Available via license: CC BY-NC-ND
Content may be subject to copyright.
RESEARCH ARTICLE
Heavy macrophage infiltration identified by optical
coherence tomography relates to plaque rupture
Xuan Shi
1,
* , Tao Tao
2,
*, Yi Wang
2
, Yunfei Han
1
, Xiaohui Xu
1
, Qin Yin
1
, Fang Wang
1
, Rui Liu
1
& Xinfeng Liu
1,3
1
Department of Neurology, Nanjing Jinling Hospital, Affiliated Hospital of Medical School, Nanjing University, Nanjing, China
2
Department of Neurosurgery, Nanjing Drum Tower Hospital, Affiliated Hospital of Medical School, Nanjing University, Nanjing, China
3
Stroke Center and Department of Neurology, First Affiliated Hospital of University of Science and Technology of China, Division of Life Sciences
and Medicine, University of Science and Technology of China, Hefei, China
Correspondence
Xinfeng Liu and Rui Liu, Department of
Neurology, Nanjing Jinling Hospital, Affiliated
Hospital of Medical School, Nanjing
University, 305 East Zhongshan Rd, Nanjing
210002, China. Tel +86 2584801861; Fax
+86 2584805169; E-mail: xfliu2@vip.163.
com (X. L.) and E-mail: liurui8616@163.com
(R. L.)
Received: 1 July 2023; Revised: 31 August
2023; Accepted: 2 October 2023
doi: 10.1002/acn3.51923
*These authors contributed equally to
this work.
Abstract
Objective: Risk stratification plays a critical role in patients with asymptomatic
carotid atherosclerotic stenosis. Heavy macrophage infiltration (HMC) is an
important factor of plaque destabilization. However, in vivo imaging technolo-
gies and screening criteria for HMC remain limited. We aimed to (i) introduce
algorithms for in vivo detection of macrophage infiltrations using optical coher-
ence tomography (OCT) and (ii) to investigate the threshold of HMC and its
association with plaque vulnerability. Methods: Ex vivo OCT images were co-
registered with histopathology in 282 cross-sectional pairs from 19 carotid end-
arterectomy specimens. Of these, 197 randomly selected pairs were employed to
define the parameters, and the remaining 85 pairs were used to evaluate the
accuracy of the OCT-based algorithm in detecting macrophage infiltrations.
Clinical analysis included 93 patients receiving carotid OCT evaluation. The
prevalence and burden of macrophage infiltration were analyzed. Multivariable
and subgroup analysis were performed to investigate the association between
HMC and plaque rupture. Results: The sensitivity and specificity of algorithm
for detecting macrophage infiltration were 88.0% and 74.9%, respectively. Of
93 clinical patients, ruptured plaques exhibited higher prevalence of macro-
phage infiltration than nonruptured plaques (83.7% [36/43] vs 32.0% [16/50],
p<0.001). HMC was identified when the macrophage index was greater than
60.2 (sensitivity =74.4%, specificity =84.0%). Multivariable analysis showed
that HMC and multiple calcification were independent risk factors for non-
lipid-rich plaque rupture. Interpretation: This study provides a novel approach
and screening criteria for HMC, which might be valuable for atherosclerotic
risk stratification.
Introduction
Carotid atherosclerotic stenosis remains an important
cause of ischemic stroke. About 10%–15% of all first ever
stroke patients will have an unheralded ischemic stroke
from a previously untreated asymptomatic carotid artery
stenosis.
1,2
The current treatment of patients with asymp-
tomatic carotid atherosclerotic stenosis is based on the
degree of stenosis.
3
However, asymptomatic patients with
high-risk plaques had significantly higher risk of ipsilat-
eral ischemic cerebrovascular events than those without
high-risk plaques with similar degrees of stenosis.
4
In this
context, risk stratification of carotid atherosclerotic pla-
ques becomes critical. Macrophage infiltration (MC) plays
a pivotal role in both atherogenesis and destabilization of
carotid plaques.
5
Histopathological study has shown that
the abundance of MC is associated with higher predicted
stroke risk.
6,7
The assessment of the intraplaque MC bur-
den seems inseparable from risk stratification of carotid
atherosclerotic plaques. Nevertheless, current imaging
technologies of in vivo identify heavy macrophage infiltra-
tions (HMC) remain limited. Traditional imaging
ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and
distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
1
modalities such as ultrasound, multidetector-row CT
angiography, and MR angiography, with the aid of con-
trast enhancement, could provide indications of the pres-
ence of MC by revealing plaque enhancement. However,
their specificity is somewhat constrained.
8
Recently,
18
F-
fluorodeoxyglucose PET/CT and PET/MR have displayed
potential in detecting active inflammation within
plaques,
7,9
but they are incapable of evaluating vascular
anatomy, plaque composition, and additional morpholog-
ical characteristics.
Optical coherence tomography (OCT), a real-time
intravascular imaging technique with ultra-high resolu-
tion, can assist in assessing the microarchitecture of ves-
sels, including vasa vasorum, cholesterol crystals, erosion,
and more.
10–12
In the last decade, studies performed by
OCT have allowed us to shed new light on the pathologic
substrate of cerebrovascular disease.
13–15
What’s more,
MC could be visible on OCT images as signal-rich, dis-
tinct, or confluent punctate regions that exceed the inten-
sity of background speckle noise.
16
The overall superiority
of OCT provides us with a reliable imaging tool to screen
for HMC and help achieve plaque risk stratification.
Therefore, the purpose of this study was two-fold: (i) to
introduce and validate a set of OCT-based algorithms for
automatic assessment of MC and (ii) to determine the
threshold of HMC and certify the association between
HMC and plaque rupture (PR).
Material and Methods
Study design
The focus of this study was two-fold: (i) histological vali-
dation of OCT-based automatic MC identification algo-
rithms (OCT-MCI) in carotid endarterectomy (CEA)
specimens; (ii) quantitative assessment of MC in vivo in
human, and exploration the risk threshold of HMC and
its association with PR (Fig. 1). This study was approved
by the Ethics Committee of Jinling Hospital and Nanjing
Drum Tower Hospital. Witten informed consent was
obtained from all patients prior to CEA and/or OCT
assessment.
