Issac Tseng.
Phase-contrast MR hemodynamic evaluation in basilar artery for posterior circulation ischemia:
Preliminary reproducibility assessment using a three-point localization technique. Proc ISMRM
13th Ann Meeting, Miami, USA, May 7-13, 2005.
12. V. J. Wedeen1, R. Wang, T. Benner, P. Hagmann, W-Y. Tseng, T. G. Reese1, A. G. Sorensen,
A. de Crespigny. DSI Tractography of CNS Fiber Architecture and Cortical Architectonics. Proc
ISMRM 13th Ann Meeting, Miami, USA, May 7-13, 2005.
13. V.J. Wedeen, S.K.V. Song, L .Wald, T.G. Reese, T. Benner, W.Y.I. Tseng. Diffusion spectrum
MRI of cortical architectonics: visualization of cortical layers and segmentation of cortical areas
by analysis of planar structure. Proc ISMRM 12th Ann Meeting, Kyoto, Japan, May 15-21, 2004.
14. Hsu-Hsia Peng, Hsi-Yu Yu, Hsiao-Wen Chung, Wen-Yih Issac Tseng. Estimation of
pulmonary windkessel volume in patients with pulmonary hypertension using MR phase contrast
imaging. Proc ISMRM 12th Ann Meeting, Kyoto, Japan, May 15-21, 2004.
15. Jun-Cheng Weng, Van J. Wedeen, Jyh-Horng Chen, Timothy G. Reese, Li-Wei Kuo, Wen-Yih
Isaac Tseng. Optimization of diffusion spectrum magnetic resonance imaging for clinical scanner.
56
Proc ISMRM 12th Ann Meeting, Kyoto, Japan, May 15-21, 2004.
16. Li-Wei Kuo, Van J Wedeen, Jun-Cheng Weng, Timothy G. Reese, Jyh-Horng Chen, Wen-Yih
Isaac Tseng. Mapping white matter connectivity with BOLD activated regions using diffusion
spectrum imaging and fMRI. Proc ISMRM 12th Ann Meeting, Kyoto, Japan, May 15-21, 2004.
17. Hagmann P, Reese TG, Tseng WY, Meuli R, Thiran J-P, Wedeen VJ. Diffusion Spectrum MRI
addresses the problem of tracking in complex fiber architecture. Proc ISMRM 12th Ann Meeting,
Kyoto, Japan, May 15-21, 2004.
18. C-P Lin, W-Y I. Tseng, Li-Wei Kuo, V.J. Wedeen, J.H. Chen. Orientation distribution function
acquisition with spherical encoding. In Proceedings: ISMRM 11th Scientific Meeting and
Exhibition. Toronto, Canada, May 10-16, 2003.
19. C-P Lin, W-Y I. Tseng, Jun-Cheng Weng, V.J. Wedeen, J.H. Chen. Reduced encoding of
diffusion spectrum imaging with cross-term Correction. In Proceedings: ISMRM 11th Scientific
Meeting and Exhibition. Toronto, Canada, May 10-16, 2003.
20. Li-Wei Kuo, Sheng-Kwei Song, Van J Wedeen, Ching-Po Lin, Jyh-Horng Chen, Wen-Yih
Isaac Tseng. Mean diffusivity and anisotropy index mapping of diffusion spectrum imaging in a
stroke model. In Proceedings: ISMRM 11th Scientific Meeting and Exhibition. Toronto, Canada,
May 10-16, 2003.
21. Jun-Cheng Weng, Ching-Po Lin, Li-Wei Kuo, Jyh-Horng Chen, Wen-Yih Isaac Tseng.
Diffractogram analysis of microstructures using MRI. In Proceedings: ISMRM 11th Scientific
Meeting and Exhibition. Toronto, Canada, May 10-16, 2003.
22. Wen-Yih Isaac Tseng, Chih-Chuan Chen, Ching-Po Lin, Horng-Huei Liou, Jyh-Horng Chen,
Li-Wei Kuo, Van J Wedeen. Diffusion spectrum imaging reveals disorganized cytoarchitecture in
the hippocampus of pilocarpine-induced epileptic rats. In Proceedings: ISMRM 11th Scientific
Meeting and Exhibition. Toronto, Canada, May 10-16, 2003.
23. Shu-Hsia Peng, Hsi-Yu Yu, Wen-Yih Isaac Tseng. Estimation of pulse wave velocity in main
pulmonary artery with phase contrast MRI. In Proceedings: ISMRM 11th Scientific Meeting and
Exhibition. Toronto, Canada, May 10-16, 2003.
24. Hsi-Yu Yu, Shu-Hsia Peng, Wen-Yih Isaac Tseng. Visualization and quantification of
pulse wave velocity of aorta by phase contrast cine MRI. In Proceedings: ISMRM 11th
Scientific Meeting and Exhibition. Toronto, Canada, May 10-16, 2003.