Добавил:
Sekretar
kiopkiopkiop18@yandex.ru
t.me/Prokururor I Вовсе не секретарь, но почту проверяю
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз:
Предмет:
Файл:Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_3592_Библиотеки_им_академика_М_И_Перельмана
.pdf
Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
[31] Saeed M, Hetts S W, English J and Wilson M 2012 MR fluoroscopy in vascular and cardiac
interventions (review) Int. J. Cardiovasc. Imaging
[32] Sahu S 2012 Comparative analysis of image enhancement techniques for ultrasound liver
image Int. J. Electr. Comput. Eng.
[33] Scath (2010-2012). http://www.scath.net/
[34] Schoonenberg G, Schrijver M, Duan Q, Kemkers R and Laine A 2005 Adaptive spatial–
temporal filtering applied to x-ray fluoroscopy angiography Proc. SPIE
[35] Shi C, Giannarou S, Lee S L and Yang G Z 2014 Simultaneous catheter and environment
modeling for trans-catheter aortic valve implantation 2014 IEEE/RSJ Int. Conf. on
Intelligent Robots and Systems (IROS 2014) (Piscataway, NJ: IEEE) pp
[36] Shibin W, Qingsong Z and Yaoqin X 2013 Evaluation of various speckle reduction filters on
medical ultrasound images Conf. Proc. IEEE Eng. Med. Biol. Soc.
[37] Shoa T, Madden J D W, Munce N R and Yang V 2010 Analytical modeling of a conducting
polymer-driven catheter Polym. Int.
[38] Stephens D N et al 2006 5G-6 forward looking intracardiac imaging catheters for electro-
physiology Proc. IEEE Ultrasonics Symp. (Piscataway, NJ: IEEE)
1109/ULTSYM.2006.191
[39] Tangen G A, Manstad-Hulaas F, Brekken R and Hernes T A N 2013 Navigation in endovascular
aortic repair aortic Aneurysm—Recent Advances (Rijeka: Intech)
books/aortic-aneurysm-recent-advances/navigation-in-endovascular-aortic-repair
[40] Tu S, Holm N R, Koning G, Huang Z and Reiber J H C 2011 Fusion of 3D QCA and IVUS/
OCT Int. J. Cardiovasc. Imaging
[41] Unal G, Gurmeric S and Carlier S G 2010 Stent implant follow-up in intravascular optical
coherence tomography images Int. J. Cardiovasc. Imaging
[42] Vanithamani R and Umamaheswari G 2014 Speckle reduction in ultrasound images using
Neighshrink and bilateral filtering J. Comput. Sci.
[43] Wahle A, Prause G P M, DeJong S C and Sonka M 1999 Geometrically correct 3-D
reconstruction of intravascular ultrasound images by fusion with biplane angiography-
methods and validation IEEE Trans. Med. Imaging
[44] Wahle A, Prause G P M, Von Birgelen C, Erbel R and Sonka M 1999 Fusion of angiography
and intravascular ultrasound in vivo: establishing the absolute 3-D frame orientation IEEE
Trans. Biomed. Eng.
[45] Wang P, Ecabert O, Chen T, Wels M, Rieber J, Ostermeier M and Comaniciu D 2013
Image-based co-registration of angiography and intravascular ultrasound images IEEE
Trans. Med. Imaging
[46] Wang Y, Chen T, Wang P, Rohkohl C and Comaniciu D 2012 Automatic localization of
balloon markers and guidewire in rotational fluoroscopy with application to 3D stent
reconstruction Computer Vision—ECCV 2012 (Lecture Notes in Computer Science vol 7577)
(Berlin: Springer) sect 31, pp
[47] Wood B J et al 2005 Navigation with electromagnetic tracking for interventional radiology
procedures: a feasibility study J. Vasc. Interv. Radiol.