CEA specimens
Nineteen specimens from patients who underwent CEA
for severe (≥70%) atherosclerotic internal carotid artery
stenosis were collected between June 2019 and June 2021
in the Department of Neurosurgery at Nanjing Drum
Tower Hospital. OCT examinations were conducted
within 72 hours postoperatively. Following OCT image
acquisition as described below, the specimens were
pressure-fixed by formalin and decalcified with
ethylenediaminetetraacetic acid to maintain their orienta-
tion and size for comparison with OCT images.
17,18
These
specimens were serially sectioned to the longitudinal axis
of the vessel at 1-mm block. From the distal side of each
1-mm block, histopathology slides of 4-μm at 16-μm
intervals were prepared. Slides were serially selected at
0.1-mm intervals and stained with HE and Movat penta-
chrome, respectively. Each histopathology slide was digi-
tized using a microscope at low magnification (×1.25)
and scanned using the 3D Panoramic Viewer system
when necessary.
Patients in the clinical study
We conducted a retrospective review of cases that under-
went OCT examination of the internal carotid artery at
Jinling Hospital from January 2017 to June 2022. The
inclusion criteria were as follows: 1) the presence of an
atherosclerotic stenosis lesion; and 2) the target vessel had
not undergone endovascular treatment (balloon dilation
or stenting) or endarterectomy before OCT examination.
The exclusion criteria consisted of nonanalyzable images,
which were defined as such if more than 3/4 of target seg-
ment could not be visualized due to serious artifact or
intraluminal blood.
19
A total of 102 cases met the inclu-
sion criteria, of which nine cases were classified as having
nonanalyzable images. Ultimately, 93 cases were included
in the clinical study. Baseline information and clinical
data of patients during hospitalization were collected and
recorded, including age, gender, history of hypertension,
diabetes, coronary heart disease, stroke, current smoking,
and alcohol consumption, as well as biochemical
parameters.
OCT image acquisition
OCT images were acquired using a frequency domain
OCT system (ILUMEN OPTIS System or C7-XR, St. Jude
Medical, USA) and a 2.7-F Dragonfly OCT Imaging Cath-
eter (C7 Dragonfly Catheter or Dragonfly Duo Catheter,
St. Jude Medical, USA). In the histopathology study, the
OCT imaging catheter was guided by a guidewire through
the CEA specimen (Fig. S1). After the images were cali-
brated for Z-bias, an automatic retraction at a speed of
25 mm/s was performed to obtain a serial set of OCT
images. In the clinical study, the OCT imaging catheter
was delivered distal to the stenotic lesion in the internal
carotid artery under the guidance of a 0.3556 mm (0.014
in) PT microwire (Fig. S1). Images were acquired by
injecting 20 ml of 100% contrast medium at a flow rate
of 10 ml/s through the guiding catheter to flush out the
blood, with the pressure set at 200 psi (pounds per square
inch, 1 psi =6.895 kPa). Calibration was completed, and
2ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.
Macrophage Infiltration and Plaque Vulnerability X. Shi et al.
images were obtained after an automatic retraction. All
images were stored in digital format using proprietary
software (St. Jude Medical, USA) for subsequent offline
analysis.
OCT image analysis
OCT images were independently analyzed by two experi-
enced OCT investigators who were blinded to the clinical
data. In cases where there was disagreement between the
two results, re-evaluation was performed by a third OCT
investigator, and the result was recorded after consensus
was reached by all three. Lipid plaque was defined as a
diffusely bordered signal-poor region with an overlying
signal-rich band. Lipid-rich plaque was identified if the
arc of >90 °within a plaque. PR was characterized by the
discontinuity of the fibrous cap or/and cavitation within
the plaque. Calcification was defined as an area with low
backscattering signal and a sharp border inside of a
plaque.
20
The number of calcifications within each plaque
was recorded, and they were then divided into single and
multiple calcifications. Thrombus appeared as a backscat-
tering floating in or protruding into the carotid lumen
with a dimension of at least 250 μm. Neovascularization
Figure 1. Study design. CEA, carotid endarterectomy. MC, macrophage infiltration. OCT, optical coherence tomography. OCT-MCI, OCT-based
automatic MC identification algorithms.
ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 3
X. Shi et al. Macrophage Infiltration and Plaque Vulnerability
was defined as signal-absent holes within a plaque
between 50 and 300 μm in diameter, and visible for at
least three consecutive frames on pullback imaging.
10
Cholesterol crystals were defined as thin linear structures
with high signal and low signal attenuation.
12
OCT and histology co-registration
A total of 12,982 histopathology slides were obtained, and
image matching was achieved by one OCT investigator
and one pathology investigator. The thickness of each his-
tology slice was 4 μm with a 16 μm interval between two
slices. The thickness of each OCT cross-sectional frame
was 100 μm, allowing for approximately 5 histological
slices per 1 OCT frame. The electronic histological images
were first sequentially picked out at 1-mm interval (about
50 slices, 10 OCT cross-sectional frames). Preliminary
pairing of histological images with OCT images was per-
formed according to the site of carotid bifurcation, the
size and morphology of the lumen contour, and the char-
acteristics of the plaque, to roughly confirm the matching
range and serial number.
18
Subsequently, secondary exact
matching was performed at 0.1-mm intervals (approxi-
mately 5 slices, 1 OCT cross-sectional frames). The
matched group was defined as an OCT cross-sectional
frame and several pairing histological slices. Finally, there
were 1559 matched groups, which included 1559 OCT
cross-sectional frames and 6951 histological slices. Con-
sidering potential morphological and compositional simi-
larities between two consecutive OCT frames, the
matched groups were selected sequentially at 0.5-mm
intervals (5 OCT cross-sectional frames). For each
matched group, one slice with the best lumen morphol-
ogy was chosen for CD 68 immunohistochemical staining.
Slices were electronically stored using the 3D Panoramic
Viewer system. In the end, a total of 311 pairs of 1:1
matched OCT images and CD 68 immunohistochemical
stained histology images were obtained.
Considering the inevitable extrusion and deformation
during specimen sectioning, the “Big Wrap” plug-in in
Fiji software was used for image registration.
21
The
“Color Deconvolution” function and “Threshold (auto-
set)” function of Fiji software were utilized to extract the
CD 68 +regions of each slice. If the CD 68 +region was
less than 0.5%, the slice was considered to have no MC
region. Thus, a total of 282 matched pairs were included
in the analysis. The extracted CD 68 +region was then
assigned to the matched OCT image in the same propor-
tion to obtain the histology-based classification images
(Fig. S2). According to these histology-based classification
images, the accuracy of OCT-MCI classification was
tested by computational binary vector methods (Supple-
mentary Methods).
From the 282 pairs of histopathological cross section
and corresponding OCT image, 197 pairs were randomly
selected for a training dataset and used to investigate the
best thresholds of parameters in the OCT-MCI. Subse-
quently, the remaining 85 pairs were used as testing data-
set. The OCT-MCI algorithm whose parameters were
determined based on the training dataset was prospec-
tively applied for the testing dataset.
OCT-based MC identification
Python 3.10 software was used for OCT image preproces-
sing and MC identification. The main objectives of OCT
raw image preprocessing were to improve image quality
and automatically extract the effective-analyzed regions
(ER). Speckle noise reduction was performed using
median filtering, mean filtering, Gaussian filtering, and
bilateral filtering. The effect of speckle noise removal was
evaluated by contrast to noise ratio and equivalent num-
ber of looks, with bilateral filtering showing the best
results. Dynamic programming (DP) was utilized for
catheter guidewire removal and automatic identification
of lumen boundaries. To ensure the accuracy and repeat-
ability of the algorithm in identifying lumen boundaries,
dice similarity coefficient (DSC) and Jaccard coefficient
were used to compare manual segmentation results by
OCT researchers and automatic segmentation results by
the algorithm.
DSC is used to measure the similarity of a set, and a
higher value represents a higher similarity of the set.
22
DSC A,MÞð¼
2ATM
jj
AjjþMjj
Jaccard is used to evaluate the similarity and difference
between the two datasets.
23
Jaccard A,MÞð¼
ATM
jj
ASM
jj
Fifteen pullbacks were randomly chosen from the OCT
image dataset, and 150 images (10 frames per pullback)
were randomly selected. Lumen boundaries were manu-
ally delineated by two independent OCT investigators for
each image set. The results indicated that the agreement
between Investigator 1 and Investigator 2 was high, and
their respective results were highly correlated with the
results of DP algorithm (Table S1). Therefore, the DP
algorithm was determined to be an effective algorithm for
vessel lumen detection. Considering the penetration depth
of the OCT itself around 1–1.3 mm, the ER of the carotid
OCT images was defined as the region where the identi-
fied lumen boundaries expand 1 mm outward.
Within the ER of each cross-sectional frame, the result-
ing OCT raw-intensity image was normalized from 0 to
4ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.
Macrophage Infiltration and Plaque Vulnerability X. Shi et al.
1. The normalized standard deviation (NSD) was calcu-
lated by applying a standard deviation spatial filter:
NSDROI ¼σROI
SmaxSmin
σ2
ROI ¼1
N1∑x∑ySx,y
ðÞ
S
2;Sx,y
ðÞ
∈ROI
where S
max
and S
min
are the maximum and minimum
intensity values of the ER, Nis the total number of the
pixels in the region of interest (ROI), S(x,y)istheOCT
sign as a function of x and y locations within the ROI, and
Sis the mean value of the OCT signal within the ROI.
24,25
Normalized standard deviation ratio (NSDR) was calculated
as the ratio of NSD values of the pixel (ROI1, punctuated
signal-rich areas) and the pixel located at a predefined span
deeper in the axial direction (ROI2, signal-attenuated areas
or shadows).
26
The NSDR value was computed as
NSDRatio ¼NSDROI1
NSDROI2
The optimal parameter values of ROI’s span and ROI’s
size were determined in the histology-based training data-
set. The performance of the algorithm was evaluated in
the histology-based testing dataset.
OCT-based MC quantitative evaluation
A series of consecutive OCT cross-sectional images
acquired from one case was performed for automated
quantitative assessment of MC. Based on the prevalence
of MC, the case was classified as having or not having
MC. The total number of MC-identified frames was
recorded. The longitudinal length of MC was defined as
OCT frame thickness multiplied by the total number of
MC-identified frames. The algorithm automatically calcu-
lated the MC burden (MC%
slice
) for each frame, and out-
putted the mean MC burden (MC%
plaque
) and maximum
MC burden (MC%
Max
) for each plaque. The MC index
for each plaque was obtained by multiplying the mean
MC%
plaque
and the longitudinal length. A receiver operat-
ing characteristic curve (ROC) was performed to deter-
mine the threshold of the MC index for predicting PR.
Then, HMC was defined as plaques with an MC index
larger than this threshold.
Statistical analysis
Clinical statistical analyses were carried out using IBM
SPSS 23.0 software (IBM, Armonk, NY) and R version
4.2.3 (R Development Core Team, Vienna, Austria). Cate-
gorical variables were presented as frequencies (percent-
ages) and compared between groups using the chi-square
test or Fisher’s exact test. The distribution of continuous
variables was tested by the Kolgormonov–Smirnov test.
Continuous variables with normal distribution were
described as mean SD, and comparisons were per-
formed using an independent samples t-test. Variables
with skewed distribution were described as median (the
25th to the 75th percentile), and comparisons were per-
formed using Mann–Whitney U-test. Interobserver and
intra-observer reliability was assessed using Cohen Kappa
test for categorical variables and intraclass correlation test
for continuous variables.
ROC analysis was utilized to determine the best cutoff
value of MC index of each plaque to predict PR. Interac-
tion terms were used to explore whether the association
between HMC and PR differed according to MLA,
lipid-rich plaque, cholesterol crystal, and calcification. A
p-value of less than 0.05 was regarded as indicating
interaction on the multiplicative scale. The relative excess
risk due to interaction (RERI), attributable proportion
due to interaction (AP), the synergy index (S), and
corresponding 95% confidence intervals were employed as
measures of additive interaction. Multivariate analysis was
conducted to assess the independent impact of HMC on
PR within non-lipid-rich plaques by adjusting for vari-
ables including MLA <9.36, with cholesterol crystals and
with multiple calcifications. The significance level was set
to p<0.05, and all tests of hypotheses were 2-sided.
Results
Parameter optimization and performance of
OCT-based MC identification
In the examination of the histological training dataset, the
accuracy of the OCT-MCI, measured by the area under
the curve area (AUC), varied depending on the size and
span of the ROI (Fig. 2). The maximum AUC of 0.928
was achieved when the ROI size was 21 pixels and the
ROI span was 230 pixels. The accuracy, sensitivity, and
specificity for OCT of identifying MC was 89.0%, 95.6%,
and 80.9%, respectively. Subsequently, the performance of
the algorithm was evaluated in the histological testing
dataset. As a result, with histology as the gold standard,
the accuracy of the OCT-MCI was 82.3%, with sensitivity
of 88.0% and specificity of 74.9%.
