Automat ed Diagnosis of Cardiovascular Disease s from Cardiac Magnetic Resonance Imaging Using Deep Learning Models A Review Mahbo obeh Jafari1 Afshin Shoeibi12 Marjane Khodatars3 Navid Ghassemi1 Parisa Moridian1

2025-05-02 0 0 1.25MB 56 页 10玖币

侵权投诉

Automated Diagnosis of Cardiovascular Diseases from Cardiac Magnetic Resonance Imaging Using

Deep Learning Models: A Review

Mahboobeh Jafari1, Afshin Shoeibi1,2,*, Marjane Khodatars3, Navid Ghassemi1, Parisa Moridian1,

Niloufar Delfan4, Roohallah Alizadehsani5, Abbas Khosravi5, Sai Ho Ling6, Yu-Dong Zhang7,

Shui-Hua Wang7, Juan M. Gorriz2,8, Hamid Alinejad Rokny9,10,11, U. Rajendra Acharya12,13,14

1 Internship in BioMedical Machine Learning Lab, The Graduate School of Biomedical Engineering, UNSW Sydney,

Sydney, NSW, 2052, Australia.

2 Data Science and Computational Intelligence Institute, University of Granada, Spain.

3 Department of Medical Engineering, Mashhad Branch, Islamic Azad University, Mashhad, Iran.

4 Faculty of Computer Engineering, Dept. of Artificial Intelligence Engineering, K. N. Toosi University of

Technology, Tehran, Iran.

5 Intelligent for Systems Research and Innovation (IISRI), Deakin University, Victoria 3217, Australia.

6 Faculty of Engineering and IT, University of Technology Sydney (UTS), Australia.

7 School of Computing and Mathematical Sciences, University of Leicester, Leicester, UK.

8 Department of Psychiatry, University of Cambridge, UK.

9 BioMedical Machine Learning Lab, The Graduate School of Biomedical Engineering, UNSW Sydney, Sydney,

NSW, 2052, Australia.

10 UNSW Data Science Hub, The University of New South Wales, Sydney, NSW, 2052, Australia.

11 Health Data Analytics Program, AI-enabled Processes (AIP) Research Centre, Macquarie University, Sydney,

2109, Australia.

12 Ngee Ann Polytechnic, Singapore 599489, Singapore.

13 Dept. of Biomedical Informatics and Medical Engineering, Asia University, Taichung, Taiwan.

14 Dept. of Biomedical Engineering, School of Science and Technology, Singapore University of Social Sciences,

Singapore.

* Corresponding author: Afshin Shoeibi (Afshin.shoeibi@gmail.com)

Abstract

In recent years, cardiovascular diseases (CVDs) have become one of the leading causes of mortality

globally. CVDs appear with minor symptoms and progressively get worse. The majority of people

experience symptoms such as exhaustion, shortness of breath, ankle swelling, fluid retention, and other

symptoms when starting CVD. Coronary artery disease (CAD), arrhythmia, cardiomyopathy, congenital

heart defect (CHD), mitral regurgitation, and angina are the most common CVDs. Clinical methods such

as blood tests, electrocardiography (ECG) signals, and medical imaging are the most effective methods

used for the detection of CVDs. Among the diagnostic methods, cardiac magnetic resonance imaging

(CMR) is increasingly used to diagnose, monitor the disease, plan treatment and predict CVDs. Coupled

with all the advantages of CMR data, CVDs diagnosis is challenging for physicians due to many slices of

data, low contrast, etc. To address these issues, deep learning (DL) techniques have been employed to the

diagnosis of CVDs using CMR data, and much research is currently being conducted in this field. This

review provides an overview of the studies performed in CVDs detection using CMR images and DL

techniques. The introduction section examined CVDs types, diagnostic methods, and the most important

medical imaging techniques. In the following, investigations to detect CVDs using CMR images and the

most significant DL methods are presented. Another section discussed the challenges in diagnosing CVDs

from CMR data. Next, the discussion section discusses the results of this review, and future work in CVDs

diagnosis from CMR images and DL techniques are outlined. The most important findings of this study are

presented in the conclusion section.

KeyWords: Cardiovascular Disease, Diagnosis, CMR, Deep Learning, Classification, Segmentation

1. Introduction

CVDs are one of the most common causes of death and endanger the health of many people around the

world annually [1-2]. According to the World Health Organization (WHO), CVDs are the leading cause of

human death worldwide [3-4]. According to this statistics, 17.9 million people died from CVDs in 2016,

accounting for 31% of all global deaths [5-7]. In addition, coronary heart disease and stroke are responsible

for four out of five deaths from CVDs, and one-third of these deaths occur in people under 70 years [8-10].

