Measurement of in vivo and ex vivo ADC values of Lipid Core and Hemorrhage

in Human Carotid Atherosclerotic Plaque

Seong-Eun Kim1,2, Gerald S. Treiman 4,5, John A. Roberts1,2, Eun-Kee Jeong1,2,

Xianfeng Shi1,3, J Rock Hadley 1,2, Dennis L. Parker1,2

1
Utah Center for Advanced Imaging Research, and 3Brain Research Institute
Department of 2Radiology, and Department of 4 Surgery, University of Utah,
5
Department of Veterans Affairs, VASLCHCS
Salt Lake City, Utah

Corresponding author: Seong-Eun Kim, Ph.D.

Utah Center for Advanced Imaging Research
Department of Radiology, University of Utah,
729 Arapeen Dr., Salt Lake City, Utah 84108
(phone) 801-581-3144 (FAX) 801-585-3592
e-mail: sekim@ucair.med.utah.edu

ACKNOWLEDGEMENTS
This work has been supported by NIH grants R01 HL 48223, R01 HL 57990, R21 NS052424,
R21 EB005705, Clinical Merit Review Grant from the Veterans Administration Health Care System,
Siemens Medical Solutions, the Mark H. Huntsman Endowed Chair, and the Ben B. and Iris M.
Margolis Foundation.
ABSTRACT
Purpose: Determine the ADC values of lipid and hemorrhage in atherosclerotic plaque in human carotid
arteries in vivo and compare the values obtained from ex vivo carotid endarterectomy specimens.
Materials and Methods: In vivo diffusion weighted imaging (DWI) of carotid arteries was performed
using a 2D single shot- Interleaved Multislice Inner Volume- Diffusion Weighted Echo Planar Imaging
(2D ss-IMIV-DWEPI) sequence on eight subjects who subsequently underwent carotid endarterectomy.
The in vivo ADC maps were calculated based on DWI of b=10 and 300 s/mm2. A total of 32 slices used
to construct the ADC maps were reviewed for the measurement of the mean ADC values in vessel wall,
hemorrhage and lipid necrotic core. Eight ex vivo carotid endarterectomy specimens were scanned using
by 3D ultishot- Inner Volume-DWEPI (3D ms-IV-DWEPI) sequence. After the ADC maps were created
based on DWI of b=10 and 1000 s/mm2, the mean ADC values in the same locations selected for in vivo
values were calculated.
Results: The mean ADC values obtained from in-vivo DWI in normal vessel wall, lipid rich core, and
hemorrhage were 1.27±0.16x10-3, 0.38±0.10x10-3 and 0.98±0.25x10-3 mm2/s, respectively. The mean
ADC values in ex-vivo lipid necrotic core, and hemorrhage were 0.33±0.08x10-3, 1.28±0.10x10-3 mm2/s,
respectively. These components mean ADC values obtained from in vivo and ex vivo ADC maps were
compared.
Conclusions: ADC values of the normal carotid wall and carotid plaque components in vivo are
consistent with values obtained from ex vivo endarterectomy specimens. The ability to obtain consistent
plaque ADC values in vivo indicates that this technique could be an integral part of the basis for plaque
component identification in conjunction with other MRI/MRA techniques.
Key Words: MRI, Lipid necrotic core, Hemorrhage, Carotid endarterectomy specimens, Diffusion
weighted imaging, ADC values, 2D ss-IMVI-DWEPI, 3D ms-IV-DWEPI.
INTRODUCTION
Disease in the cervical carotid artery is a major cause of stroke and subsequent disability and mortality 1.
Development and progression of atherosclerosis may cause obstruction of the arterial lumen resulting in
stenosis and impaired cerebral perfusion. However, evidence suggests that unstable, or vulnerable,
plaques may lead to cerebral embolization and this is thought to be the more common etiology of focal
cerebral ischemia1. Vulnerable plaques areis characterized by specific components that cause
predispositione to thrombosis or embolization. Atherosclerotic plaques isare composed predominantly of
lipid/necrotic core with overlying fibrous cap. Erosion or disruption of the overlying fibrous layer with
exposure of the lipid/necrotic core is thought to represent an important etiology of thrombus formation
and subsequent embolization. Intraplaque hemorrhage is thought to contribute to rapid enlargement of
the plaque core and may be a cause of cap disruption 2 3. Vulnerable plaques with the components
described are not reliably identifiable in vivo by conventional imaging. Determination of percent
stenosis, which is the traditional focus of carotid disease evaluation, is limited for identifying vulnerable
plaque because it neglects characterization of the plaque ultrastructure3.
High-resolution magnetic resonance imaging (MRI) has increasingly been used as a noninvasive
modality for atherosclerotic plaque characterization because it permits both assessment of plaque burden
and determination of plaque constituents in vivo, with good ex vivo correlation. MRI sequences that are
currently used to characterize atherosclerotic plaque composition include 3D time of flight 4, steady-state
free-precession (FIESTA, True FISP), and 2D turbo/fast spin echo (TSE/FSE) black blood techniques.
Although each sequence provides different image contrast that helps discriminating the various plaque
components, significant limitations remain in indentifying unambiguous plaque components 5-9.
Atherosclerotic plaque characterization by these sequences is generally based on the signal intensities in
multi-contrast images such as T1w, T2w and proton density. However, the in vivo quantification of high
risk plaque components by conventional MRI signal intensity suffers from low sensitivity and specificity
10 11 12
.
Diffusion weighted imaging (DWI) relies on thermally activated molecular motion, which can be
quantified by the apparent diffusion coefficient (ADC). The ADC is a phenomenological measure that
combines the actual diffusion of the observed water molecules with the geometric restrictions imposed
on molecular motion by barriers such as cell membranes, organelles, and other tissue microstructures.
DWI is highly sensitive to the change in the physical environment in which water molecules reside, and
this allows discrimination between different tissues on a quantitative basis13. The calculated water
diffusion coefficients in different components within atherosclerotic plaques suggest that DWI might
provide a tool for the discrimination of intraplaque hemorrhage and lipid/necrotic core from other
components14. Previous investigation of DWI characteristics of ex-vivo plaque revealed restricted
diffusion within intraplaque thrombus relative to the vessel wall of the rabbit aorta, and this technique
provided excellent quantitative contrast between these two tissues15. Further, there is evidence that DWI
can uniquely discriminate lipid/necrotic core from fibrous tissue. Protons in lipid molecules, even within
a liquid phase, undergo diffusion much more slowly than do protons in water molecules. For instance,
human breast adipose tissue, which consists mainly of triglyceride, is characterized by a very low
diffusion coefficient16.
However, in spite of the potential use of DWI in characterizing carotid plaques, most data on DWI is
based on imaging of ex vivo specimens. Good quality imaging with DWI is difficult to achieve in vivo,
especially in the vicinity of the carotid bifurcation where flow characteristics and motion are significant
limitations. In an attempt to eliminate the effects of flow and motion and at the same time allow
increased resolution with reduced geometric distortion and blurring, we have developed a 2D single shot
Interleaved Multislice Inner Volume Diffusion Weighted Echo Planar Imaging (2D ss-IMIV DWEPI)
technique in which interleaved multislice inner volume (IMIV) imaging was combined with 2D ss-
DWEPI to obtain minimally distorted high resolution DW images of the cervical carotid artery17,18. This
2D ss-IMIV-DWEPI sequence allows high-resolution interleaved multiple slices DWI in a localized
region of the carotid artery.
The first DWI of in vivo human carotid artery using 2D ss-IMIV-DWEPI technique has been
successfully reported17. This prior work focused on the feasibility of the 2D ss-IMIV DWEPI technique
to obtain good quality in vivo DW images of carotid arteries but did not measure the ADC for specific
plaque components. Building on this prior feasibility demonstration, this study investigates the utility of
the Apparent Diffusion Coefficient (ADC) in characterization of atherosclerotic plaques, including lipid
core and hemorrhage. In this study, the mean ADC values of lipid necrotic core and hemorrhage were
measured in vivo and then correlated with the ex-vivo values measured from histologically registered
carotid endarterectomy specimens. Therefore, this study reports on in vivo and corresponding ex vivo
ADC values of hemorrhage and lipid core and seeks to improve the characterization of vulnerable
plaque components.