[48] Yushkevich P A, Piven J, Hazlett H C, Smith R G, Ho S, Gee J C and Gerig G 2006 User-
guided 3D active contour segmentation of anatomical structures: significantly improved
efficiency and reliability Neuroimage
46 1176–80
32 2238–49
2 792–7
59 343–51
27 197–207
428–41
31 1116–28
28 117–37
5744
2024–9
2013 1148–51
doi:https://doi.org/10.
https://www.intechopen.com/
26 809–16
10 623–31
18 686–99
16 493–505
11-23
Section IV
https://t.me/medicina_free
Risk stratification in carotid and coronary artery
https://t.me/medicina_free
IOP Publishing
https://t.me/medicina_free
Vascular and Intravascular Imaging Trends, Analysis, and
Challenges, Volume 1
Stent applications
Petia Radeva and Jasjit S Suri
Chapter 12
A cloud-based smart IMT measurement tool for
multi-center clinical trial and stroke risk
stratification in carotid ultrasound
Luca Saba, Sumit K Banchhor, Harman S Suri, Narendra D Londhe, Tadashi Araki,
Nobutaka Ikeda, Klaudija Viskovic, Shoaib Shafique, John R Laird, Ajay Gupta,
Andrew Nicolaides and Jasjit S Suri
This study presents AtheroCloud™—a novel cloud-based smart carotid intima–
media thickness (cIMT) measurement tool using B-mode ultrasound for stroke/
cardiovascular risk assessment and its stratification. This is an anytime–anywhere
clinical tool for routine screening and multi-center clinical trials. In this pilot study,
the physician can upload ultrasound scans in one of several formats (DICOM,
JPEG, BMP, PNG, GIF or TIFF) directly into the proprietary cloud of
AtheroPoint from the local server of the physician’soffice. They can then run the
intelligent and automated AtheroCloud™ cIMT measurements in point-of-care
settings in less than five seconds per image while saving the vascular reports in the
cloud. We statistically benchmark AtheroCloud™ cIMT readings against sonographer (a registered vascular technologist) readings and manual measurements derived
from the tracings of the radiologist.
Scans of one hundred patients (75 M/5 F, mean age: 68 ± 11 years; institutional
review board (IRB) approved, Toho University, Japan), of the left/right (L/R)
common carotid artery (CCA; 200 ultrasound scans), were collected using a 7.5
MHz transducer (Toshiba, Tokyo, Japan). The measured cIMTs for the L/R carotid
were as follows (in millimeters): (i) AtheroCloud™ (0.87 ± 0.20, 0.77 ± 0.20);
(ii) sonographer (0.97 ± 0.26, 0.89 ± 0.29) and (iii) manual (0.90 ± 0.20, 0.79 ± 0.20),
respectively. The coefficient of correlation (CC) between the sonographer and
manual for L/R cIMT was 0.74 (P < 0.0001) and 0.65 (P < 0.0001), while between
AtheroCloud™ and manual the CC was 0.96 (P < 0.0001) and 0.97 (P < 0.0001),
doi:10.1088/2053-2563/ab01fach12 12-1 ª IOP Publishing Ltd 2019
Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
respectively. We observed that 91.15% of the population in AtheroCloud™ had a
mean cIMT error less than 0.11 mm compared to 68.31% for the sonographer. The
area under the curve for receiving operating characteristics was 0.99 for
AtheroCloud™ against 0.81 for the sonographer. Our Framingham risk score
stratified the population into three bins as follows: 39% in low-risk, 70.66% in
medium-risk and 10.66% in the high-risk bins. Statistical tests were performed to
demonstrate the consistency, reliability and accuracy of the results. The proposed
AtheroCloud™ system is a completely reliable, automated, fast (3–5 s depending
upon the image size with an internet speed of 180 Mbps), accurate and intelligent
web-based clinical tool for multi-center clinical trials and routine telemedicine
clinical care.
12.1 Introduction
Cardiovascular diseases (CVDs) have been predicted as the main cause of morbidity
globally. On an average, 7.4 million deaths were due to CVDs and 6.7 million were
due to stroke [1]. It was found that over three-quarters of CVD deaths take place in
low and middle-income countries. In particular, the South-East Asia region is
showing a rapid increase of CVDs in the young and middle-aged population.
Between 2000 and 2030, it is estimated that about 35% of all CVD deaths in India
will occur among the 35–64 year age group [2]. Coronary artery disease and carotid
artery disease are two primary examples of diseases caused by the build-up of
atherosclerotic plaque that falls under the broader category of CVDs [3].
Atherosclerosis is a progressive process that damages the endothelium due to the
deposition of plaque in the arteries [4, 5]. Atherosclerosis narrows the arteries,
restricting the flow of oxygenated blood in the body [6]. As atherosclerosis
progresses, the blockage can rupture, causing the clot to dislodge and travel
downstream (figure 12.1)[7]. This results in myocardial infarction or stroke.
Figure 12.1. Illustration of plaque formation in the carotid artery. (Courtesy of AtheroPoint™, Roseville, CA,
USA.)
12-2
Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
People with cardiovascular disease or high cardiovascular risk may significantly
benefit from early detection, monitoring and management [8, 9].
cIMT is one of the most popular methods for monitoring CVD and predicting the
occurrence of major adverse cardiovascular events [10–22]. The prediction of CVD
has been tied to cIMT in previous studies, with the aim of foreseeing cardiovascular
events (CVE). Several studies have shown a relationship between threshold values of
cIMT and CVD: (cIMT > 0.7 mm) [23], (cIMT > 0.85 mm) [ 24], (cIMT > 0.9 mm)
[25, 26], (cIMT > 1.0 mm) [27] and (cIMT > 1.26 mm) [28]. Recent studies have also
revealed a strong relationship between cIMT values and the severity of coronary
artery disease (CAD) [8, 25, 26, 29]. The above studies have clearly shown that
cIMT is a risk biomarker for CVEs.
In spite of the strong relationship between cIMT and CAD, clinicians have not
routinely benefitted from automated processing of carotid ultrasound scans. Several
studies [17–20] emphasized the need for an automated system for cIMT computation. This is mainly because current manual [19] or semi-automated systems [30]
used by sonographers are subjective and associated with operator or observer bias.
Current systems are not fully automated [30] and lack advanced image-based
features for risk assessment [31, 32]. Often, these systems lack reliability, accuracy
and reproducibility, and provide no comparative reference marker, which is needed
for monitoring. Furthermore, there is no standardization towards clinical trials [33].
The lack of reproducibility is due to the methodology used to take the readings, such
as (i) caliper-based and (ii) readings taken manually at a limited number of locations
(positions) along the CCA [34]. We assume that a robust and validated automated
system is more accurate and reliable compared to sonographer readings taken by a
registered vascular technologist (RVT) in a vascular ultrasound laboratory. This
assumption can be proven if the error between the automated AtheroCloud™
software-based cIMT readings and the gold standard (manual tracings taken by the
radiologist) is lower compared to the error between the sonographer’s reading and
the gold standard (manual).
Through automated systems, the operator variability and subjectivity can be
controlled, but there are still challenges in stroke/cardiovascular risk monitoring,
such as the ability to operate in remote areas of the world. The current methods for
cIMT measurement use a cart-based ultrasound machine or portable machines
which are bulky to carry, in contrast to pocket-sized machines [35], putting a strain
on mobile-based infrastructure. Patients and doctors are physically con
fined to a
machine and a clinical protocol cannot be executed if the patient, for example, is in a
rural area without access to a physician [36]. The concept of home-healthcare [37]is
not prevalent and the traditional approach lacks an ‘anytime–anywhere’ solution.
There are two major challenges in current cIMT system designs in cloud-based
settings:
(i) The design of a two-pronged system, with a single (routine) mode and a batch
(pharmaceutical) mode, that can recognize the far wall of the carotid artery
[38] and detect the lumen–intima and media–adventitia interfaces [39, 40].
(ii) The design of a multicenter clinical tool which is completely automated
(which can extract cIMT measurements over thousands of studies without
12-3
Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
interruption), which can handle the variability in image characteristics such
as resolution, contrast, size, quality and formats, and be able to process
images from different countries. A synopsis of previous techniques is
presented in the discussion section.
The proposed patented system is ‘smart’ in the sense that intelligent cloudcomputing is adapted for recognition of carotid anatomy in carotid scans (having
multiple vascular beds) and computing the lumen–intima (LI)/media–adventitia
(MA) (details discussed in the next section) interfaces along with the cIMT
measurements. The system is intelligent in the sense that it is able to automatically
recognize the far (posterior) wall of the carotid artery even in the presence of the
near and far walls of the jugular vein [41].
In a cloud-based approach, the physician can upload ultrasound scans in one of
several formats (DICOM, JPEG, BMP, PNG, GIF or TIFF) directly into the
proprietary cloud of AtheroPoint from the local server of the physician’soffice, and
then run the intelligent and automated AtheroCloud™ cIMT measurements in
point-of-care settings in less than five seconds per image, while saving the vascular
reports in the cloud. We then compare and validate AtheroCloud™ readings against
a sonographer’s measurements and manual (gold standard) readings by computing
the precision-of-merit (PoM) and CC between these methods. The performance of
the AtheroCloud™ system is then analyzed by computing the area under the curve
(AUC) of the receiver operating characteristics (ROC) [42]. We also compare the
Routine mode (a single image at a time) against the Pharmaceutical trial mode
(a batch of images at a time without interruption) using the AtheroCloud™ software
system, showing the reliability and reproducibility. Further, we benchmark
AtheroCloud against the commercially available desktop-based systems such as
AtheroEdge™ (AtheroPoint™, Roseville, CA, USA), that (i) previously has been
benchmarked against original equipment manufacturer (OEM) vendors such as
Siemens [43], (ii) is used for epidemiological studies [44] and (iii) is a 510(K) cleared
medical device [45], establishing the standard for cIMT measurement [33], demonstrating the error difference which follows the criteria of acceptance under
regulatory conditions.
12.2 Patient demographics and data acquisition
12.2.1 Patient demographics
Two hundred and four (204) patients underwent both (i) percutaneous coronary
interventions using iMap (Boston Scientific
tion and (ii) B-mode carotid ultrasound scans. Both left and right CCA ultrasound
scans (a total of 407 images, one patient had one image missing) were obtained from
Toho University, Japan and retrospectively analyzed (ethics approved by the IRB).
For this pilot study, due to cost and manual tracing constraints, we randomly
selected 100 patients (200 CCA ultrasound scans) for this design study. No special
criteria were adopted in choosing the 100 patients. There were 75 male and 25 female
patients with a mean age 68 ± 11 ranging from 29 to 88 years. Of these, 53 patients
®
, Marlborough, MA) IVUS examina-
12-4
Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
had a proximal lesion location, 27 a middle location and 20 a distal location. These
100 patients have a mean HbA1c of 6.40 ± 1.2 mg dl
lipoprotein (LDL) cholesterol of 104.60 ± 30.4 mg dl
oprotein (HDL) cholesterol of 51.5 ± 15.9 mg dl
35.4 mg dl
−1
. Thirty-nine of the pool of one hundred were smokers. These data were
−1
and total cholesterol of 179.40 ±
−1
, mean low-density
−1
, mean high-density lip-
acquired from July 2009 to December 2010.
12.2.2 Ultrasound image data acquisition
All the patients were scanned using an ultrasound scanner (model: Aplio XV, Aplio
XG, Xario) equipped with a 7.5 MHz linear array transducer from Toshiba, Inc.,
Tokyo, Japan. The same sonographer (with 15 years of experience) scanned all the
patients. The American Society of Echocardiography Carotid Intima–Media
Thickness Task Force protocol was used and high-resolution images of the CCA
were acquired. First, the subjects were examined in the supine position and the head
was tilted backward. The probe first located the carotid arteries using transverse
scans (perpendicular to the blood flow) and then the probe was rotated by 90° to
acquire the longitudinal carotid ultrasound scans. Two views were collected:
anterior and posterior walls. The sonographer also acquired internal carotid artery
(ICA) and carotid bulbs scans in addition to CCA images in order to calculate the
plaque score (PS). This study does not use PS measurements and is discussed
elsewhere. This study underwent a full ethics review by the IRB of Toho University
Hospital and written informed consent was provided by all the patients. The average
resolution factor for carotid ultrasound scans was 0.0529 mm/pixel.
12.2.3 Sonographer’s cIMT readings
These are the measurements taken by the sonographer who is present in the vascular
ultrasound laboratory during the digital acquisition of the ultrasound scan of the
patient’s carotid artery. A sonographer is a registered vascular technologist (RVT)
and has a background in vascular ultrasound and is qualified enough to measure the
cIMT in carotid ultrasounds. The sonographer uses the software integrated with the
ultrasound scanning device for measuring the cIMT. The sonographer places two
points manually: one along the lumen–intima (LI) interface and the second along the
media–adventitia (MA) interface. The LI point is the transition point when going
from the lumen region to the intima region. The MA point is the transition point
when going from the media region to the adventitia region. The software then
processes these points to compute the distance between them, which is called the
cIMT. The sonographer places these points visually by looking at the plaque
distribution in the carotid B-mode ultrasound scans. It is important to note that the
sonographer places these points manually and uses his/her experience and judgment
when placing these points.
Since experience and judgment can be different between different sonographers,
the cIMT measurement can also differ between sonographers. That is why
sonographer readings have larger variability. For this reason, there is a clear
motivation to design an automated cIMT measurement system. It is also important
12-5
Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
to note that sometimes it is the physician who takes these measurements if the
sonographer is not present. Several countries follow their own methodologies for
cIMT measurements due to direct costs and overheads [46]. For example, in some
European countries (excluding the United Kingdom and Ireland), it is the physician
who takes these cIMT measurements. This places fewer financial burdens on the
hospital or clinic. In Asia (especially in India), it is the sonologist who takes these
measurements. A sonologist, who has a medical degree, is qualified to take these
measurements. Our ultimate goal is to collect these measurements, taken either by a
sonographer or a sonologist or the physician. Here we call these measurements
sonographer readings.
12.2.4 Manual cIMT readings
For a performance evaluation of the AtheroCloud™ software system, the gold
standard was created by manually tracing the LI/MA interfaces for the distal wall of
the carotid artery in the ultrasound scan. We used a commercial software package
ImgTracer™ (courtesy of AtheroPoint™, Roseville, CA, USA) for manual tracing
of LI/MA interfaces. The LI/MA interface is the border between the lumen–intima
(LI) and media–adventitia (MA). The carotid ultrasound image shows the carotid
artery with a lumen at the center and walls on both sides of the lumen region. These
walls are called the near (proximal) and far (distal) walls of the carotid artery. The
carotid artery wall is surrounded by the lumen region on one side and the adventitia
region on the other side, while the wall consists of the intima and media regions. The
interface between the lumen and intima walls is called the LI interface or LI border.
The interface between the media region and adventitia region is the MA interface or
MA border. Between the LI and MA interface are found the plaque or atherosclerotic diseased components. The mean distance between the LI border and the
MA border is the cIMT or plaque burden. The greater the plaque burden, the larger
is the cIMT measurement, and the AtheroCloud™ software system allows us to
measure this plaque burden. A research scholar (SKB, the second author of this
chapter), currently a doctoral candidate in the field of atherosclerosis imaging,
traced the IM and MA interfaces. SKB was trained under the guidance of JSS
(principal investigator on this project and corresponding author for this chapter),
who is an expert in vascular ultrasound imaging, with experience of 25 years in
imaging sciences. The gold standard tracings were finally checked and endorsed by
LS (the first author of this chapter), a neuroradiologist with 15 years of experience in
radiology and the author of over 100 international journal articles [47] in carotid
ultrasound and radiology.
A sample view of ImgTracer™ is shown in figure 12.2. The top yellow line
indicates the LI and the bottom yellow line the MA interface. The tracing protocol
was adopted for all 100 patients, consisting of a total of 200 carotid scans.
ImgTracer™ software has been installed at several geographical locations around
the world such as India (NIT-Raipur), Malaysia (UTM Razak School of
Engineering and Advanced Technology), Italy (University of Cagliari, Italy),
Croatia (Department of Radiology and Ultrasound, Zagreb) and the USA
12-6
Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
Figure 12.2. Manual tracings (yellow) of the carotid intima–media thickness region showing LI and MA
borders using ImgTracer™. (Courtesy of AtheroPoint™, Roseville, CA, USA.)
(AtheroPoint, Roseville, CA), and has been successfully used for several anatomic
applications.
12.3 Methodology and cloud-based workflow
AtheroCloud™ 1.0 (courtesy of AtheroPoint™, Roseville, CA, USA) is a cloudbased stroke monitoring software system which can be used in (a) completely
automated or (b) user-interactive semi-automated modes for computing (i) intima–
media thickness (IMT) and its variability (IMTV), (ii) lumen diameter (LD) and its
variability (LDVar), and (iii) stenosis severity index (SSI) in carotid ultrasound
scans.
12.3.1 Workflow architecture of the AtheroCloud™ 1.0 system
Even though there are automated desktop-based intelligent cIMT systems created by
HSS (the principal investigator on this project) and his team [9, 11, 13, 14, 30, 48, 49],
this study is the first of its kind in which cIMT is measured in carotid ultrasound scans
through intelligent cloud-computing, with automated analysis occurring in the cloudbased settings. The workflow is shown in figure 12.3. It consists of a three-layer
architecture:
(i) The GUI layer, where, the doctor can interact with the AtheroCloud™
software using a laptop or PC. This is mainly used for accessing ultrasound
12-7
Соседние файлы в папке Библиотека им академика М.И. Перельмана