Some of the most important CVDs include coronary arteries disease (CAD) [11-12], rheumatoid arthritis

[13-14], myocarditis [15-17], cardiovascular diabetes [18-19], etc. Figure (1) shows the patients with

cardiovascular diabetes in the world.

Fig. 1. Patients with cardiovascular diabetes in the world.

The human heart is responsible for pumping blood and circulating it throughout the body [18], so any

abnormality in it results in CVDs [19]. CAD is considered the most common type of CVD [20-22]. CAD

is the plaque accumulation in the arteries that supply oxygen-rich blood to the heart [20-22]. Plaque causes

narrowing or blockage, which restricts blood flow and thus reduces blood oxygen to parts of the heart [20-

22]. Some of the most significant symptoms of CAD involve chest pain or discomfort and shortness of

breath [20-22]. Cardiac arrhythmia is another of the most prevalent CVDs caused by atrial fibrillation and

ventricular arrhythmias [23-24]. A cardiac arrhythmia occurs due to a non-uniform heartbeat. Weakness

and pain in the chest area are the most important symptoms of arrhythmia [23-24]. Congenital heart disease

(CHD) is another human CVDs. There is a defect in the structure of the heart or large arteries are present

by birth [25-26]. Signs and symptoms of CHD include rapid breathing, a blue tingein the skin (cyanosis),

poor weight gain, and tiredness [25-26]. Figure (2) shows the common CVDs together with their details. In

recent years, major advances in cardiac research have been made to improve the diagnosis and treatment of

CVDs as well as decline their case fatality.

Fig. 2. Common types of CVDs with details.

Cardiac ultrasound or echocardiography (Echo) works by utilizing sound waves which is a non-invasive

modality to image heart tissue [27-28]. In this method, ultrasound waves are taken advantage of to produce

echocardiography images of the heart [27-28]. Echo helps physicians detect various types of CVDs by

assessing the heart's structure, analyzing how the blood flows in them, and evaluating the heart's pumping

cavities [27-28]. Advantages of echocardiography include readily accessible, portability, high temporal

resolution, and no ionizing radiation [29].

CT is a non-invasive imaging technique that can be applied to detect a variety of CVDs, brain diseases, etc.

[30-31]. In particular, cardiac CT provides the anatomical evaluation of the heart, especially CAD [32].

This imaging technique involves two techniques: non-contrast CT and contrast-enhanced coronary CT

angiography (CTA) [30-31]. Non-contrast CT makes use of the density of tissues to generate the image so

that various densities can be simply distinguished using different attenuation values [30-31] [33-34]. In

addition, the amount of calcium in the coronary arteries can be calculated using non-contrast CT [30-31]

[33-34]. In comparison, contrast-enhanced coronary CTA provides the ability to generate extraordinary

images of the heart, arteries, and coronary arteries [30-31] [33-34]. Radiation exposure is one of the major

weaknesses of cardiac CT imaging. Frequent exposure to radiation is associated with deleterious health

effects, including an increased cancer risk [30-31] [33-34].

CMR imaging offers an excellent quantitative assessment of cardiac chamber volume/function [374] and

the extent of myocardial infarction/fibrosis [375]. It is a guideline-recommended modality for the diagnosis

of diverse CVDs, including ischemic heart disease [37, 38], heritable or acquired cardiomyopathy [39],

myocarditis [40], congenital heart disease [41], etc. For measurement of ventricular volume, function, and

mass, accurate segmentation of the endocardial (and, in the case of myocardial mass, epicardial) contours

on standard cine CMR images is a necessary prerequisite. Typically, the contours are drawn ―either

manually or software-assisted―on a stack of contiguous parallel slices of two-dimensional (2D) short-axis

time-series cine CMR images of the ventricles at desired phases of the cardiac cycle, e.g., end-diastole and

-systole, to derive the corresponding time-aligned three-dimensional (3D) ventricular volumes using

Simpson’s method of disc without the need for geometric assumption [376]. Indeed, cine CMR analysis is

the gold standard for right ventricular (RV) volume/function measurement as the RV can be optimally

visualized on CMR without being limited by issues of acoustic window access, as with echocardiography

[377]. Late gadolinium enhancement (LGE) [378] is an established CMR imaging technique in which

images acquired ten to twenty minutes after gadolinium-based contrast administration are used to define in

granular detail regions of myocardial infarct, fibrosis, infiltrate, etc. Indeed, segmentation can also be

performed to outline and quantitate areas of abnormal tissue, e.g., myocardial infarct [379], microvascular

obstruction [380], and non-infarct fibrosis [381], which may have prognostic significance.