MATERIALS AND METHODS

In vivo Imaging Studies: This study was approved by the institutional review board. All subjects signed
informed consent. All studies were performed on a Siemens Trio 3T MRI scanner (Siemens Medical
Solutions, Erlangen, Germany). Diffusion weighted images of bilateral carotid arteries of 8 subjects with
atherosclerosis were acquired using 2D ss-IMVI-DWEPI at the slice locations centered approximately at
the carotid bifurcation. These subjects were selected as having 70% carotid stenosis by duplex
ultrasound examination and subsequently underwent carotid endarterectomy (CEA). Figure 1A displays
the pulse diagram of 2D-ss-IMIV DWEPI. The carotid bifurcation was localized using a low resolution
multiple overlapping thin slab acquisition (MOTSA) scan. Custom designed bilateral dual-element
phased array surface coils were placed on each side of the neck and centered over the region of the
carotid bifurcation19. The imaging parameters in DWI were: receiver bandwidth = 1.086 kHz/pixel, FOV
= 160x40 mm2, imaging matrix = 160x40, 2 mm slice thickness, effective TE = 56 ms, TR = 3 s, 41
echoes per each slice, 32 averages (magnitude) and interleaved acquisition of 16 contiguous slices. The
in-plane spatial resolution for typical data acquisition was 1x1 mm2 with display resolution 0.5x0.5 mm2,
after zero-filled interpolation. DW images were acquired with b =10 and 300 s/mm2 along the slice
selection direction, with interleaved acquisition of the two b value images. The total imaging time for
both b-values in one diffusion direction with 32 averages was 3 min 20 sec.
T1w images were acquired with 2D TSE using our modified version of the double inversion
preparation20,21 with TR = 850 ms, TI = 500 ms, TE = 9.5 ms, and ETL =11, spatial resolution of
0.5x0.5x2 mm3. T2w images were acquired at the same slice locations using 2D TSE with 0.5x0.5x2
mm3, TR = 2.0 s, TE = 65 ms and ETL = 15. For the T2w images, spatial saturation RF pulses were
applied superior and inferior to the imaging location to suppress both arterial and venous blood signal.
For both T1w and T2w acquisition, a chemical-shift fat saturation RF pulse was applied to eliminate
perivascular fat.
Magnetization prepared 3D gradient echo (3D MP-RAGE) sequence was modified to utilize sequential
phase encoding order in the slice selection direction and fat saturation RF application after nonselective
inversion preparation. The inversion time (TI), the delay time from inversion to the center of k space in
the slice direction is controlled by proper selection of the number of slices to be imaged, thereby
maximizing the efficiency of inversion prepared acquisition On the basis of the simulation, the
optimized imaging parameters were determined and then selected to achieve optimal levels of blood
suppression and maximum contrast between intraplaque hemorrhage and vessel wall. Heavily T1w
images were acquired with the optimized 3D MP-RAGE with TR = 650 ms, TI = 350 ms, TE = 1.5 ms,
and segmentation=42, spatial resolution of 0.5x0.5x1 mm3.
Ex vivo Imaging Studies: 3D multishot (ms) Inner Volume (IV) Diffusion Weighted (DW) EPI
technique was implemented from the 3D segmented EPI. Figure 1 B shows the pulse sequence diagram
of 3D ms-IV-DWEPI. The slice-selection gradient of the 180o pulse is applied in the phase-encoding
direction to achieve the inner volume imaging. Eight carotid endarterectomy specimens were obtained
from surgery and imaged at same day. They were immersed in saline solution on the test tube and
maintained at the room temperature during MR scan. The fresh specimens were imaged with a
commercial wrist coil on the same 3T system. Ex –vivo 3D T1w, T2w images were obtained by using
3D TSE sequence with the following parameters: 0.3 mm isotropic voxel dimension, ETL=15,
TE/TR=9.2 or 63/ 700 or 2000 ms for T1w or T2w, respectively. 3D MPRAGE images were acquired
with same parameters used for in vivo studies except 0.3 mm voxel dimension. 3D DWI of specimens
were acquired by using 3D ms-IV-DWEPI with following parameters; b=10, 1000 mm2/sec, ETL=5, 0.5
mm isotropic voxel dimension. After MRI, the endarterectomy samples were fixed in formalin for
histology.

Carotid Endarterectomy (CEA) Processing and Evaluation

After ex vivo MR scan the specimens were fixed in 10% neutral buffered formalin for 3 days. The ratio of fixative
to specimen was at least 10:1. Specimens were decalcified in 1% EDFTM (Enhanced Decalcification Formulation).
The sections were submitted individually and sequentially from inferior to superior in plastic tissue cassettes.
Tissue cassettes were processed on the automated Sakura Vacuum Infiltrating Processor (VIP), embedded in
paraffin wax, sectioned at 3-4 mm intervals and stained with a combined Mason elastic (CME) and hematoxylin
and eosin (H&E). Plaque components were traced as regions of interest and histologic slides were matched with
in vivo MR images. The carotid bifurcation and gross morphology were used as anatomic guides between MRI
and histological cross sections. Major components in the selected sections of histologically stained tissues were
classified as fibrous cap (FC), fibrosis mainly consisting of collagen fibers, calcification, myxomatous tissue rich
in smooth muscle cells and extracellular matrix such as lipid core. Some of lipid core component were overlapped
with fresh intraplaque hemorrhage. Fresh intraplaque hemorrhage was defined as plaque components containing
hemorrhagic debris and intact or degenerating red blood cells, but without a histologic tissue reaction. The
matching histological cross sections were also classified for necrosis, calcification and hemorrhage.

ADC Measurements
In vivo ADC Measurement: The ADC maps were calculated using the images from the two b-
values acquired. Images with b=10 and b=300 mm2/sec were acquired in two sets and
magnitude images at each b-value were averaged. Using two images (S0, S1) from the two b-

values (b0, b1), ADC maps were calculated using the following equation
(1)
and displayed using algorithms developed in IDL (ITT Visual Information, Boulder, CO).
A total of 128 slices of ADC maps were created from in vivo DW images obtained from the eight
subjects. For each slice location, the ADC maps and corresponding T1w, T2w, and MPRAGE
images were displayed. In vivo MR images including ADC maps were manually compared
and registered to the histologic findings in all eight specimens. Slice locations of MR images
and ADC maps were matched to the section of plaque identified on histology by reviewing
histologic features based on distance from the bifurcation of the common carotid artery and
the size and shape of the lumen and plaque. Thirty two axial ADC maps which had
corresponding histology from carotid endarterectomy were selected for the measurement of
the mean ADC values in vessel wall, lipid core and hemorrhage.
There were eighteen axial locations on ADC maps that had hemorrhage with no lipid core. While
lipid core component without hemorrhage was seen on eight slices of ADC maps, lipid core
with hemorrhage was found on six maps. A total of sixty two regions of interests (ROIs) were
manually drawn on those thirty two slices of ADC maps (fourteen ROIs for lipid necrotic core,
twenty four ROIs for hemorrhage, and twenty four ROIs for normal vessel wall, as well).
They were subsequently selected and correlated based upon the conformation, location and
shape of plaque component shown in the histological section and other MR images in the
corresponding location. The presence and location of hemorrhage shown in 2D black blood
T1w and 3D MP RAGE images were reviewed for selection of lipid core ROI. Therefore, the
histology report served as a standard for determining the accuracy of ROI selection on ADC
maps with help from the other MR images. Although the in-plane resolution of the ADC map
is 1.0x1.0 mm2, most ROIs drawings identified on ADC maps typically include only two to
three pixels due to the small size of plaque components. Thus, for each plaque component, the
mean ADC value was obtained by averaging all of the pixel-based ADC values for the
component for all ROI’s rather than averaging the mean value from each ROI.
Ex vivo ADC Measurement: The ex vivo ADC maps were calculated using the images from the two b-values (10
and 1000 mm2/sec) acquired. 3D T1w, T2w and MP RAGE images and ADC maps at the same location were
matched to histological sections. Slices were matched by observing the overall morphology of the plaque, as well
as the known distance of each slice from the common carotid bifurcation. After the ex vivo ADC maps and 3D
MR were matched to the corresponding histological slices, they were registered to the in vivo slices by scaling,
translation and rotation transformation using algorithms developed in IDL. The scaling factor was known from the
relative sizes of the ex vivo and in vivo pixels. Translation and rotation were obtained to match the orientation of
a user-defined straight line that was drawn on both of the ex vivo, in vivo maps and images. Thirty eight ROIs
representing lipid necrotic core or hemorrhage on the corresponding in vivo ADC and histological slices, were
selected. Fourteen and twenty four ROIs were selected for the measures of ex vivo mean ADC in lipid core
hemorrhage, respectively. The location and size of each ROI was carefully selected to match the location and size
of ROI selected on in vivo maps.

Statistical Analysis
Mean and standard deviation (SD) ADC values were computed using all pixels identified by ROI for
each component, in vivo and ex vivo. Overall, one hundred seventy eight pixels on wall tissue, lipid
necrotic core, and hemorrhage were identified by ROI for in vivo measurements. For ex vivo measures,
two hundred sixty six pixels were selected. No standardized techniques were employed to normalize the
ADC value between components such as lipid core, hemorrhage and wall tissue. Quantitative statistical
comparison of ADC values from each plaque component and wall tissue was conducted separately for in
vivo and ex vivo cases using unbalanced 1-way analysis of variance (ANOVA). An unbalanced 2-way
ANOVA was performed comparing both in vivo and ex vivo ADC values from hemorrhage and lipid.
Differences were considered significant when P < 0.05. Box-and-whisker plots were generated for the
in-vivo and ex-vivo samples separately to provide graphical assessment of the component distributions.

RESULTS
Among the 128 axial ADC maps, 32 axial ADC maps were selected for in vivo ADC
measurements. We used the average of 75 pixel based ADC values of wall to calculate the
mean ADC value of wall tissue. It was measured as 1.27±0.16x10-3 mm2/s. To calculate the mean
ADC values of lipid necrotic core and hemorrhage, 14 and 24 ROIs were selected, respectively. Total of thirty
four (lipid necrotic core) and sixty nine (hemorrhage) pixels were found on these 14 and 24 ROI selections.
Therefore, 34 and 69 pixel based ADC values were averaged to calculate the mean ADC values for lipid and
hemorrhage, respectively. The mean ADC values of lipid, hemorrhage wall composition obtained from eight
patient subjects are summarized in Table 1.
Similarly, for ex vivo measures, 34 and 69 ROI selections on the lipid core and hemorrhage components located
at the corresponding 32 ex vivo ADC slices were performed. For ex vivo ADC measurements, the total of 91 and
175 pixel based ADC values obtained from the ROI selections of lipid core and hemorrhage were averaged,
respectively. Table 2 displays the summarization of the mean ADC values of lipid and hemorrhage obtained from
the eight carotid endarterectomy specimens.
Fig. 2 shows the typical example of ROI selection for a lipid core component on in vivo and ex vivo
ADC slices. Figure : (A) 2D T1w, 2D T2w and ADC maps of three contiguous slices from a subject with
a lipid core component. (B) Cropped 2D T1w, T2w, images and ADC map in the same location as in the
first row in (A). (C) Corresponding ex vivo 3D T1w, T2w images and ADC map shown in (B). (D)
Histology of the same location seen in (B and C). Red –Hemorrhage, Blue – Collagen Rich, Green-
Necrosis, Light Blue – Lipid.A displays 2D T1w, 2D T2w images and ADC maps at the same slices
obtained from a patient before surgery. The displayed in vivo ADC maps demonstrate that the arterial
wall and plaque can be seen in most slices. The vessel wall structure that can be seen in the
corresponding T1w and T2w images is clearly shown in ADC maps. The in vivo ADC values near the
plaque area are reduced compared to those of the surrounding vessel wall area. The plaque area in a
single slice indicated by red arrows in Fig. 2A is hyperintense on T1w, intermediate intensity on T2w
and hypointense on the ADC map. The histological findings on the corresponding slice shown in Fig. 2D
confirmed a lipid necrotic core component. The lipid rich core component indicated by light blue color
on the annotated histology slide is spatially well matched with the low signal area of the ADC map. The
cropped images including ADC maps of in vivo and ex vivo plaque with a lipid core on the
corresponding slice are displayed in Fig. 2B and 2C, respectively. Green ROIs in maps of in vivo and ex
vivo plaque were drawn and corrected according to the location and shape of a lipid core seen in
histology. The mean ADC values obtained form in vivo and ex vivo green ROIs were measured as
0.42±0.12 and 0.35±0.09 x10-3 mm2/s, respectively.
Figs. 3A and B display the T1w, T2w, and 3D MP RAGE images and the ADC map obtained from a
patient with intraplaque hemorrhage and a lipid core. The hemorrhage is indicated by the red arrows and
shows a very bright signal on 3D MP RAGE. Histology confirmed that this area represented recent
hemorrhage (red area on Fig. 3D). The lipid necrotic core component also was identified on the bottom
slice location. The ROI drawn by the red line in the in vivo and ex-vivo ADC map of Figs. 3B and C
demonstrates a typical selection for hemorrhage. The mean ADC values obtained by averaging the in
vivo and ex vivo pixel based values within the red ROIs were 0.95±0.23 and 1.21±0.31x10-3 mm2/s,
respectively. The green ROIs indicate the necrotic core confirmed by histology shown Fig. 3D and in
vivo and ex vivo ADC values of 0.51±0.28 and 0.38±0.19 x10-3mm2/s were measured.
Figs. 4 and 5 show the boxplots comparing the distributions of in vivo and ex vivo component ADC
values. Value minima and maxima are shown in black with the lower and upper quartiles indicated in
blue and median values shown in red. In Fig. 5, two points for ex vivo hemorrhage were identified as
possible outliers (distance from mean > 2.7σ) and shown as red crosses. The unbalanced 1-way ANOVA
for the in vivo ADC measurements of hemorrhage, lipid and wall showed a significant difference
between the mean values of these components ( F (2,175)=249, p <0.001). Similarly, for ex vivo ADC
measuremeants of lipid and hemorrhage a significant difference between the mean values was observed
( F(1,264)=1692 , p <0.001). The unbalanced 2-way ANOVA between components both in vivo and ex
vivo showed significant difference between component mean values, ( F(1,365)=1128, p <0.001).
Differences between in vivo and ex vivo measurements were determined to be significant ( F(1,365)=18,
p<0.001). Additionally, an interaction between the component considered and whether measured in vivo
or ex vivo was observed ( F(1,365)=43, p <0.001). Separate post-hoc 1-way ANOVA of hemorrhage
and lipid showed significant dependence of the mean ADC values of hemorrhage upon in vivo vs. ex
vivo (F(1,242)=62, p <0.001), and slightly less but still significant dependence of the mean value of
lipid upon in vivo vs. ex vivo (F(1,123)=9.8, p <0.01).

DISCUSSION
From a comparison of multi contrast carotid MRI sequences including T1w, T2w, MPRAGE and ADC
maps, we demonstrated that ADC maps have a high sensitivity for the identification of lipid rich necrotic
core and hemorrhage. The enhanced detection of lipids and hemorrhage will allow more precise analysis
of the quantity and location of lipid or hemorrhage in relation to other plaque constituents, and allow for
future correlations of these components with clinical data on patients. Most previous DWI studies have
been performed on ex-vivo specimens22,23. DWI is difficult to perform in vivo, especially in the vicinity
of the carotid bifurcation where B0 variation, flow characteristics, and motion pose major challenges.
2D single shot or multishot EPI techniques have been used mainly in intracranial DWI. However, EPI
acquisition of the carotid artery leads to image distortion, drop-out, blurring and signal loss. In an
attempt to reduce the geometric distortion and blurring, we have developed a 2D ss-IMIV DWEPI
technique, in which interleaved multislice inner volume (IMIV) imaging 24 25 is combined with 2D ss-
DWEPI. 2D ss-IMIV-DWEPI sequence can perform high-resolution interleaved multiple slice DWI of
localized regions of the carotid artery. To our knowledge, this study is the first report of in-vivo
measurement of plaque components in DWI of the human carotid artery26.
DWI obtains images weighted by the molecular diffusion of water. The ADC is a measure that
combines the actual diffusion coefficient of the observed water molecules with the geometric restrictions
imposed on molecular motion by barriers such as cell membranes, organelles, and other tissue
microstructures. Measuring displacement of water molecules by calculating ADC can identify the
microstructure of the environment in which these displacements take place. There is evidence that DWI
can accurately discriminate lipid and necrotic core from fibrous tissue27. Lipid molecules, even within a
liquid phase, undergo diffusion much slower than water molecules. Lipids in human breast adipose
tissue, which consist mainly of triglyceride, are characterized by a very low diffusion coefficient16.
Water diffuses slowly within the microenvironment of the atherosclerotic core, where the abundant
lipids can restrict its diffusion relative to that in fibrous tissues. Therefore, the diffusion of lipids within
the plaque is extremely slow. Using 2D ss-IMIV DWEPI, these results demonstrated that ADC values
are much lower in that region than in the other components of normal and diseased arterial walls. The
lower ADC of lipid core to normal wall and hemorrhage was statistically significant with p-value of
p<0.01.
The low contrast to noise ratio (CNR) of T1W and T2W images of the carotid plaques can often lead to
difficulty in correlating MR images with histology in the identification of carotid plaque components. It
is not surprising that previous ex-vivo studies have compared T1w, T2w, and DWI images of carotid
plaque specimens with histology and found that DWI provides the highest CNR for lipid detection.
Reports of the appearance of lipids in carotid plaques in conventional MR images such as T1w and T2w
images have not been consistent. In some studies, image comparisons between different sequences were
inconsistent because of the dependence of relaxation on magnetic field strengths. Various studies have
reported that lipid rich regions are more hypointense than fibrous tissue in T2W images or, conversely
that lipids and fibrous tissue were isointense 11,28,29. An approach that can accurately separate lipid from
other components of atherosclerotic plaque will allow more precise quantification and localization of
lipid in relation to other plaque constituents.
The role of plaque hemorrhage as a potential marker for symptomatic carotid plaque had been
suggested as early as 1982 by Lusby et al. 30. Their retrospective study of carotid endarterectomy
specimens identified a significantly higher association of plaque hemorrhage in symptomatic patients
when compared to asymptomatic patients. Previous ex-vivo studies showed that DWI sequences readily
identified restricted water diffusion in intraplaque thrombus relative to diffusion in the vessel wall of the
rabbit aorta, and the sequences provided excellent contrast between thrombus and normal vessel wall22,14
. To evaluate for plaque hemorrhage in vivo, heavily T1 weighted sequences including 3D MP-RAGE
have been introduced. In the acute/subacute phase of thrombus, formation of methemoglobin results in
shortening of T1, producing an increase in signal on this T1-weighted sequence 31. By properly selecting
the inversion time(TI) based on the T1 relaxation time of blood, the blood signal can be eliminated to
reach the maximum contrast between the lumen and vessel wall. Despite the ability to detect
methemoglobin within intraplaque hemorrhage, 3D MP RAGE is limited in its ability to stage cerebral
hemorrhage such as fresh (early), recent and old32. Further, even with these improvements in MPRAGE,
studies have found that DWI and the calculated ADC greatly enhanced image contrast and allowed
better differentiation of the hemorrhage from the other components of the vessel wall.
A previous ex vivo DWI study reported that ADC in a hemorrhage varies according to the processes
that occur during the successive phases of aging. An early phase occurring after 1 week in vitro showed a
reduction of ADC (0.36x 10-3 mm2/sec). However, larger hemorrhage after partial resolution later
showed increased water diffusion (1.33x 10-3 mm2/sec) 14. In our measurement, ADC values in
hemorrhage ranged from 0.59 to 1.12x 10-3 mm2/sec. This indicates that ADC maps acquired using 2D
ss-IMIV DWEPI may be capable of identifying not only the presence but also the stage of intraplaque
hemorrhage. This would permit in vivo studies correlating the acuity of intraplaque hemorrhage with the
development of ischemic complications. In addition, the detection and acuity determination of
intraplaque hemorrhage and its stage may provide important information on the instability of the plaque.
Future work to further improve the accuracy of ADC maps in the differentiation of the stage of
hemorrhage should improve our ability to determine the risk of subsequent cerebrovascular neurologic
events based on these values.
In previous DWI studies of ex-vivo specimens, higher b values (usually 1000 mm2/sec) were used to
create the ADC maps 14,15,22,27. To achieve this b value, the diffusion weighted gradient lobe is typically
several tens of milliseconds in length, which inevitably leads to a rather long TE. This long TE reduces
the SNR and adds unwanted T2w contrast in in-vivo DWI of carotid artery. To obtain ADC maps with
reasonable SNR, a b-value of 300 sec/mm2 was used in this study. Signal in DWI decays exponentially
with b-value as shown in Eq. (1). ADC values of different plaque components from previous ex vivo
studies have been reported to be at least of order 10-4 mm2/sec. Thus, the signal difference obtained from
DWI with two b values (10 and 300 mm2/sec) provides enough information to differentiate the ADC
values of different plaque components. According to our results the in vivo ADC values were measured
as the similar values obtained from ex vivo DWI with b=10 and 1000 mm2/sec. It also indicated that our
in vivo protocol can provide the reliable ADC measurements of plaque components.
In the spite of the potential utility of DWI in plaque component identification, clinical application of
DWI is hampered by bulk physiological motion and magnetic susceptibility. Our previous work
demonstrated that the 2D ss-IMIV-DWEPI technique can obtain good quality in vivo diffusion weighted
images of the carotid arteries with a spatial resolution of 1x1x2 mm3. In the visual assessment of image
quality, an average of about 75% of the normal vessel wall could be visualized in the resulting ADC
map. Plaque is typically larger and more easily visualized in the ADC map. However, for in vivo
measurement of a lipid core component only 34 pixels were included from 14 ROIs selections, meaning
that only 2 or 3 pixels were covered by most ROI selections. To overcome this limitation, a higher
resolution DWI technique is desirable. Usually, 3D techniques can provide both increased SNR and
spatial resolution. Increased SNR and spatial resolution in in-vivo diffusion weighted imaging can
improve the sensitivity of ADC maps in the plaque component identification. 3D DWI techniques may
further improve the resolution of ADC maps.
The diffusion values we obtained for normal arterial wall (1.27±0.16x10-3 mm2/s) were consistent with
previously reported ex vivo measurements (1.50x10-3 mm2/s)27. A limited range of diffusion values was
observed for lipid (0.38±0.10x10-3 mm2/s) and hemorrhage (0.98±0.25x10-3 mm2/s). These values were
both significantly lower than the ADC of normal wall. In addition, the ADC values of hemorrhage and
lipid were significantly different from each other as observed in the boxplots of Figures 4 and 5 as well
as the ANOVA findings. This is consistent with diffusion measurements in ex vivo studies14,15. Our goal
in this study was not to provide the definitive in vivo measurements for these components, but to show
that such measurements are possible in vivo.
Regarding the ANOVA analysis, there are three important observations. First, while the results support
the conclusion that the measured component mean ADC values were distinct, it should be noted that the
analysis does not support the conclusion that the components would be separable based solely on their
ADC values without some other a priori information such as that provided here by histology. For
example, Fig. 4 suggests that the distribution of in vivo ADC values for hemorrhage and wall show
significant overlap even though their means are statistically distinct. Second, our pixel based analysis
does not account for possible bias in groups of pixels which could lead to low p-value ANOVA results.
The limited numbers of pixels per ROI per in vivo MR image per subject at this point limits the ability
to detect such biases in vivo, but similarly helps mitigate the effect of possible bias upon the global
distribution and subsequent 1-way ANOVA analysis of the in vivo results. We believe that the pixel
counts support our basic finding that the in vivo component mean values of ADC are distinct. In
addition, the ADC values of hemorrhage and lipid were significantly different from each other as
observed in the boxplots of Figs. 4 and 5 as well as the ANOVA findings.
Finally, we consider three possible sources for the significant interaction term of the 2-way ANOVA
and significant difference observed between in vivo and ex vivo measurements. Primarily, we believe
these differences are due to the difference in resolutions between in vivo and ex vivo images.
The lower resolution in vivo images likely suffer from greater partial volume effects. Secondarily, the
differences may represent a true difference between in vivo and ex vivo measurements that will require
further experiments to confirm. Finally, some of the difference might be explained by errors in ROI
placement. Because of changes in specimen shape due to surgery and post-surgical processing, the ex
vivo MR images are more likely to agree with histology than the in vivo images. One possible approach
would be to use other in vivo MR contrasts, with their likely superior co-registration with the in vivo
ADC map, rather than histology to identify components for measurement.

CONCLUSION

This study provides the first reported reference range for the ADC values of overall plaque, lipid and
hemorrhage in vivo. The different diffusion values of the normal carotid wall compared with lower ADC
values in intraplaque lipid and hemorrhage suggest that ADC values may be of substantial value in
plaque component discrimination. We also believe that this technique can be used to further investigate
the ADC of other plaque components. The results obtained indicate that an ADC map may be of
substantial value in identifying lipid and hemorrhage within overall plaque burden.
REFERENCES

1. Fleiner M, Kummer M, Mirlacher M, al. e. Arterial neovascularization and inflammation in

vulnerable patients: early and late signs of symptomatic atherosclerosis. Circulation 2004;110:2843-
2850.
2. Spagnoli LG, Mauriello A, Sangiorgi G, et al. Extracranial thrombotically active carotid plaque as a
risk factor for ischemic stroke. JAMA 2004;292(15):1845-1852.
3. Molloy J, HS. M. Asymptomatic embolization predicts stroke and TIA risk in patients with carotid
artery stenosis. Stroke 1999;30:1440-1443.
4. Kristoffersen A. Diffusion measurements and diffusion tensor imaging with noisy magnitude data. J
Magn Reson Imaging 2009;29(1):237-241.
5. Saam T, Ferguson MS, Yarnykh VL, al. e. Quantitative evaluation of carotid plaque composition by
in vivo MRI. . Arterioscler Thromb Vasc Biol 2005;25:234-239.
6. Fayad Z, Fallon J, Shinnar M, et.al. Noninvasive in vivo high-resolution magnetic resonance
imaging of atherosclerotic lesions in geneticallt engineered mice. Ciculation 1998;98:1541-1547.
7. Demarco JK, Rutt BK, Clarke SE. Carotid plaque characterization by magnetic resonance imaging:
review of the literature. Top Magn Reson Imaging 2001;12(3):205-217.
8. Chu B, Kampschulte A, Ferguson MS, al. e. Hemorrhage in the atherosclerotic carotid plaque: a
high-resolution MRI study. Stroke 2004;35:1079-1084.
9. Cai J, Hatsukami T, et.al. Classification of human carotid atherosclerotic lesions in vivo
muticontrast magnetic resonance imaging. Ciculation 2002;106:1368-1373.
10. Gortler M, Goldmann A, Mohr W, B. W. Tissue characterisation of atherosclerotic carotid plaques
by MRI. . Neuroradiology 1995;37:631-635.
11. Shinnar M, Fallon J, Wehri S, al e. The diagnostic accuracy of ex vivo MRI for human
atherosclerotic plaque characterization. Arterioscler Thromb Vasc Biol 1999;19:2756-2761.
12. Toussaint JF, Southern JF, Fuster V, HL. K. T2-weighted contrast for NMR characterization of
human atherosclerosis. . Arterioscler Thromb Vasc Biol 1995;15 1533-1542.
13. Bammer R. Basic principles of diffusion-weighted imaging. Eur J Radiol 2003;45(3):169-184.
14. Toussaint JF, Southern JF, Fuster V, Kantor HL. Water diffusion properties of human atherosclerosis
and thrombosis measured by pulse field gradient nuclear magnetic resonance. Arterioscler Thromb
Vasc Biol 1997;17(3):542-546.
15. Viereck J, Ruberg FL, Qiao Y, et al. MRI of atherothrombosis associated with plaque rupture.
Arterioscler Thromb Vasc Biol 2005;25(1):240-245.
16. Tanner JE. Intracellular diffusion of water. Arch Biochem Biophys 1983;224(2):416-428.
17. Kim SE, Jeong EK, Shi XF, Morrell G, Treiman GS, Parker DL. Diffusion-weighted imaging of
human carotid artery using 2D single-shot interleaved multislice inner volume diffusion-weighted
echo planar imaging (2D ss-IMIV-DWEPI) at 3T: diffusion measurement in atherosclerotic plaque.
J Magn Reson Imaging 2009;30(5):1068-1077.
18. Jeong EK, Kim SE, Guo JY, Kholmovski EG, DL. P. High-resolution DTI with 2D interleaved
multislice reduced FOV singleshot diffusion-weighted EPI (2D ss-rFOV-DWEPI). Magn Reson
Med 2005;54:1575-1579.
19. Hadley JR, Roberts JA, Goodrich KC, Buswell HR, Parker DL. Relative RF coil performance in
carotid imaging. Magn Reson Imaging 2005;23(5):629-639.
20. Kim SE, Kholmovski EG, Jeong EK, Buswell HR, Tsuruda JS, Parker DL. Triple contrast technique
for black blood imaging with double inversion preparation. Magn Reson Med 2004;52(6):1379-
1387.
21. Parker D, Goodrich K, Masiker M, Tsuruda J, et.al. Improved efficiency in double-inversion fast
spin-echo imaging. Magn Reson Med 2002;47:1017-1021.
22. Clarke SE, Hammond RR, Mitchell JR, Rutt BK. Quantitative assessment of carotid plaque
composition using multicontrast MRI and registered histology. Magn Reson Med 2003;50(6):1199-
1208.
23. Clarke SE, Beletsky V, Hammond RR, Hegele RA, Rutt BK. Validation of automatically classified
magnetic resonance images for carotid plaque compositional analysis. Stroke 2006;37(1):93-97.
24. Feinberg D, Hoenninger J, Crooks L, Kaufman L, Watts J, M. A. Inner volume MR imaging:
technical concepts and their application. Radiology 1985;156:743-747.
25. Conturo TE, Price RR, AH. B. Rapid local rectangular views and magnifications: reduced phase
encoding of orthogonally excited spin echoes. Magn Reson Med 1988;6(4):418-429.
26. Kim TH, Zollinger L, Shi XF, et al. Quantification of Diffusivities of the Human Cervical Spinal
Cord Using a 2D Single-Shot Interleaved Multisection Inner Volume Diffusion-Weighted Echo-
Planar Imaging Technique. AJNR Am J Neuroradiol 2009.
27. Qiao Y, Ronen I, Viereck J, Ruberg FL, Hamilton JA. Identification of atherosclerotic lipid deposits
by diffusion-weighted imaging. Arterioscler Thromb Vasc Biol 2007;27(6):1440-1446.
28. Underhill HR, Yarnykh VL, Hatsukami TS, et al. Carotid plaque morphology and composition:
initial comparison between 1.5- and 3.0-T magnetic field strengths. Radiology 2008;248(2):550-
560.
29. Cappendijk VC, Kessels AG, Heeneman S, et al. Comparison of lipid-rich necrotic core size in
symptomatic and asymptomatic carotid atherosclerotic plaque: Initial results. J Magn Reson
Imaging 2008;27(6):1356-1361.
30. Lusby RJ, Ferrell LD, Ehrenfeld WK, Stoney RJ, Wylie EJ. Carotid plaque hemorrhage. Its role in
production of cerebral ischemia. Arch Surg 1982;117(11):1479-1488.
31. Moody AR, Murphy RE, Morgan PS, et al. Characterization of complicated carotid plaque with
magnetic resonance direct thrombus imaging in patients with cerebral ischemia. Circulation
2003;107(24):3047-3052.
32. Chu B, Kampschulte A, Ferguson MS, et al. Hemorrhage in the atherosclerotic carotid plaque: a
high-resolution MRI study. Stroke 2004;35(5):1079-1084.
Figure Captions

Figure : Pulse Sequence diagram of 2D ss-IMIV DWEPI (A) and 3D ms-IV-DWEPI (B).

Figure : (A) 2D T1w, 2D T2w and ADC maps of three contiguous slices from a subject with a lipid core
component. (B) Cropped 2D T1w, T2w, images and ADC map in the same location as in the first row in
(A). (C) Corresponding ex vivo 3D T1w, T2w images and ADC map shown in (B). (D) Histology of the
same location seen in (B and C). Red –Hemorrhage, Blue – Collagen Rich, Green- Necrosis, Light Blue
– Lipid.

Figure 3 (A) 2D T1w, T2w, 3D MPRAGE images and ADC maps of two contiguous slices obtained
from a subject with intraplaque hemorrhage and lipid core. (B) Cropped 2D T1w, T2w, 3D MPRAGE
and ADC map of same slice location as in the second column in (A). (C) Corresponding 3D T1w, T2w,
MPRAGE and ADC map of ex vivo plaque. The red and green lines shown in both of in vivo and ex
vivo ADC maps present the ROI selection for hemorrhage and lipid core confirmed by histology,
respectively. (D) Histology of the same location seen in (B and C).

Figure 4 in vivo ADC values for lipid necrotic core, hemorrhage and wall tissue. Standard boxplot
format shows the lower and upper quartiles in blue, range in black, and median in red.

Figure 5 ex vivo ADC values for lipid necrotic core and hemorrhage. Standard boxplot format shows the
lower and upper quartiles in blue, range in black, and median in red. Red crosses indicate possible
outliers greater than 2.7σ away from the mean.

Table 1 Mean ADC

value of normal
arterial wall and in
vivo plaque
components
ADC value(10-3 mm2/s) Wall Lipid Necrotic Core Hemorrhage
MEAN 1.27 0.38 0.98
STDV(Standard Deviation) ±0.16 ±0.10 ±0.25
Number of Pixels 75 34 69
Number of ROIs 24 14 24

Table1: Mean ADC values of normal arterial wall and plaque components
Table 2 Mean ADC
value of ex vivo plaque
components
ADC value(10-3 mm2/s) Lipid Necrotic Core Hemorrhage
MEAN 0.33 1.23
STDV(Standard Deviation) ±0.08 ±0.20
Number of Pixels 91 175
Number of ROIs 14 24
Table2: Mean ADC values of ex vivo plaque components

Figure : Pulse Sequence diagram of 2D ss-IMIV DWEPI (A) 3D ms-IV DWEPI (B).
Figure : (A) 2D T1w, 2D T2w and ADC maps of three contiguous slices from a subject with a lipid core
component. (B) Cropped 2D T1w images, T2w images, and ADC map in the same location as in the first
row in (A). (C) Corresponding ex vivo 3D T1w images, T2w images, and ADC map shown in (B). (D)
Histology of the same location seen in (B and C). Red –Hemorrhage, Blue – Collagen Rich, Green-
Necrosis, Light Blue – Lipid.

Figure 3 (A) 2D T1w, T2w, 3D MPRAGE images and ADC maps of two contiguous slices obtained
from a subject with intraplaque hemorrhage and lipid core. (B) Cropped 2D T1w, T2w, 3D MPRAGE
images and ADC map of same slice location as in the second column in (A). (C) Corresponding 3D
T1w, T2w, MPRAGE images, and ADC map of ex vivo plaque. The red and green lines shown in both
the in vivo and ex vivo ADC maps present the ROI selection for hemorrhage and lipid core confirmed
by histology, respectively. (D) Histology of the same location seen in (B and C).
Figure 4 in vivo ADC values for lipid necrotic core, hemorrhage and wall tissue. Standard boxplot
format shows the lower and upper quartiles in blue, range in black, and median in red.

Figure 5 ex vivo ADC values for lipid necrotic core and hemorrhage. Standard boxplot format shows the
lower and upper quartiles in blue, range in black, and median in red. Red crosses indicate possible
outliers greater than 2.7σ away from the mean.