MC and plaque characteristics
In the clinical analysis, out of 93 carotid atherosclerotic
stenotic lesions, 55.9% (52/93) of plaques were detected
with MC by OCT-MCI. Analysis of plaque type via OCT
revealed that plaques with MC were more frequently clas-
sified as lipid-rich plaques (67.3% [35/52] vs 4.9% [2/41],
p<0.001; Table 1). Compared to plaques without MC,
ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 5
X. Shi et al. Macrophage Infiltration and Plaque Vulnerability
those with MC had a smaller minimal luminal area
(8.2 4.7 vs 10.9 6.9, p=0.04) and a higher preva-
lence of PR (69.2% [36/52] vs 17.1% [7/41], p<0.001).
Detailed characteristics were assessed between the two
groups. Cholesterol crystals were more commonly
detected in plaques with MC (36.5% [19/52] vs 14.6%
[6/41], p=0.02). In contrast, neovascularization (17.3%
[9/52] vs 14.6% [6/41], p=0.73), intraluminal thrombus
(19.2% [10/52] vs 12.2% [5/41], p=0.36), and calcifica-
tions (44.2% [23/52] vs 29.3% [12/41], p=0.14) were
not significantly different between plaques with or
without MC.
Association between MC and PR
Of 93 atherosclerotic stenotic lesions in the clinical study,
46.2% (43/93) of plaques were detected with fibrous cap
disruption through OCT (Table 2). Among them, 83.7%
(36/43) of ruptured plaques were detected with MC by
the algorithm, whereas only 32.0% (16/50) of nonrup-
tured plaques had MC. In the ruptured plaques, the MC
index was significantly higher than in the nonruptured
plaques (125.2 [55.5–199.4] vs 0 [0–50.0], p<0.001).
The mean MC%
plaque
(19.6 [14.3–23.6] vs 0 [0–11.3],
p<0.001) and longitudinal length (6.0 [3.5–9.4] vs 0
[0–2.1], p<0.001) were also higher in ruptured plaques
compared to nonruptured plaques. The ROC curve was
utilized to determine the threshold of MC index for pre-
dicting PR. The Youden index was largest when the MC
index was 60.2, with the AUC of 0.85, sensitivity of
74.4%, and specificity of 84.0%. Based on the cutoff value
of MC index (>60.2), 40 plaques were classified as HMC
plaques. Of these, 80% (32/40) of HMC plaques were
combined with fibrous cap disruption, while only 16%
Figure 2. (A) Representative case of OCT-based automatic MC detection. (B) Steps of OCT-based automated MC analysis. (C) ROC-AUC value as
a function of ROI size and span for the NSDR-based classification. The asterisk mark indicates optimal ROI size (21 pixels) and span (230 pixels)
values that led to the highest ROC-AUC value. ER, effective-analyzable region; NSD, normalized standard deviation; NSDR, normalized standard
deviation ratio.
6ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.
Macrophage Infiltration and Plaque Vulnerability X. Shi et al.
(8/43) of nonruptured plaques exhibited HMCs
(p<0.001). Additionally, compared to nonruptured pla-
ques, ruptured plaques had a larger lipid core (69.8%
[30/50] vs 14.0% [7/43], p<0.001]), a higher prevalence
of thrombus (30.2% [13/50] vs 4.0% [2/43], p=0.001),
and multiple calcifications (32.6% [14/50] vs 8.0% [4/43],
p=0.003).
Association between HMC and PR: subgroup
analysis
Subgroup analyses were conducted based on plaque char-
acteristics to further investigate the association between
HMC and PR (Fig. 3). Subgroup analyses were based on
the full data set in the clinical study. HMC was shown to
be a risk factor for PR after adjusting for lipid-rich pla-
ques (adjusted odds ratio [OR], 6.34 [95% CI 1.93–
20.85]). Subgroup analysis showed no significant statisti-
cal association between HMC and PR in lipid-rich pla-
ques (OR, 0.83 [95% CI 0.08–8.52]; p=0.88). However,
HMC remained a risk factor for PR in non-lipid-rich pla-
ques (OR, 23.92 [95% CI 3.99–143.22]; p=0.003). There
was a multiplicative interaction between lipid-rich plaques
and HMC (p=0.03), but not additive interaction (RERI,
28.61 [95% CI -118.51 –61.28]; AP, 1.00 [95%
CI -4.41 –2.40]; S, 0.49 [95% CI 0.09–2.78]). In the
other subgroups, both overall and subgroup analyses
showed that HMC represented as a risk factor for PR,
including MLA, cholesterol crystals, and calcifications,
and none of interactions were found.
Association between HMC and PR in
non-lipid-rich plaques
In the clinical analysis, variables that were found to be
statistically different between ruptured plaques and non-
ruptured plaques in the univariate analysis, including
MLA (<9.36), cholesterol crystals, multiple calcifications,
and HMC, were included in the multivariate model
(Fig. 4). The multivariate analysis revealed that HMC
(OR, 54.15[95% CI 5.98–490.15]; p<0.001) and multi-
ple calcifications (OR, 17.72 [95% CI 1.89–165.77];
p=0.01) were independently associated with non-lipid-
rich plaque rupture, whereas MLA (<9.36) and
Table 1. Comparison of clinical character-
istic and OCT findings based on the pres-
ence of MC detected by OCT-MCI.
Patients with MC
(n=52)
Patients without MC
(n=41) pvalue
Demographic characteristics
Age, years 68 (61–72) 66 (58–69) 0.13
Male sex, n (%) 18 (34.6) 16 (40.0) 0.60
Body mass index, kg/m
2
25.1 3.1 25.0 2.7 0.81
Clinical features, n(%)
Hypertension 41 (78.8) 32 (80.0) 0.89
Diabetes mellitus 20 (58.8) 14 (41.2) 0.73
Prior stroke/TIA 16 (30.8) 12 (30.0) 0.94
Coronary heart disease 15 (28.8) 9 (22.5) 0.49
Current smoking 21 (40.4) 15 (37.5) 0.78
Biochemical parameters
Triglycerides, mmol/L 1.4 (1.0–2.0) 1.2 (1.1–1.7) 0.63
Total cholesterol, mmol/L 3.4 (2.9–4.3) 3.5 (2.8–4.1) 0.91
HDL, mmol/L 1.0 (0.9–1.1) 1.0 (0.9–1.1) 0.50
LDL, mmol/L 1.8 (1.5–2.4) 2.0 (1.4–2.4) 0.68
Creatinine, μmol/L 66.0 (55.9–78.8) 60.6 (55.8–70.2) 0.09
Serum glucose, mmol/L 5.6 (4.8–7.1) 5.4 (4.9–6.3) 0.33
OCT findings
MLA, mm
2
8.2 4.7 10.9 6.9 0.04
Plaque rupture, n(%) 36 (69.2) 7 (17.1) <0.001
Lipid-rich plaque, n(%) 35 (67.3) 2 (4.9) <0.001
Thrombus, n(%) 10 (19.2) 5 (12.2) 0.36
Cholesterol crystals, n(%) 19 (36.5) 6 (14.6) 0.02
Neovascularization, n(%) 9 (17.3) 6 (14.6) 0.73
Calcifications, n(%) 23 (44.2) 12 (29.3) 0.14
Multiple calcifications, n(%) 12 (23.1) 6 (14.6) 0.31
HDL, high-density lipoprotein; LDL, low-density lipoprotein; MC, macrophage infiltrations; MLA, min-
imum luminal area; TIA, transient ischemic attack.
ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 7
X. Shi et al. Macrophage Infiltration and Plaque Vulnerability
cholesterol crystals were not significant. This suggested
that HMC represented as an independent risk factor for
fibrous cap disruption in non-lipid-rich plaques after
adjusting for multiple associated factors.
Discussion
The main findings of this study were as follows: (1)
OCT-based algorithms for automatic MC identification in
vivo were introduced. The sensitivity and specificity of
OCT-MCI for detecting MC were 88.0% and 74.9%,
respectively. (2) MC was identified in 83.7% of ruptured
plaques. The burden of MC, calculated as MC index, was
significantly higher in ruptured plaques than in
nonruptured plaques. (3) HMC was identified as an inde-
pendent risk factor for PR in non-lipid-rich plaques.
In this study, an algorithm was established for OCT-
based automatic MC identification in vivo using open-
source Python language. Previous study by Tearney’s
team showed a correlation between punctuated signal-rich
bright spots on OCT images and the presence of MC in
fibrous cap of fibroatheromas plaques, and introduced
NSD as an imaging marker for MC identification.
24
NSD
was then used to identify MC in preselected regions on
OCT, which were mainly within fibrous cap or plaque
shoulder.
25,27,28
Researchers have compared the effective-
ness of MC identification based on three different param-
eters: NSD, granulometry index, and signal attenuation,
Patients with PR
(n=43)
Patients without PR
(n=50) pvalue
Demographic characteristics
Age, years 67 (61–70) 66 (60–70) 0.45
Male sex, n(%) 28 (65.1%) 30 (61.2) 0.70
Body mass index, kg/m
2
25.1 3.0 25.0 2.9 0.87
Systolic blood pressure,
mmHg
140.6 19.7 136.2 16.5 0.25
Diastolic blood pressure,
mmHg
78 (72–86) 77 (69–80) 0.16
Clinical features, n(%)
Hypertension 32 (74.4) 41 (83.7) 0.27
Diabetes mellitus 17 (39.5) 17 (34.7) 0.63
Prior stroke/TIA 12 (27.9) 16 (32.7) 0.62
Coronary heart disease 14 (32.6) 10 (20.4) 0.19
Current smoking 17 (39.5) 19 (38.8) 0.94
Biochemical parameters
Triglycerides, mmol/L 1.4 (1.1–1.9) 1.1 (1.0–1.8) 0.11
Total cholesterol, mmol/L 3.5 (2.9–4.5) 3.5 (2.8–4.0) 0.36
HDL, mmol/L 1.0 0.2 1.0 0.2 0.75
LDL, mmol/L 1.8 (1.5–2.4) 1.9 (1.5–2.5) 0.69
Creatinine, μmol/L 64.0 (54.0–74.2) 65.5 (57.3–77.2) 0.58
Serum glucose, mmol/L 5.6 (4.9–7.0) 5.5 (4.8–6.4) 0.54
MC features
With MC, n(%) 36 (83.7) 16 (32.0) <0.001
Longitudinal length, mm 6.0 (3.5–9.4) 0 (0–2.1) <0.001
Maximum area, % 27.3 (17.9–34.0) 0 (0–15.0) <0.001
Mean area, % 19.6 (14.3–23.6) 0 (0–11.3) <0.001
MC index 125.2 (55.5–199.4) 0 (0–50.0) <0.001
HMC, n(%) 32 (74.4) 8 (16.0) <0.001
OCT findings
MLA, mm
2
8.0 5.7 10.5 5.9 0.04
Lipid-rich plaque, n(%) 30 (69.8) 7 (14.0) <0.001
Thrombus, n(%) 13 (30.2) 2 (4.0) 0.001
Cholesterol crystals, n(%) 14 (32.6) 11 (22.0) 0.25
Neovascularization, n(%) 8 (18.6) 7 (14.0) 0.55
Calcifications, n(%) 24 (55.8) 11 (22.0) 0.001
Multiple calcifications, n(%) 14 (32.6) 4 (8.0) 0.003
HDL, high-density lipoprotein; LDL, low-density lipoprotein; MC, macrophage infiltrations; MLA, min-
imum luminal area; PR, plaque rupture; TIA, transient ischemic attack.
Table 2. Comparison of clinical character-
istics, MC features, and OCT finding based
on the presence of plaque rupture.
8ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.
Macrophage Infiltration and Plaque Vulnerability X. Shi et al.
Figure 3. Subgroup analyses. HMC was shown to be a risk factor for PR after adjusting for lipid-rich plaques. No significant statistical association
was found between HMC and PR in lipid-rich plaques. However, HMC remained a risk factor for PR in non-lipid-rich plaques. In the other
subgroups, both overall and subgroup analyses showed that HMC represented as a risk factor for PR, including MLA, cholesterol crystals, and
calcifications, and none of interactions were found. CI, confidence interval; HMC, heavy macrophage infiltration; MLA, minimum lumen area; OR,
odds ratio.
Figure 4. Multivariable logistic regression analysis for plaque rupture. HMC and multiple calcifications were independently associated with non-
lipid-rich plaque rupture, whereas MLA (<9.36) and cholesterol crystals were not significant. CI, confidence interval; HMC, heavy macrophage
infiltration; MLA, minimum lumen area; OR, odds ratio.
ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 9
X. Shi et al. Macrophage Infiltration and Plaque Vulnerability
and found that NSD analysis exhibited the highest
accuracy.
27
However, other plaque components such as
the external elastic lamina and plaque healing could also
produce high NSD values.
29
Recently, Javier’s team pro-
posed an improved method based on NSD analysis by
adding the ratio processing of punctuated signal-rich
areas (with high NSD values) to signal-attenuated areas
or shadows (with low NSD values) in a subsequently dee-
per axial region, which led to significant improvements in
the accuracy, specificity, and sensitivity of NSD-based
recognition.
26
The study has made several improvements by incorpo-
rating the concept of NSDR and utilizing the characteris-
tics of cerebrovascular OCT images. Firstly, filter
preprocessing was employed to improve the image quality
for the inherent speckle on OCT images. Secondly, unlike
previous investigations conducted on coronary arteries,
the penetration depth of OCT in carotid plaques is lim-
ited, which poses challenges in accurately defining the
analyzable range of images during manual reading in clin-
ical settings. To mitigate excessive analysis and minimize
errors in manual readings, ER was used to quantitatively
evaluate MC in carotid plaques. Although the whole-
plaque macrophage density rarely exceeds a few
percent,
5,30
the superficial macrophage density is usually
high.
31
Thus, the MC% calculated based on ER may bet-
ter reflect the clinical focus and provide clinicians with
reference values. In the future, combining OCT with
imaging techniques that offer improved penetration, such
as intravascular ultrasound, may enable more comprehen-
sive evaluation of deeper plaques structure.
In the present study, HMC was identified as an inde-
pendent risk factor for PR in non-lipid-rich plaques. This
suggests that even in plaques lacking a lipid-rich core,
HMC may pose a threat to plaque homeostasis by
degrading the collagen-rich cap matrix and inhibiting
new collagen synthesis.
32,33
This discovery introduces a
novel finding that the presence of HMC could serve as a
crucial indicator of atherosclerosis progression in non-
lipid-rich plaques. While initial ruptures in this type of
plaques may be healed and clinically silent, persistent
HMC not only triggers further inflammation but also
contributes to the re-rupture of the fibrous cap.
34
Such de
novo ruptures, compared to first-time ruptures, are more
likely to generate artery-to-artery embolisms or result in
severe stenosis, ultimately leading to stroke.
35,36
Con-
versely, there was no significant independent association
found between HMC and PR in lipid-rich plaques. We
hypothesized that lipid-rich plaques might require reach-
ing a relatively higher risk threshold to disrupt the initial
temporary equilibrium within the plaque. Future studies
with larger sample sizes are necessary to validate this
hypothesis. Nevertheless, in lipid-rich plaques, sustained
HMC also contributes to increased inflammation, which
subsequently leads to rupture and thrombosis.
5
Although
the timing and mechanism of rupture may differ between
HMC plaques with and without large lipid cores, persis-
tent HMC remains a significant trigger that jeopardizes
plaque homeostasis. Therefore, early consideration of
aggressive management might be recommended once
HMC is detected, regardless of whether patients have
lipid-rich plaques or not. For instance, promoting anti-
inflammatory diets in the daily health management of
these patients could be beneficial, as studies have indi-
cated that such diets may be associated with reduced pla-
que vulnerability and vascular events.
37,38
Additionally,
more intensive cholesterol lowering therapy or the addi-
tion of anti-inflammatory agents may need to be
considered.
More importantly, our findings may inspire the design
of forthcoming clinical trials aimed at risk stratification
for identifying patients at increased risk of stroke while
undergoing medical therapy. Growing evidence suggests
that stroke could be attributed to the presence of vulnera-
ble plaques, even in the absence of moderate or severe
stenosis.
4,7,39–41
As medical therapy advances and risk fac-
tor control improves, the benefits of surgery or stenting
may diminish.
42,43
Hence, it becomes imperative to iden-
tify patients with asymptomatic carotid stenosis with sta-
ble and with unstable plaques and to select those patients
who would benefit from carotid intervention. A recent
study has indicated that persistent increasing active
inflammation may hinder plaques from exhibiting a
favorable response and rendering them vulnerable despite
statin therapy. This is evident by the development of new
layered plaques, which undergo a cycle of rupture and
healing, ultimately contributing to plaque progression.
44
Incorporating plaque inflammation into current selection
strategies may target patients who are most likely to
derive long-term benefits from carotid revascularization,
both in the early and late stages. However, the limited
methods available for assessing MC hinder further analy-
sis regarding the extent of MC that would benefit most
from carotid revascularization. Our exploratory study
introduces a novel approach for quantifying MC and ini-
tially determines the risk threshold of HMC based on a
limited number of clinical samples. This may provide a
valuable reference for designing future trials that investi-
gate risk models enriched with imaging data.
Over the past decade, OCT has emerged as a promising
technology for cerebrovascular assessment, offering
enhanced possibilities for detailed evaluation of both
extracranial and intracranial vessels.
13,45,46
Due to its high
resolution, OCT holds the potential to serve as an effec-
tive tool for precision medicine, including automatic
image recognition and machine learning.
47,48
However,
10 ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.
Macrophage Infiltration and Plaque Vulnerability X. Shi et al.
commercially available OCT imaging catheters were pri-
marily designed for coronary arteries, which imposes cer-
tain limitations when employed in cerebrovascular
assessment. Firstly, when assessing middle cerebral arteries
with tortuous pathways, the imaging catheter may be vul-
nerable to damage caused by excessive tortuosity.
49
Fortu-
nately, this limitation appears to be circumvented when
evaluating extracranial vessels and vertebrobasilar arteries
that exhibit relatively straight pathways. Furthermore, the
depth of penetration of OCT imaging catheters remains
limited in cerebrovascular assessment. This hinders their
capacity to assess deeper structures and vascular remodel-
ing. In the future, this challenge could be addressed by
incorporating intravascular ultrasound assessment, which
offers greater depth of penetration. Overall, it seems
imperative to develop OCT imaging catheters specifically
tailored for cerebrovascular assessment.
This study has some limitations. Firstly, the resolution
of OCT imaging limited the identification of specific sub-
types of inflammatory cells in the OCT image-based algo-
rithm. However, this limitation could potentially be
addressed in the future through the implementation of
high-frequency OCT or micro-OCT.
50
Secondly, the algo-
rithm has not yet been able to analyze lesions of in-stent
neo-atherosclerosis due to the high signal emitted by the
stent wire within plaques, which might cause errors.
Lastly, the sample size for the clinical analysis remains
limited. Future studies with larger sample sizes may pro-
vide more robust data to support the findings of this
study.
Conclusively, this study introduced an OCT-based
algorithm for automatic assessment of MC in vivo. HMC
(MC index >60.2) identified by the algorithm was an
independent risk factor for PR even in non-lipid-rich pla-
ques. This study provides a novel approach for in vivo
identification of MC and new screening criteria for HMC
plaques, which might be valuable for early detection and
monitoring of high-risk atherosclerotic plaques.
Acknowledgements
This project was supported by National Natural Science
Foundation of China (81530038, 81901218), and National
Key Research and Development Program of China
(2017YFC1307901).
Author Contributions
Conceptualization: Xinfeng Liu, Rui Liu, Xuan Shi, and
Tao Tao; Methodology: Yi Wang, Yunfei Han, Qin Yin,
and Rui Liu; Formal analysis and investigation: Xuan Shi,
Tao Tao, Xiaohui Xu, Fang Wang, and Rui Liu; Writing
—original draft preparation: Xuan Shi and Tao Tao;
Writing—review and editing: Rui Liu; Funding acquisi-
tion: Xinfeng Liu and Rui Liu; Supervision: Xinfeng Liu.
Conflict of interest
The authors declare that there is no conflict of interest.
Data Availability Statement
Data that support the findings of this study are available
for sharing with qualified investigators on reasonable
request.
References
1. Naylor AR. Why is the management of asymptomatic
carotid disease so controversial? Surgeon. 2015;13:34-43.
2. Hogberg D. Asymptomatic carotid stenosis: the silent
threat. Eur J Vasc Endovasc Surg. 2021;61:374.
3. Abbott AL, Paraskevas KI, Kakkos SK, et al. Systematic
review of guidelines for the management of asymptomatic
and symptomatic carotid stenosis. Stroke.
2015;46:3288-3301.
4. Kamtchum-Tatuene J, Noubiap JJ, Wilman AH, Saqqur
M, Shuaib A, Jickling GC. Prevalence of high-risk plaques
and risk of stroke in patients with asymptomatic carotid
stenosis: a meta-analysis. JAMA Neurol.
2020;77:1524-1535.
5. Bentzon JF, Otsuka F, Virmani R, Falk E. Mechanisms of
plaque formation and rupture. Circ Res.
2014;114:1852-1866.
6. Howard DP, van Lammeren GW, Rothwell PM, et al.
Symptomatic carotid atherosclerotic disease: correlations
between plaque composition and ipsilateral stroke risk.
Stroke. 2015;46:182-189.
7. McCabe JJ, Camps-Renom P, Giannotti N, et al. Carotid
plaque inflammation imaged by pet and prediction of
recurrent stroke at 5 years. Neurology. 2021;97:
e2282-e2291.
8. Brinjikji W, Huston J 3rd, Rabinstein AA, et al.
Contemporary carotid imaging: from degree of stenosis to
plaque vulnerability. J Neurosurg. 2016;124:27-42.
9. Robson PM, Kaufman A, Pruzan A, et al. Scan-rescan
measurement repeatability of (18)f-fdg pet/mr imaging of
vascular inflammation. J Nucl Cardiol. 2022;29:1660-1670.
10. Li L, Dmytriw AA, Krings T, Feng Y, Jiao L. Visualization
of the human intracranial vasa vasorum in vivo using
optical coherence tomography. JAMA Neurol.
2020;77:903-905.
11. Montone RA, Vetrugno V, Camilli M, et al. Macrophage
infiltrates in coronary plaque erosion and cardiovascular
outcome in patients with acute coronary syndrome.
Atherosclerosis. 2020;311:158-166.
ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 11
X. Shi et al. Macrophage Infiltration and Plaque Vulnerability
12. Shi X, Cai H, Wang F, et al. Cholesterol crystals are
associated with carotid plaque vulnerability: an optical
coherence tomography study. J Stroke Cerebrovasc Dis.
2020;29:104579.
13. Xu X, Li M, Liu R, et al. Optical coherence tomography
evaluation of vertebrobasilar artery stenosis: case series
and literature review. J Neurointerv Surg. 2020;12:809-813.
14. Xu X, Huang FH, Shi X, et al. Optical coherence
tomography evaluation of carotid artery stenosis and
stenting in patients with previous cervical radiotherapy.
Front Neuroscience. 2022;16:861511.
15. Xu R, Yang B, Li L, et al. Macrocalcification of
intracranial vertebral artery may be related to in-stent
restenosis: lessons learned from optical coherence
tomography. J Neurointerv Surg. 2021;14:475-479.
16. Tearney GJ, Regar E, Akasaka T, et al. Consensus
standards for acquisition, measurement, and reporting of
intravascular optical coherence tomography studies: a
report from the international working group for
intravascular optical coherence tomography
standardization and validation. J Am Coll Cardiol.
2012;59:1058-1072.
17. Prabhudesai V, Phelan C, Yang Y, Wang RK, Cowling
MG. The potential role of optical coherence tomography
in the evaluation of vulnerable carotid atheromatous
plaques: a pilot study. Cardiovasc Intervent Radiol.
2006;29:1039-1045.
18. Shimokado A, Matsuo Y, Kubo T, et al. In vivo optical
coherence tomography imaging and histopathology of
healed coronary plaques. Atherosclerosis. 2018;275:
35-42.
19. Liu R, Jiang Y, Xiong Y, et al. An optical coherence
tomography assessment of stent strut apposition based on
the presence of lipid-rich plaque in the carotid artery. J
Endovasc Ther. 2015;22:942-949.
20. Nakajima A, Araki M, Kurihara O, et al. Predictors for
rapid progression of coronary calcification: an optical
coherence tomography study. J Am Heart Assoc. 2021;10:
e019235.
21. John A, Bogovic PH, Wong A. Stephan Saalfeld. Robust
Registration of Calcium Images by Learned Contrast
Synthesis. IEEE; 2015.
22. Ughi GJ, Verjans J, Fard AM, et al. Dual modality
intravascular optical coherence tomography (oct) and
near-infrared fluorescence (nirf) imaging: a fully
automated algorithm for the distance-calibration of nirf
signal intensity for quantitative molecular imaging. Int J
Cardiovasc Imaging. 2015;31:259-268.
23. Miyagawa M, Costa MGF, Gutierrez MA, Costa JPGF,
Filho CFFC. Lumen segmentation in optical coherence
tomography images using convolutional neural network.
Annu Int Conf IEEE Eng Med Biol Soc.
2018;2018:600-603.
24. Tearney GJ. Quantification of macrophage content in
atherosclerotic plaques by optical coherence tomography.
Circulation. 2002;107:113-119.
25. MacNeill BD, Jang IK, Bouma BE, et al. Focal and multi-
focal plaque macrophage distributions in patients with
acute and stable presentations of coronary artery disease. J
Am Coll Cardiol. 2004;44:972-979.
26. Rico-Jimenez JJ, Campos-Delgado DU, Buja LM, Vela D,
Jo JA. Intravascular optical coherence tomography method
for automated detection of macrophage infiltration within
atherosclerotic coronary plaques. Atherosclerosis.
2019;290:94-102.
27. Di Vito L, Agozzino M, Marco V, et al. Identification and
quantification of macrophage presence in coronary
atherosclerotic plaques by optical coherence tomography.
Eur Heart J Cardiovasc Imaging. 2015;16:807-813.
28. Scalone G, Niccoli G, Refaat H, et al. Not all plaque
ruptures are born equal: an optical coherence tomography
study. Eur Heart J Cardiovasc Imaging. 2017;18:1271-1277.
29. Phipps JE, Vela D, Hoyt T, et al. Macrophages and
intravascular oct bright spots: a quantitative study. JACC
Cardiovasc Imaging. 2015;8:63-72.
30. Falk E. Pathogenesis of atherosclerosis. J Am Coll Cardiol.
2006;47:C7-C12.
31. Kolodgie FD, Burke AP, Farb A, et al. The thin-cap
fibroatheroma: a type of vulnerable plaque: the major
precursor lesion to acute coronary syndromes. Curr Opin
Cardiol. 2001;16:285-292.
32. Gough PJ. Macrophage expression of active mmp-9
induces acute plaque disruption in apoe-deficient mice. J
Clin Investig. 2005;116:59-69.
33. Virmani R, Burke AP, Farb A, Kolodgie FD. Pathology of
the vulnerable plaque. J Am Coll Cardiol. 2006;47:
C13-C18.
34. Russo M, Kim HO, Kurihara O, et al. Characteristics of
non-culprit plaques in acute coronary syndrome patients
with layered culprit plaque. Eur Heart J Cardiovasc
Imaging. 2020;21:1421-1430.
35. Burke AP, Kolodgie FD, Farb A, et al. Healed plaque
ruptures and sudden coronary death. Circulation.
2001;103:934-940.
36. Kramer MC, Rittersma SZ, de Winter RJ, et al.
Relationship of thrombus healing to underlying plaque
morphology in sudden coronary death. J Am Coll Cardiol.
2010;55:122-132.
37. Peng M, Wang L, Xia Y, et al. High dietary inflammatory
index is associated with increased plaque vulnerability of
carotid in patients with ischemic stroke. Stroke.
2020;51:2983-2989.
38. Estruch R, Ros E, Salas-Salvado J, et al. Primary
prevention of cardiovascular disease with a mediterranean
diet supplemented with extra-virgin olive oil or nuts. N
Engl J Med. 2018;378:e34.
12 ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.
Macrophage Infiltration and Plaque Vulnerability X. Shi et al.
39. Takaya N, Yuan C, Chu B, et al. Association between
carotid plaque characteristics and subsequent ischemic
cerebrovascular events: a prospective assessment with mri
—initial results. Stroke. 2006;37:818-823.
40. Zhao X, Hippe DS, Li R, et al. Prevalence and
characteristics of carotid artery high-risk atherosclerotic
plaques in chinese patients with cerebrovascular
symptoms: a chinese atherosclerosis risk evaluation ii
study. J Am Heart Assoc. 2017;6:6.
41. Yamada K, Kawasaki M, Yoshimura S, et al. High-
intensity signal in carotid plaque on routine 3d-tof-mra is
a risk factor of ischemic stroke. Cerebrovasc Dis.
2016;41:13-18.
42. Reiff T, Eckstein HH, Mansmann U, et al. Carotid
endarterectomy or stenting or best medical treatment
alone for moderate-to-severe asymptomatic carotid artery
stenosis: 5-year results of a multicentre, randomised
controlled trial. Lancet Neurol. 2022;21:877-888.
43. Keyhani S, Cheng EM, Hoggatt KJ, et al. Comparative
effectiveness of carotid endarterectomy vs initial medical
therapy in patients with asymptomatic carotid stenosis.
JAMA Neurol. 2020;77:1110-1121.
44. Nakajima A, Minami Y, Araki M, et al. Optical coherence
tomography predictors for a favorable vascular response to
statin therapy. J Am Heart Assoc. 2020;10:e018205.
45. Li T, Xu R, Ma Y, Wang T, Yang B, Jiao L. Calcification is
a risk factor for intracranial in-stent restenosis: an optical
coherence tomography study. J Neurointerv Surg.
2023;jnis-2023-020624. Epub ahead of print. doi: 10.1136/
jnis-2023-020624
46. Funatsu N, Enomoto Y, Egashira Y, et al. Tissue
protrusion with attenuation is associated with ischemic
brain lesions after carotid artery stenting. Stroke.
2020;51:327-330.
47. Park S, Araki M, Nakajima A, et al. Enhanced diagnosis of
plaque erosion by deep learning in patients with acute
coronary syndromes. JACC Cardiovasc Interv.
2022;15:2020-2031.
48. Shibutani H, Fujii K, Shirakawa M, et al. Diagnostic
accuracy of optical frequency domain imaging for
identifying necrotic cores with intraplaque hemorrhage in
advanced human carotid plaques. Am J Cardiol.
2021;156:123-128.
49. Lopes DK, Johnson AK. Evaluation of cerebral artery
perforators and the pipeline embolization device using
optical coherence tomography. J Neurointerv Surga.
2012;4:291-294.
50. Ughi GJ, Marosfoi MG, King RM, et al. A neurovascular
high-frequency optical coherence tomography system
enables in situ cerebrovascular volumetric microscopy. Nat
Commun. 2020;11:3851.
Supporting Information
Additional supporting information may be found online
in the Supporting Information section at the end of the
article.
Appendix S1.
ª2023 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association. 13
X. Shi et al. Macrophage Infiltration and Plaque Vulnerability