In addition to quantitative measurements, CMR must be qualitatively interpreted by medical experts, which

is time-consuming and subject to human bias. The presence of noise and imaging artifacts can further

confound the interpretation, potentially resulting in misdiagnosis. However, CMRI data is the gold standard

and most popular procedure for diagnosing cardiac diseases among physicians. To address CMRI

challenges, researchers have proposed artificial intelligence (AI) techniques for the automatic diagnosis of

CVDs using CMRI data [1-10]. In the presented papers, the main objective of the researchers is to achieve

a tool for rapid detection of CVDs using CMRI images along with AI techniques. For this purpose, the

researchers have conducted extensive research on ML-based approaches for diagnosing CVDs from CMRI

data, including introducing various segmentation and classification approaches [42-44]. However, ML

methods presented satisfactory results in early research on the diagnosis of CVDs. Nevertheless, due to

high computational complexity, and inefficient performance with huge databases these methods were not

able yield good performances. To tackle the challenges of ML methods, AI researchers introduced DL

methods [45-47]. DL networks were able to overcome the limitations of ML methods [45-47]. The DL

models were employed in various medical applications, including the diagnosis of CVDs [4], and reported

satisfactory results. Researchers hope that in the near future, an accurate software platform for diagnosing

CVDs using MRI data and DL techniques will be realized.

In this study, papers in diagnosis of CVDs using CMRI images and DL techniques were examined. The

section 3 describes search strategy papers regarding preferred reporting items for systematic reviews and

meta-analyses (PRISMA) guidelines [48]. In section 4, the conducted review papers in diagnosis of CVDs

are studied. The computer aided diagnosis system (CADS) and their steps for diagnosis CVDs from CMRI

images are provided in Section 5. This section discusses in datasets, preprocessing, and popular DL models

for diagnosis of CVDs. Also, in this section, segmentation, classification, and fusion research based on DL

methods are summarized in different Tables. Section 6 is allocated to the most important challenges in

diagnosis of CVDs using CMRI data. The discussion of this paper, along with its details, is provided in

section 7. Future work is also presented in section that suggest potential directions for future works. Finally,

the conclusion and the findings of this study are discussed in the section 7.

2. Search strategy

This section searches papers based on PRISMA guidelines [48]. We have searched the papers published

between 2016 and 2022 in the field of heart diseases using the general keywords "CVDs", "deep learning",

"Segmentation", "classification", and "CMRI”. Keyword searches are performed in repositories such as

Science Direct, Frontiers, MDPI, IEEE Xplore, Nature, Springer, ArXiv, and Wiley citation databases.

The selection method of important articles for diagnosing CVDs with AI techniquesis presented in this

section. The selection process of papers related to this field has been done in three levels. In Figure (3), the

review process of papers based on PRISMA guidelines is provided. First, 324 articles were collected and

then 38 articles were filtered out as they are not related to this area of research. In the following, 33 papers

did not use the CMRI datasets and filtered. Finally, 21 articles were filtered out as they did not use DL

techniques in their studies. Figure (3) displayed the PRISMA guidelines for diagnosis of CVDs from CMR

images using DL methods. In addition, the exclusion and inclusion criteria used in this work are provided

in Table (1).

Table 1. Exclusion and inclusion criteria used for the diagnosis of CVDs.

Inclusion

Exclusion

1. CMRI Images

1. Treatment of CVDs

3. Different types of CVDs.

2. Clinical methods for CVDs treatment

3. CVDs detection

3. Rehabilitation systems for CVDs detection (Without AI

techniques)

4. DL models (CNNs, RNNs, AEs, CNN-RNN, CNN-AE,

GAN, Transfer Learning, etc.)

Fig. 3. Literature search procedure.

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

AutomatedDiagnosisofCardiovascularDiseasesfromCardiacMagneticResonanceImagingUsingDeepLearningModels:AReviewMahboobehJafari1,AfshinShoeibi1,2,*,MarjaneKhodatars3,NavidGhassemi1,ParisaMoridian1,NiloufarDelfan4,RoohallahAlizadehsani5,AbbasKhosravi5,SaiHoLing6,Yu-DongZhang7,Shui-HuaWang7,JuanM.Gorriz2,...

展开>> 收起<<

Automat ed Diagnosis of Cardiovascular Disease s from Cardiac Magnetic Resonance Imaging Using Deep Learning Models A Review Mahbo obeh Jafari1 Afshin Shoeibi12 Marjane Khodatars3 Navid Ghassemi1 Parisa Moridian1.pdf

共56页,预览5页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Automat ed Diagnosis of Cardiovascular Disease s from Cardiac Magnetic Resonance Imaging Using Deep Learning Models A Review Mahbo obeh Jafari1 Afshin Shoeibi12 Marjane Khodatars3 Navid Ghassemi1 Parisa Moridian1

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: