Sage Journals: Discover world-class research

Abstract

Data mining provides the methodology and technology to transform huge amount of data into useful information for decision making. It is a powerful process to extract knowledge and discover new patterns embedded in large data sets. Data mining has been increasingly used in medicine, particularly in cardiology. In fact, data mining applications can greatly benefits all parts involved in cardiology such as patients, cardiologists and nurses. This article aims to perform a systematic mapping study so as to analyze and synthesize empirical studies on the application of data mining techniques in cardiology. A total of 142 articles published between 2000 and 2015 were therefore selected, studied and analyzed according to the four following criteria: year and channel of publication, research type, medical task and empirical type. The results of this mapping study are discussed and a list of recommendations for researchers and cardiologists is provided.

Keywords

cardiology data mining knowledge extraction mapping study medical tasks

Introduction

Heart disease is a broad term used to describe a range of diseases that affect the heart. The various diseases that fall under the umbrella of heart disease include diseases caused by damage to the blood vessels (coronary artery disease), valvular heart disease, abnormal heartbeats (heart arrhythmias), thick heart muscle (cardiomyopathy), heart defect and heart infection.¹ The consequences of heart disease are very diverse. They can range, for instance, between high blood pressure, cardiac insufficiency, arrhythmia, stroke, heart attack and sudden cardiac arrest, depending on which part of the cardiovascular system is affected. Thereby, heart disease is a very serious disease that, if not treated properly and on time, can lead to very serious complications, including death. This is why heart disease is considered as one of the main priorities in medical research, which in turn generates huge amounts of data. These increasing volumes of data are very well suited to be processed using data mining (DM) techniques that can handle them with efficiency.

Using DM methods in heart disease studies is one of the best ways in which to use the large volumes of heart disease data sets available to extract knowledge. In fact, according to Fayyad et al.,² DM is defined as “a process of nontrivial extraction of implicit, previously unknown and potentially useful information from the data stored in a database.” DM can therefore be considered as a powerful process to extract knowledge and discover new patterns embedded in large data sets.³ DM techniques have already been applied in various areas for more than four decades. Recent studies show an increase in the amount of applications using DM techniques in several domains such as education,^4,5 clinical medicine,⁶ financial fraud detection^7,8 and genetic data analyzing.⁹

The studies on the application of DM techniques in medicine in general and cardiology in particular have recently undergone a rapid growth in order to assist physicians, improve public health and support patients. To the best of our knowledge, there is no existing systematic mapping study (SMS) that focuses on DM-based empirical studies carried out in cardiology covering all DM tasks and all cardiology disciplines. We were therefore motivated to perform an SMS regarding the application of DM techniques in cardiology in order to analyze the empirical studies available, synthesize the outcomes and identify the challenges in this field of research. We therefore selected and studied 142 papers published in the period from 1 January 2000 to 31 December 2015. Each of these studies was classified with regard to the following: publication channel, year of publication, research type, medical task and empirical type. The objective of this SMS is to summarize and analyze evidence concerning the following: (1) the research type most frequently adopted by researchers, (2) the medical tasks that receive most attention and (3) the empirical type most frequently used.

This article is organized as follows: Section “Related work” presents an overview of the existing reviews in literature regarding the DM-based studies in cardiology. Section “Research methodology” describes the research methodology used in our study. Section “Results” reports the results. Section “Discussion” discusses the findings of this SMS. Section “Implications for researchers” presents implications for researchers. Finally, our conclusion and future work are presented in section “Conclusion.”

Related work

In this section, previous reviews associated with DM in cardiology are outlined. This section aims to better understanding of cardiac knowledge extraction background. For this reason, five review papers are analyzed.

Kumari and Godara¹⁰ reviewed in their study four DM classification techniques to predict cardiovascular disease. They analyzed and compared RIPPER classifier, decision tree (DT), artificial neural networks (ANNs) and support vector machine (SVM) based on their structure and efficiency. The results of this analysis showed that (1) the classification carried out by RIPPER is computationally inexpensive, (2) DT results are easier to read and interpret, (3) ANN draws consistent conclusion and it can be built and evaluated using a large number of cases and (4) SVM is highly accurate, owning to their ability to model complex nonlinear decision boundaries.¹⁰ Shouman et al.¹¹ identified in their study gaps in the research on heart disease diagnosis and treatment by reviewing a set of related papers. This review has identified that the application of DM techniques in the diagnosis of heart disease has a significant success, unlike the treatment stage where less attention has been given. Sivagowry et al. presented in their study of application of DM in heart disease prediction based on a limited number of selected studies (13 papers). The review results showed that classification task in DM plays a vital role in heart disease prediction when compared with prediction, association and clustering. Moreover, text mining the medical data was identified as another extension in predicting the health care data.¹²

Bharti and Singh¹³ studied and compared various DM algorithms that can be used to predict the heart disease in order to identify the best method of prediction of disease. The techniques analyzed include the following: genetic algorithm (GA), particle swarm optimization (PSO), ANN, DT, K-nearest neighbors and Naïve Bayes. The analysis results showed that combining GA, ANN and PSO with various DM techniques will enhance the performance and achieve high accuracy rate. Feng et al.¹⁴ provided in their study insights for the main features of DM and their applications of integrative medical studies in coronary heart disease (CHD). The goal of this study is to analyze the progress and prospect in this field. The results showed that the application of DM methods in clinical intervention to CHD is still in the developing stage, and its own methodology and practical application are continuously improved. Moreover, the study identified four development directions for CHD in the future including (1) following the general rules of DM methods, (2) combining it with the characteristics of clinical practice, (3) continuing to explore suitable DM methodology and (4) getting continuous optimization improvement in practice on the basis of the existing databases.¹⁴

Research methodology

The objective of an SMS is to identify and classify all research related to a broad research area.¹⁵ It provides an overview of a topic area and identifies the quantity and type of research and results available within it. The research questions (RQs) for such a study are quite high level and include issues such as which sub-topics have been addressed, what empirical methods have been used and what sub-topics have sufficient empirical studies to support a more detailed systematic review.

Figure 1 presents the mapping process which involves five steps, namely, definition of RQs, conducting the search for relevant papers, screening of papers, keywording of abstracts and data extraction and mapping. Each process step has an outcome; the final outcome of the process is the systematic map. A mapping study differs from a literature review¹⁶ which reviews primary studies of a specific research area in depth and describes their methodology and results.

Figure 1.

Systematic mapping study process.

RQs

The main goal of our mapping study is to identify, analyze and synthesize articles published since 2000 regarding the application of DM techniques in cardiology. To obtain a more detailed and comprehensive view of the topic, the overall objective was divided into four concrete RQs, which are presented in Table 1 along with their motivations.

Table 1.

Research questions.

No.	Research question	Rationale
Q1	In which sources and in which years were selected studies on the application of DM techniques in cardiology published?	To identify where studies concerning this field of research can be found and whether there are specific publication channels; it also indicates when effort regarding this research area was made.
Q2	What types of investigations are carried out and published in cardiac DM studies?	To provide an overview of different types of research outlined in literature regarding the application of DM in cardiology.
Q3	What are the disciplines of cardiology in which DM techniques were applied?	To determine in which cardiology tasks DM techniques were carried out and researchers were interested.
Q4	What types of empirical studies were conducted when applying DM techniques in cardiology?	To identify the kind of evidence that was developed in studies concerning the application of DM in cardiology in order to evaluate the maturity of the approaches.

Search strategy

Once the RQs have been determined, it is essential to identify and precisely state the search strategy to be followed. The primary studies were therefore identified by means of a search string which was used in the following eight digital databases: IEEE Xplore, SpringerLink, DBLP Computer Science Bibliography, ScienceDirect, ACM Digital Library, MEDLINE (Medical Literature Analysis and Retrieval System Online), Embase (Excerpta Medica Database) and CINAHL (Cumulative Index to Nursing and Allied Health Literature). The reasons for choosing these digital libraries were as follows: (1) they collect articles covering the fields of computer science, medical informatics and cardiology, which represent the areas of interest for this study; (2) they index millions of articles from diverse types of publication channels such as journals, conferences and workshops and (3) they have been used as sources for previous systematic reviews related to computer science in cardiology.³

The search string should provide the maximum coverage but be of a manageable size. Thus, in order to determine the search string, the major terms matching the RQs presented in Table 1 were derived and the synonyms for the main terms were identified. The Boolean OR was used to join alternative terms and the Boolean AND was used to link the major parts.¹⁷

The complete set of search string was designed as follows:

(cardi* OR Electrocardiogram OR echocardiographic OR “Coronary artery disease” OR “arrhythmia” OR myocardial OR (Heart AND (disease OR failure OR diagnosis OR prognosis OR treatment OR monitoring OR beat OR rate variability)))

AND

(“data mining” OR intelligent OR predict* OR classificat* OR cluster* OR associat*)

AND

(model* OR algorithm* OR technique* OR rule* OR method* OR tool* OR framework*)

Study selection

The selection process aims to filter the candidate papers that were selected by applying the search string to the libraries chosen in order to identify the most relevant articles, based on their titles, abstracts and/or keywords, as regards the objective of this study. This signifies that when several papers report the same study, only the most recent should be considered and included. Moreover, in the case of repeated studies found in different publication channels, only one study is considered.

Inclusion and exclusion criteria were applied to the candidate papers so as to identify and include studies that would be relevant to answer the RQs in Table 1. We therefore defined the following inclusion criteria in order to select the most appropriate publications. The inclusion criteria (IC) are linked using the OR Boolean operator:

IC1: Empirical studies presenting newly developed DM-based models in cardiology, either evaluated or not.

IC2: Empirical studies validating an existing DM-based model in cardiology.

IC3: Papers providing only comparisons between DM techniques in cardiology without proposing any new developed models.

With regard to the exclusion criteria, papers that met at least one of the following exclusion criteria (EC) were excluded:

EC1: Papers presenting studies on the application of DM techniques in medical disciplines other than cardiology.

EC2: Papers published earlier than 2000 and later than 2015.

EC3: Papers presenting an overview of existing approaches that apply DM techniques in cardiology.

Each candidate paper was evaluated by two authors in order to decide whether it should be included, by considering its title, abstract and keywords. Articles that were judged differently were discussed by the two authors using either the partial text or the full text until an agreement was reached. The reference lists of the relevant papers were also scanned by the authors to find extra relevant papers and add them to the set.

Quality assessment

In order to support the inclusion/exclusion process, it is important to analyze and assess the quality of primarily selected studies for them to be eventually included in the data extraction and reporting process. In fact, the aim of assessing the quality is to ensure that the study findings are relevant and unbiased.¹⁸ We therefore identified a set of quality assessment (QA) questions so as to enhance our study and evaluate the relevance and rigorousness of the candidate studies. The design of this questionnaire was inspired by previous SMSs.^18–20

QA1: The study presents empirical evidence that is analyzed quantitatively or qualitatively. The possible answers are “quantitative research (+1)” and “qualitative research (+0).”

QA2: The study presents an experimental design that is justifiable and detailed. The possible answers are “yes: +1,” “no: 0” and “partially: 0.5.”

QA3: The study measures and reports the accuracy rate of the experimental method. The possible answers are “yes: +1” and “no: 0.”

QA4: The study presents a comparison between the proposed empirical method and other methods. The possible answers are “yes: +1” and “no: 0.”

QA5: The study explicitly analyzes the benefits and limitations of the study. The possible answers are “yes: +1,” “no: 0” and “partially: 0.5.”

QA6: The study has been published in a recognized and stable publication source. The possible answers to this question were as follows:

For conferences, workshops and symposia:

(+1.5) if it is ranked CORE 2014 A;

(+1) if it is ranked CORE 2014 B;

(+0.5) if it is ranked CORE 2014 C;

(+0) if it is not in CORE 2014 ranking.

For journals:

(+2) if it is ranked JCR 2014 Q1;

(+1.5) if it is ranked JCR 2014 Q2;

(+1) if it is ranked JCR 2014 Q3 or Q4;

(+0) if it has no JCR 2014 ranking and for other journals.

QA2 is rated partially when the experimental design and the techniques used are explained in a very limited way or one of the techniques used is not detailed. QA5 is scored partially when only one and not both of the benefits and challenges of the study are reported. QA6 is rated based on the CORE Conference Ranking Exercise 2014²¹ and the Journal Citation Reports (JCR) 2014.²² The score for QA6 was identified based on the fact that journals are more advantageous than conferences, workshops and symposia because the possibility of publishing works in Q1 or Q2 journals may be more difficult than in other publication channels.²³

In order to ensure the efficiency of the outcomes of this study, we considered the quality score as an exclusion criterion and therefore selected only those relevant studies with a quality score that exceeded 3.5, which is 50 percent of the perfect score: 7. The complete list of the studies selected and details of their QA can be found in Table A.1 (Appendix).

Data extraction strategy and synthesis

A data extraction form was used to collect all relevant data from the selected studies in order to provide answers to the RQs in Table 1.

RQ1: To answer this question, it is necessary to identify the publication source and channel for each paper. The articles should also be classified per publication year.

RQ2: A research type can be categorized as being of the following types:¹⁶

Evaluation research (ER): An assessment of one or several DM-based techniques in cardiology is conducted.

Solution proposal (SP): The development of a new DM-based technique for cardiology. This can also be an important improvement to an existing DM-based technique.

Experience papers (EP): The researchers report their personal experience when testing the DM-based model developed in a cardiac unit and present the results obtained.

RQ3: The disciplines of cardiology can be classified as follows:³

Screening: This is used to identify the existence of a disease even though its symptoms have not yet appeared. This is carried out with people who appear in good health.

Diagnosis: This is considered to be one of the crucial tasks in cardiology since it attempts to identify the nature of a disease. The selection of an appropriate treatment method is therefore based on the results of a diagnosis procedure.

Prognosis: This consists of predicting current disease outcomes in terms of mortality and morbidity and the patient’s chances of recovery.

Treatment: This aims to allow the patient to recover and prevent the disease from progressing.

Monitoring: This concerns the observation phase of the disease and the patient’s condition over the time.

Management: This is related to health promotion and medical services.

RQ4: The empirical research type can be classified as follows:^24,25

Survey: An approach that allows the necessary information to be gathered by means of a questionnaire that asks several questions.

Historical-based evaluation (HBE): This consists of looking for patterns in historical data.

Case study: An empirical validation of DM techniques in cardiology using data sets collected from hospitals and cardiac units.

In order to synthesize the data extracted to answer the RQs in Table 1, we employed different strategies. The studies selected were therefore ranked on the basis of their QAs and classified in each of the RQs’ answers. A narrative synthesis method was used for all RQs. This method was supported by visualization tools such as tables, a pie chart and a bar chart in a manner consistent with the questions.

Threats to validity

The main threats to the validity of this study are analyzed in the following three respects:

Study selection bias: The selection of relevant studies for this mapping study depended on the search strategy, the literature sources, the selection criteria and the quality criteria. We have therefore constructed the search string to match the RQs identified and have used them to retrieve the relevant studies in the five electronic databases listed above. However, it is possible that some relevant studies were not returned in the autonomic search. We addressed this threat by manually scanning the reference lists of the selected studies to look for extra relevant studies that were not covered by the search in the five electronic databases. Moreover, in order to prevent the relevant studies from being excluded incorrectly, we defined the selection criteria and QAs that strictly comply with the RQs. The final decision as regards study selection was made by means of a double confirmation. In fact, two researchers initially conducted the process of selecting the relevant studies separately, and any disagreements between these researchers were then discussed until a final agreement was reached.

Publication bias: It is well known that in each study, researchers attempt to prove the performance of their DM-based model and may tend to claim that their model outperforms others. There is consequently a risk of overestimating the performance of the DM-based models developed. Fortunately, the fourth inclusion criterion in this study may reduce this threat because it aims to identify the studies that simply conduct comparisons of the DM-based models developed for cardiology and does not propose a new developed model. These comparative studies report the comparison results in an objective way. In these studies, therefore, both positive and negative comparison results are likely to be reported, which to some extent helps reduce publication bias.

Data extraction bias: Data extraction was also one of the most critical tasks in this study. In order to reduce the threat of inaccurate data extraction, the studies were read and analyzed independently by two researchers and the data required to provide answers to the RQs were collected. All disagreements between the extractor and checker were resolved by discussion among all the researchers.

Results

This section presents the findings related to this study. First, we introduce an overview of the result of the selection process; and second, all the results for each RQ are reported.

Overview of the selected studies

A search in the eight electronic databases identified above permitted the selection of 2392 candidate papers, as shown in Figure 2. Of the 2392 papers, 1992 papers were discarded after applying the exclusion criteria, thus leading to the identification of 400 articles regarding the use of DM techniques in cardiology. In fact, the duplicate papers were first discarded and only one study was considered. Besides, the papers reporting the same study were also excluded and only the most recent one was included. Then, the exclusion criteria EC1 and EC2 were applied on the remaining papers leading to the exclusion of 1992 papers. These papers were analyzed in depth in order to select only the most relevant articles. In fact, after scanning the titles, abstracts and keywords, only papers that met at least one of the inclusion criteria were included. At the end of this stage, 166 papers were discarded and 234 articles were included. In the next phase, the reference lists of these relevant papers were scanned and this resulted in the discovery of 21 extra relevant papers that had been missed in the initial search. Finally, in order to assess the quality of the selected papers based on their full texts, the QAs were applied to the remaining studies, resulting in the exclusion of 113 papers with low-quality scores; 142 papers were therefore eventually identified as the final selected studies, which were then used to obtain data extraction information in order to provide answers to the RQs of this review.

Figure 2.

Result of selection process.

We used QA to evaluate the quality of the relevant studies. Hence, since the perfect quality score was 7, the quality score of each selected paper had to exceed 3.5. Table 2 summarizes the results of the QA process. Of the articles selected, 56 percent had a higher or equal average score and 44 percent of the papers were under the average score and were therefore discarded. Furthermore, 15 percent had an average score and 41 percent of the studies had a high quality score. The complete list of these 142 selected papers with details of the overall classification results can be found in Table A.1 (Appendix).

Table 2.

Results of quality assessment.

Quality level	No. of studies	%
Very low (0 ⩽ score ⩽ 1.5)	13	5
Low (2 ⩽ score ⩽ 3)	100	39
Medium (3.5 ⩽ score ⩽ 4)	37	15
High (4.5 ⩽ score ⩽ 5.5)	83	32
Very high (6 ⩽ score ⩽ 7)	22	9

RQ1: In which sources and in which years were selected studies on the application of DM techniques in cardiology published?

The selected articles were published in several channels, as shown in Table 3 in which publication channels that were repeated more than once and the number of papers per publication source is listed. Two publication channels were therefore identified, which are journals and conferences. Of the selected papers, 26 percent were presented at conference proceedings and 74 percent were published in journals. According to Table 3, the most frequent targets are journal of Computer Methods and Programs in Biomedicine (7.3%), Expert Systems with Applications (6.1%) and IEEE Transactions on Information Technology in Biomedicine (4.6%); and the most frequent conferences are the International conference of the IEEE Engineering in Medicine and Biology Society (3.3%) and the International Conference on Biomedical Engineering and Informatics (1.3%).

Table 3.

Publication sources.

Publication source	Type	No. of papers	%
Computer Methods and Programs in Biomedicine	Journal	11	7.3
Expert Systems with Applications	Journal	9	6.1
IEEE Transactions on Biomedical Engineering	Journal	7	4.6
IEEE Transactions on Information Technology in Biomedicine	Journal	7	4.6
Computers in Biology and Medicine	Journal	5	3.3
Journal of Medical Systems	Journal	5	3.3
Artificial Intelligence in Medicine	Journal	4	2.6
Applied Soft Computing	Journal	3	2.1
BMC Medical Informatics and Decision Making	Journal	3	2.1
Medical & Biological Engineering & Computing	Journal	2	1.3
International Journal of Computer Applications	Journal	2	1.3
Annual International Conference of the IEEE Engineering in Medicine and Biology Society	Conference	5	3.3
First International Conference on Computational Intelligence: Modeling Techniques and Applications	Conference	2	1.3
International Conference on Biomedical Engineering and Informatics	Conference	2	1.3
Other journals		51	34.2
Other conferences		31	20.8

Figure 3 shows the distribution of the selected studies between 2000 and 2015. It will be observed that the number of papers increases during this period of time. In fact, the frequency of papers published per year has increased from one article in 2000 to 21 papers in 2015. Moreover, there is an outlier in the years 2012, 2013, 2014 and 2015 because 55 percent of the selected studies were published in these years. These outliers were identified on the basis of the number of publications that exceeded 12 publications per year. It will also be noted that the number of publications concerning the application of DM in cardiology was low between 2000 and 2006. However, since 2007, the number of publications has increased significantly, as 91 percent of the selected studies were published in the period between 2007 and 2015.

Figure 3.

Number of papers published per year and publication channel.

RQ2: What types of investigations are carried out and published in cardiac DM studies?

The data extraction phase allowed us to identify three research types, namely, SP, ER and experience report. Of the selected studies, 25 percent were identified as SP papers, that is, papers that present new DM-based techniques to be applied in cardiology, and 10 percent belong to the ER type since they evaluate the performance of existing DM-based techniques in cardiology. Moreover, 63 percent of the selected articles were identified as being simultaneously SP and ER papers, and 2 percent of the selected studies were identified as EPs.

Figure 4 shows the evolution of the research types of the selected studies over time. In this chart, it will be noted that EPs have only begun to appear in the last few years. However, SP studies first appeared in 2000 and have increased over the years. Moreover, ER papers appeared 2 years after the appearance of SP papers and have also increased over the years.

Figure 4.

Evolution of research types identified over the years.

RQ3: What are the disciplines of cardiology in which DM techniques were applied?

All the activities in cardiology can be divided into six categories that are referred to as “medical tasks”:³ screening, diagnosis, treatment, prognosis, monitoring and management. Figure 5 provides an overview of the frequency of the distribution of the selected studies as regards cardiology disciplines and DM objectives. Of the studies, 55 percent were dedicated to the diagnosis stage, 26 percent of the selected studies were related to a screening procedure which concerns cardiologists and 8 percent of the selected papers have applied DM techniques in monitoring. Nearly 7 percent of the studies were interested in prognosis procedures. With regard to the treatment phase, 2 percent of the selected studies were interested in this discipline, while only 2 percent of the papers carried out research for the management phase.

Figure 5.

Frequency of medical tasks with DM objectives.

Figure 5 shows the frequency of published papers in six medical tasks. In each task, the frequency of DM objectives is presented. Classification algorithms were used in all cardiology disciplines with the exception of the management task. Classification algorithms would therefore appear to be those most frequently used in several tasks, and particularly, the diagnosis task in which 59 percent of the classification studies were performed. Prediction algorithms were also adopted by several researchers in all medical tasks; 47 percent of prediction-based studies were carried out for diagnosis tasks, while the remaining studies were distributed throughout the remaining medical tasks. Association rule–based studies were carried out only in diagnosis, screening, prognosis and treatment tasks and they were also more focused on diagnosis. Data clustering algorithms were used only in diagnosis, screening, prognosis and monitoring tasks and were interested in the diagnosis stage.

RQ4: What types of empirical studies were conducted when applying DM techniques in cardiology?

Figure 6 shows that all the papers were empirically evaluated using several empirical research types. Of the selected papers, 54 percent belong to the HBE type (77 of 142 papers) since they employed publicly available heart disease databases. Some of these studies are identified as being simultaneously SPs and ER papers, which is why they were double calculated in these two types of research; 57 percent of the papers using public data sets therefore proposed new solutions and 43 percent of them evaluated the models developed. Other researchers preferred to obtain the data set from care units at which they were able to carry out their experiments, and 43 percent of the selected studies (61 of 142 papers) were therefore empirically validated by means of case studies. This empirical type was used by researchers in their SPs (52 out of 90 papers), evaluations (38 out of 90 papers) or experience studies (2 out of 90 papers). Only 3 percent of the selected studies (4 out of 142 papers) were evaluated using surveys.

Figure 6.

Distribution of research types and empirical types.

Several public cardiac data sets were used to construct and evaluate the studies selected in this study. Table 4 shows the most frequently used data sets with their relevant information. The relevant information includes the number and percentage of the selected studies using the data set, the number of instances included in the data set, the number of attributes and the source of the data set. The two databases most frequently used in the selected studies were the MIT-BIH Arrhythmia Databases and Cleveland Heart Disease Database, with percentages of 23 and 20 percent, respectively.

Table 4.

Data sets used in the selected studies.

Data set	No. of studies	%	No. of example	Attribute	Source
MIT-BIH Arrhythmia Database	33	23	48	–	26
Cleveland Heart Disease Database	29	20	303	76	27
SPECT Heart Data Set	9	6	267	22	28
Hungary Heart Disease Database	7	5	294	76	27
UCI Arrhythmia Data Set	5	3	452	279	29
Switzerland Heart Disease Database	2	1	123	14	27
Specific data sets	61	43	Min: 41Max: 4850Mean: 823	Min: 6Max: 279Mean: 74	30–43

SPECT: Single Proton Emission Computed Tomography.

Moreover, 6 percent of the selected studies used the Single Proton Emission Computed Tomography (SPECT) heart data set, while the Hungarian heart disease database, the UCI Arrhythmia Data Set and the Switzerland Heart Disease Database were used in only 5, 3 and 1 percent, respectively, of the selected studies. All the aforementioned databases, with the exception of the MIT-BIH arrhythmia database, belong to the UCI Machine Learning Repository, which is a collection of databases, domain theories and data generators that are used by the machine learning community for the empirical analysis of machine learning algorithms.²⁷ Although public data sets were widely used, specific data sets were also adopted in several of the selected studies; 43 percent of the databases used were collected from hospital records or from observation per month of group of patients.

Discussion

The objective of this mapping study was to identify and analyze studies carried out on the application of DM techniques in cardiology, and 142 papers were therefore selected and analyzed from a total of 2392 studies. In this section, the principal findings of this SMS will be presented and discussed.

RQ1: In which sources and in which years were the selected studies on the application of DM techniques in cardiology published?

Table 3 shows the variety of publication sources, since 85 different sources in 142 articles were recorded, which demonstrates that there is no specialized source as regards the application of DM techniques in cardiology. In fact, these publication sources are specialized in computer science, the medical field and computer science applied to medicine. It is well known that these three research areas are very broad and encompass several research topics. There are also several publication channels that specialize in the aforementioned research areas, which is why a variety of publication sources have been recorded in this study. The publication channels selected were scored to measure their reputation, and the percentage of papers published in recognized sources is therefore quite good, at around 75 percent. However, 25 percent of the publication channels were not well recognized particularly in the case of conferences since only 21 out of 34 conferences were ranked. This can be explained by the fact that in computer science, the number of low-level conferences has undergone a tremendous growth.⁴⁴

Furthermore, Figure 3 shows the increasing number of papers since 2000, which can be explained by the fact that the research area regarding the application of DM techniques in cardiology has attracted the attention of researchers owing to its benefits. It will also be noted that the number of articles published in the period between 2000 and 2006 was low, which might be explained by the fact that the application of DM techniques in the medical field in general came to be of significantly more interest at the beginning of the last decade, although the number of publications in this context was even lower in the first 7 years of the last decade until 2007 when publications about the particular application of DM in cardiology made a significant appearance. Moreover, 91 percent of the selected studies were published in the period between 2007 and 2015, a period during which outliers were also identified because DM techniques were proven to be powerful tools that provide efficient solutions for the medical field in general³ and the cardiac domain in particular. This encouraged researchers to carry out more research in this respect. Moreover, heart disease is considered to be one of the most high-risk diseases and therefore attracted more attention than other diseases, and this is reflected in the growing number of works performed at this level.³ Nevertheless, it can be noted that the number of articles published in 2010 and 2011 have decreased due to the selection process adopted in this study. In fact, after applying the IC, several relevant studies published in 2010 and 2011 were selected. However, they were excluded after the application of QA which may explain the decrease of publications in 2010 and 2011.

What types of investigations are carried out and published in cardiac DM studies?

As can be seen in Figure 4, the first studies to appear were SP papers because researchers were more interested in applying DM techniques in cardiology, especially after the good results obtained when applying these techniques in other disciplines of medicine.³ They therefore identified the needs of the cardiac area, where DM techniques could be perfectly applied and attempted to propose new solutions that would assist physicians.^45,46 ER papers appeared 2 years later, which can be explained by the fact that researchers needed to evaluate the studies that were carried out so as to measure their performance and identify the gaps and limits in this respect. New ideas regarding those techniques that outperform others were also proposed, along with how the performance of the systems developed could be enhanced.^47,48

Moreover, some selected studies were identified as being both SP papers and ER papers, which can be explained by the fact that some researchers preferred to provide the evaluation of their developed models in conjunction with the proposed solutions so as to prove their efficiency. Moreover, Figure 4 shows that SP and ER studies have increased over the years and are still published in a considerable number of journals each year. This is owing to the fact that the research area of cardiac DM has not yet reached the level of maturity, and it is therefore still necessary to carry out a considerable amount of research in order to fill the gaps identified. Of these gaps, we can mention the following: the small size of available data sets,^32,49 the data incompleteness in databases⁵⁰ and the lack of real cases for validation purposes.³⁷

Furthermore, only 2 percent of the selected studies were identified as EPs, which may be explained by the fact that finding a real cardiac unit at which to test a DM model is not an easy task. It needs to be supervised by cardiologists to avoid any mistakes, which may not be always possible owing to their numerous commitments.

What are the disciplines of cardiology where DM techniques were applied?

Of all medical tasks, researchers were most interested in diagnosis. With regard to Figure 5, 55 percent of the studies were dedicated to the diagnosis stage, which can be explained by the importance of this procedure in medical activities. In fact, performing an efficient diagnosis allows cardiologists to select an appropriate treatment method and reduce the mortality rate of patients. This is why several researchers were interested in the diagnosis procedure in order to produce intelligent systems capable of helping cardiologists and providing an efficient and accurate diagnosis.^48,51 Of the selected studies, 26 percent were related to screening procedures, which are of interest to cardiologists since they allow the disease to be detected at an early stage. In fact, researchers at this level were focused on studying heart rate variability (HRV) signals and applying DM techniques to them in order to identify arrhythmias.^52,53

Of the selected studies, 8 percent were focused on the monitoring phase, which is owing to the importance of this stage for heart disease patients. In fact, the observation phase of the disease and the patient’s condition over time is an extremely significant process that enables a rapid response to take place in the case of an emergency. Moreover, the need to know the necessary and appropriate action to take at any time when an unusual emergency situation is occurring signifies that e-health services that improve the access, efficiency and quality of health care services are privileged. The rapid development of information and communication technologies therefore led to more studies regarding the development of home care systems being introduced.^54–56 These systems aim to send the real-time monitoring of physical information to a central health management system, which would serve as a platform on which doctors could communicate with patients who need help.

Of the selected studies, 7 percent applied DM techniques at the prognosis stage so as to provide an efficient system that is capable of predicting the patient’s chances of recovery and the disease outcomes in terms of mortality and morbidity.⁴⁶ The treatment task was not adopted in several papers, which can be explained by the fact that treatment is known to be the most difficult task for automation because it is directly associated with human life. It is, moreover, relative to the diagnosis stage: the more the diagnosis is correct, the more the treatment is appropriate. Little work was carried out on management procedures which can be explained by the fact that cardiologists pay more attention to diagnosis and screening stages than other tasks. Researchers were therefore more interested in and concentrated on producing systems that provide efficient decision support systems for diagnosis and screening procedures.

What types of empirical studies were conducted when applying DM techniques in cardiology?

Of the selected papers, 54 percent belong to historical-based ER, which can be explained by the fact that several researchers used publicly available databases for heart disease in order to evaluate their DM models. It is also easier to obtain access to a public database than a real database from a hospital owing to confidentiality constraints. Some researchers therefore preferred to start testing and evaluating the models they developed by first using public data sets and then if possible using a real database. Asl et al.⁵⁷ presented an effective cardiac arrhythmia classifier using HRV signals. The proposed algorithm was implemented using the support vector machine (SVM) as a basis and was evaluated using the MIT-BIH Arrhythmia Database.²⁹ This database includes electrocardiogram (ECG) signals and is a standard reference for studies on the cardiac arrhythmia detection and classification problem. Other publicly available databases were adopted in the selected studies, as shown in Table 4. The Cleveland heart disease database was adopted in 20 percent of the studies that used public databases because it is considered to be the most complete data set when compared to other public data sets such as the Hungary and Switzerland Heart Disease Databases.²⁷

Other researchers preferred to obtain data sets from care units at which they could carry out their experiments. In fact, 43 percent of the selected studies were empirically validated using case studies based on the data sets collected from hospitals and care units. Researchers used this means in an attempt to extract knowledge from real data sets and identify gaps that might be encountered in such databases in order to develop a better model that reflects real life. Only 3 percent of the selected studies were identified as surveys. This may be explained by the fact that collecting information directly from patients is not a simple procedure. Moreover, the cardiac field is critical because any small error could be fatal, which is why relevance is highly recommended. However, surveys rely on users’ subjective memories and rating scales in questionnaires which may sometimes have an influence on the relevance of data filled out in these questionnaires. Moreover, the relevance of the information collected must be validated by expert cardiologists who may not always be available owing to their workload. A lot of time may therefore be wasted owing to the lack of collaboration between academics and doctors.

Implications for researchers

This section shows recommendations for researchers and guidelines for practitioners (cardiologists, care units, etc.). These implications will be categorized per RQ as follows:

RQ1: Although the majority of the selected studies were published in journals, many of them were published at conferences. However, they were discarded during different stages of the study selection process because of the lack of relevant content for our research into these studies. It is therefore recommended that researchers provide relevant and concise information in their studies even if it is not detailed because of the reduced number of pages included in the proceedings of conferences. This concise and brief information will avoid the exclusion of interesting papers presented at conferences in the selection process.

RQ2 and RQ4: Several databases were used to evaluate the DM models that were developed. Some of these data sets were publicly available while the others were extracted from heart patients’ records from various hospitals. However, most of these databases include a limited number of records. Researchers are therefore encouraged to use large data sets containing a sufficient number of records or attributes to evaluate and validate DM models. The use of large data sets is necessary since the medical database is considered to be one of the largest databases. The accuracy of the DM models developed should be tested and evaluated on large databases in order to rely on the efficiency of the systems implemented. Note that a slight medical error could lead to the loss of human lives.

RQ3: Of the six medical tasks, researchers were more interested in diagnosis and screening owing to their importance for heart disease. However, few works were carried out for the treatment and management tasks. Researchers are therefore encouraged to conduct more studies on treatment procedures because even if treatment might not appear to be the most difficult procedure, this is well-supported by physicians. It would also be possible for researchers to work on both medical tasks: diagnosis and treatment. This combination could provide a powerful system since it allows the diagnosis of heart disease and the proposal of a treatment according to the diagnosis established. Such a system could be of great help to cardiologists. However, specialists’ opinions will continue to be adopted because these systems only assist cardiologists to make better clinical decisions.⁵⁸

Conclusion

The growing number of heart disease patients has generated huge amounts of data which has encouraged researchers adopting DM techniques to handle these increasing volumes of data efficiently. Several DM models were developed in order to provide efficient solutions and help cardiologists and care stakeholders. These models were developed using various DM techniques and were applied in different disciplines of cardiology. Moreover, the developed models were evaluated using various evaluation methods and different types of cardiac databases.

This mapping study investigated the studies that were performed in cardiology using DM techniques. A total of 400 papers from between 2000 and 2015 were therefore identified. After applying our study selection process, 142 empirical studies were selected for this study in order to answer the RQs shown in Table 1. These answers permitted the findings to be classified according to the following criteria: year and channel of publication, research type, cardiac task and empirical type. The results obtained showed that the amount of studies conducted on cardiac DM has risen during the last decade. These papers were published in different journals and appeared at several conferences regarding computer science, the medical field and computer science applied to medicine. The majority of the selected articles were identified as SP and/or evaluation papers. The main empirical types found were HBE and case studies. Around half of the selected studies used historical data sets to evaluate the models developed. Diagnosis and screening were identified as the most frequent medical tasks for which studies were conducted, while less attention has been paid to the other tasks.

This study provides also recommendations for researchers and cardiologists. In the case of researchers, we recommend that more effort should be made as regards medical tasks other than diagnosis and screening, and particularly the treatment phase. Moreover, in order to validate the DM models, clinical validation should be performed to be able to ensure the efficiency of the model developed. As for the weakness of application of DM in cardiology, we should mention the lack of a standard concerning the overall knowledge extraction process from data gathering to knowledge evaluation. For instance, a standard could be imposed for the data pre-processing phase in order to produce knowledge, and a means could then be found to locate this knowledge in the cardiology process. This weakness is not limited to cardiology, since no such standard exists for the entire medical field.³

Footnotes

Appendix A. Classification Results

Table A.1

Classification of the selected studies.

Paper ID	Reference	RQ1			RQ2	RQ3	RQ4	Quality assessement
Paper ID	Reference	Channel	source	year	RQ2	RQ3	RQ4	(a)	(b)	(c)	(d)	(e)	(f)	Score
S1	[23]	Journal	DBLP	2012	SP+ER	diagnosis	HBE	1	1	1	0.5	0	0	3.5
S2	[24]	Journal	sciencedirect	2012	SP+ER	diagnosis	HBE	1	1	1	0	0	1.5	4.5
S3	[25]	Journal	IEEE Xplore	2006	SP	screening	Case study	1	1	1	1	0.5	1.5	6
S4	[26]	Journal	springerLink	2013	SP	diagnosis	HBE	1	1	0	0	1	1	4
S5	[27]	Journal	ACM	2010	SP	diagnosis	HBE	1	1	1	0	0.5	2	5.5
S6	[28]	Conference	IEEE Xplore	2009	ER	diagnosis	survey	1	1	1	0	0	0.5	3.5
S7	[29]	Conference	sciencedirect	2013	SP	diagnosis	HBE	1	1	1	0.5	0	0	3.5
S8	[30]	journal	springerLink	2004	SP	screening	HBE	1	1	1	0	0.5	0	3.5
S9	[31]	Journal	sciencedirect	2013	SP	diagnosis	HBE	1	1	1	1	0	0	4
S10	[32]	Journal	IEEE Xplore	2013	SP	prognosis	HBE	1	0.5	1	1	1	0	4.5
S11	[33]	Journal	sciencedirect	2011	SP+ER	diagnosis	HBE	1	1	1	1	0	0	4
S12	[34]	Conference	IEEE Xplore	2014	SP+ER	diagnosis	Case study	1	1	1	1	0.5	0	4.5
S13	[35]	Journal	sciencedirect	2012	SP	diagnosis	Case study	1	1	1	0	1	2	6
S14	[36]	Conference	DBLP	2012	SP+ER	prognosis	HBE	1	1	1	1	0	0	4
S15	[37]	Journal	ACM	2013	SP	prognosis	HBE	1	1	1	0	0.5	0	3.5
S16	[38]	Conference	springerLink	2014	ER	treatment	HBE	1	1	1	0	0.5	1.5	5
S17	[39]	Journal	DBLP	2008	SP	screening	HBE	1	1	1	1	0.5	1.5	6
S18	[40]	Journal	DBLP	2012	SP	screening	HBE	1	1	1	1	0.5	2	6.5
S19	[41]	Journal	sciencedirect	2005	SP	diagnosis	HBE	1	1	1	1	0	1	5
S20	[42]	journal	sciencedirect	2010	SP	screening	HBE	1	1	1	0	0	1.5	4.5
S21	[43]	Conference	ACM	2007	SP+ER	diagnosis	Case study	1	1	1	0.5	0	0	3.5
S22	[44]	Journal	IEEE Xplore	2011	SP	diagnosis	HBE	1	1	1	0	0	2	5
S23	[45]	Journal	springerLink	2012	SP+ER	Screening	HBE	1	1	1	1	0	1.5	5
S24	[46]	Conference	IEEE Xplore	2013	SP	prognosis	Case study	1	1	1	0	0	0.5	3.5
S25	[47]	Conference	IEEE Xplore	2009	SP	monitoring	Case study	1	1	1	1	0	1.5	5.5
S26	[48]	Conference	IEEE Xplore	2007	EXP	screening	Case study	0	1	1	0.5	1	0	3.5
S27	[49]	Conference	IEEE Xplore	2009	SP	diagnosis	HBE	1	1	1	1	0	0	4
S28	[50]	Conference	IEEE Xplore	2012	SP	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S29	[51]	Conference	IEEE Xplore	2013	EXP	prognosis	Case study	1	1	1	0	0	0.5	3.5
S30	[52]	Conference	IEEE Xplore	2010	SP	diagnosis	Case study	1	1	1	0.5	0	0	3.5
S31	[53]	Journal	springerLink	2012	SP+ER	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S32	[54]	Journal	springerLink	2012	SP	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S33	[55]	Conference	IEEE Xplore	2012	ER	monitoring	Case study	1	1	1	1	1	0	5
S34	[56]	Conference	IEEE Xplore	2011	EXP	screening	survey	1	1	1	0	1	0.5	4.5
S35	[57]	Conference	IEEE Xplore	2011	SP+ER	diagnosis	HBE	1	1	1	0	0.5	0	3.5
S36	[58]	Journal	sciencedirect	2009	SP+ER	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S37	[59]	Journal	IEEE Xplore	2008	SP	diagnosis	Case study	1	1	1	1	0.5	2	6.5
S38	[60]	Journal	IEEE Xplore	2011	ER	screening	Case study	1	1	1	1	0.5	2	6.5
S39	[61]	Conference	IEEE Xplore	2008	ER	management	Case study	1	1	1	0	0	1.5	4.5
S40	[62]	Journal	sciencedirect	2012	SP+ER	prognosis	Case study	1	1	1	1	0	0	4
S41	[63]	Journal	sciencedirect	2010	ER	diagnosis	HBE	1	1	1	1	0	2	6
S42	[64]	Journal	sciencedirect	2012	SP+ER	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S43	[65]	Conference	IEEE Xplore	2013	ER	monitoring	HBE	1	1	1	0	0	0.5	3.5
S44	[75]	Conference	springerLink	2007	ER	diagnosis	survey	1	1	1	1	0	0.5	4.5
S45	[76]	Conference	ACM	2011	SP+ER	diagnosis	HBE	1	1	1	1	0	0	4
S46	[77]	Journal	DBLP	2009	SP	diagnosis	Case study	1	1	1	0	0	1	4
S47	[78]	Journal	springerLink	2009	SP+ER	diagnosis	Case study	1	1	1	1	0	1	5
S48	[79]	Journal	sciencedirect	2013	SP+ER	screening	HBE	1	1	1	0	0	1.5	4.5
S49	[80]	Journal	sciencedirect	2009	ER	screening	Case study	1	1	1	0	0	1.5	4.5
S50	[81]	Journal	sciencedirect	2008	ER	diagnosis	Case study	1	1	1	0	0	1.5	4.5
S51	[82]	Journal	sciencedirect	2012	SP+ER	monitoring	HBE	1	1	1	1	0	1.5	5.5
S52	[83]	journal	IEEE Xplore	2008	SP	monitoring	Case study	1	1	1	0	0	1.5	4.5
S53	[84]	Journal	sciencedirect	2010	ER	diagnosis	Case study	1	1	1	0	0	1.5	4.5
S54	[85]	Conference	IEEE Xplore	2008	SP	diagnosis	Case study	1	1	1	1	0	0	4
S55	[87]	Journal	sciencedirect	2009	SP+ER	screening	Case study	1	1	1	1	1	2	7
S56	[88]	Journal	sciencedirect	2007	SP	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S57	[89]	Journal	DBLP	2011	SP+ER	diagnosis	HBE	1	1	1	0	1	0	4
S58	[90]	Journal	DBLP	2012	SP+ER	diagnosis	Case study	1	1	1	1	0	0	4
S59	[91]	Journal	sciencdirect	2013	SP+ER	diagnosis	Case study	1	1	1	1	0	2	6
S60	[92]	conference	IEEE Xplore	2008	SP+ER	diagnosis	HBE	1	1	1	0	1	0.5	4.5
S61	[93]	Journal	DBLP	2009	SP+ER	screening	Case study	1	1	1	0	0	2	5
S62	[94]	Journal	springerLink	2011	SP+ER	treatment	survey	1	1	1	1	1	0	5
S63	[95]	Journal	IEEE Xplore	2006	SP	diagnosis	Case study	1	1	1	0	0	2	5
S64	[96]	Journal	springerLink	2014	SP+ER	screening	Case study	1	1	1	1	0.5	1	5.5
S65	[97]	Journal	sciencedirect	2009	SP+ER	diagnosis	Case study	1	1	1	1	0	2	6
S66	[98]	Journal	DBLP	2012	SP	diagnosis	Case study	1	1	1	1	0	0	4
S67	[99]	Journal	sciencedirect	2005	ER	diagnosis	Case study	0	1	1	1	0	2	5
S68	[100]	Journal	IEEE Xplore	2011	SP+ER	monitoring	HBE	1	1	1	1	0	1.5	5.5
S69	[101]	Journal	springerLink	2010	ER	diagnosis	Case study	1	1	1	1	0	1.5	5.5
S70	[102]	Journal	springerLink	2007	SP	prognosis	HBE	1	1	1	1	0	0	4
S71	[103]	Conference	IEEE Xplore	2008	SP	screening	HBE	1	1	1	1	0	0	4
S72	[106]	Journal	sciencedirect	2014	SP+ER	diagnosis	Case study	0	1	1	1	0.5	2	5.5
S73	[107]	Journal	sciencedirect	2003	SP	diagnosis	HBE	1	1	1	0	0	2	5
S74	[108]	Journal	sciencedirect	2014	SP+ER	screening	HBE	1	1	1	0	0	1	4
S75	[115]	Journal	ACM	2014	SP+ER	screening	HBE	1	1	1	1	0	1	5
S76	[116]	Journal	sciencedirect	2014	SP+ER	monitoring	HBE	1	1	1	1	0	1	5
S77	[117]	Journal	sciencedirect	2014	SP+ER	diagnosis	Case study	0	1	1	1	0.5	2	5.5
S78	[118]	Journal	sciencedirect	2014	SP	screening	HBE	1	1	1	1	0	1	5
S79	[119]	Journal	IEEE Xplore	2014	SP+ER	screening	HBE	1	1	1	1	1	1.5	6.5
S80	[120]	Journal	sciencedirect	2014	SP+ER	monitoring	HBE	1	1	1	1	0	1	5
S81	[121]	Journal	DBLP	2014	SP	treatment	Case study	0	1	1	1	0.5	0	3.5
S82	[122]	Journal	sciencedirect	2014	SP+ER	diagnosis	HBE	1	1	1	1	0.5	2	6.5
S83	[123]	Conference	DBLP	2014	SP	management	Case study	1	1	0.5	0	0.5	0.5	3.5
S84	[124]	Journal	springerLink	2014	SP+ER	diagnosis	Case study	0	1	1	1	0.5	1	4.5
S85	[125]	Journal	sciencedirect	2013	SP+ER	diagnosis	HBE	1	1	1	1	0.5	2	6.5
S86	[126]	Journal	sciencedirect	2014	SP+ER	screening	HBE	1	1	1	0	1	1	5
S87	[133]	Conference	IEEEXplore	2015	SP+ER	screening	Case study	1	1	1	1	0	0.5	4.5
S88	[134]	Journal	springerLink	2015	SP+ER	diagnosis	Case study	1	1	1	0	1	2	6
S89	[135]	Journal	ACM	2015	SP+ER	screening	Case study	1	1	1	1	0	1.5	5.5
S90	[136]	Journal	sciencedirect	2015	SP+ER	screening	HBE	1	1	1	1	0	2	6
S91	[137]	Journal	springerLink	2015	SP+ER	diagnosis	HBE	1	1	1	1	0	1	5
S92	[138]	Conference	IEEEXplore	2015	SP+ER	diagnosis	HBE	1	1	1	1	0.5	0.5	5
S93	[139]	Journal	sciencedirect	2015	SP+ER	diagnosis	HBE	1	1	1	1	0	1	5
S94	[140]	Conference	IEEE Xplore	2015	SP+ER	screening	Case study	1	1	1	1	0	0.5	4.5
S95	[141]	Conference	IEEE Xplore	2015	SP+ER	monitoring	Case study	1	1	0	1	0.5	0.5	4
S96	[70]	Conference	IEEE Xplore	2015	SP+ER	diagnosis	HBE	1	1	1	1	0	0	4
S97	[71]	Conference	IEEE Xplore	2015	SP+ER	diagnosis	HBE	1	1	1	1	0	0	4
S98	[1]	Conference	IEEE Xplore	2015	SP+ER	diagnosis	HBE	1	1	1	1	0	0.5	4.5
S99	[2]	Journal	springerLink	2015	SP+ER	diagnosis	HBE	1	1	1	1	0.5	1	5.5
S100	[3]	Journal	springerLink	2015	SP+ER	diagnosis	Case study	1	1	1	1	1	0	5
S101	[142]	Journal	PubMED	2015	SP+ER	diagnosis	Case study	0	1	1	1	0.5	0	3.5
S102	[143]	Journal	PubMED	2015	SP+ER	diagnosis	HBE	1	1	0.5	1	0	1.5	5
S103	[144]	Journal	PubMED	2015	SP+ER	diagnosis	HBE	1	1	1	1	0	0	4
S104	[146]	Journal	PubMED	2013	SP+ER	management	Case study	0	1	1	1	0.5	0	3.5
S105	[147]	Journal	PubMED	2014	SP+ER	diagnosis	HBE	1	1	1	1	0	1	5
S106	[148]	Journal	PubMED	2013	SP+ER	diagnosis	Case study	0	1	1	1	0.5	2	5.5
S107	[149]	Journal	PubMED	2012	SP+ER	prognosis	Case study	0	0.5	0.5	1	0.5	1.5	4
S108	[150]	Journal	PubMED	2002	SP+ER	screening	Case study	0	1	1	1	0	1.5	4.5
S109	[151]	Journal	PubMED	2004	SP+ER	diagnosis	Case study	0	1	1	1	0	1.5	4.5
S110	[152]	Conference	PubMED	2014	SP+ER	screening	HBE	1	1	1	1	0.5	0.5	5
S111	[153]	Journal	PubMED	2013	SP+ER	screening	HBE	1	1	1	1	0	2	6
S112	[154]	Journal	PubMED	2013	SP+ER	diagnosis	HBE	1	1	1	1	0.5	2	6.5
S113	[155]	Journal	PubMED	2012	SP+ER	screening	HBE	1	1	1	1	0.5	1	5.5
S114	[156]	Conference	PubMED	2007	SP+ER	screening	HBE	1	1	1	1	0	0.5	4.5
S115	[157]	Conference	PubMED	2007	ER	diagnosis	HBE	1	1	1	1	0	0.5	4.5
S116	[158]	Journal	PubMED	2006	SP+ER	screening	Case study	0	1	1	1	0	1.5	4.5
S117	[159]	Journal	PubMED	2004	SP+ER	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S118	[160]	Journal	PubMED	2001	SP	screening	HBE	1	0.5	1	1	0	1.5	5
S119	[161]	Journal	PubMED	2014	SP+ER	diagnosis	Case study	0	1	1	1	0	1.5	4.5
S120	[162]	Journal	PubMED	2015	SP+ER	monitoring	Case study	0	1	1	1	0.5	2	5.5
S121	[163]	Journal	PubMED	2010	SP+ER	diagnosis	Case study	0	1	1	1	0	2	5
S122	[164]	Journal	Embase	2013	SP+ER	diagnosis	Case study	0	1	1	1	0	2	5
S123	[165]	Journal	Embase	2009	SP+ER	prognosis	Case study	0	1	1	1	0	1	4
S124	[166]	Journal	Embase	2005	SP+ER	diagnosis	Case study	0	1	1	1	0	2	5
S125	[167]	Journal	Embase	2007	SP+ER	screening	Case study	0	1	1	1	0	1	4
S126	[168]	Conference	Embase	2000	SP	screening	HBE	1	1	1	1	0	1.5	5.5
S127	[169]	Journal	Embase	2015	SP+ER	diagnosis	Case study	0	1	1	1	0.5	1	4.5
S128	[170]	Journal	Embase	2014	SP+ER	prognosis	HBE	1	1	1	1	0.5	1	5.5
S129	[171]	Journal	Embase	2013	SP+ER	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S130	[172]	Journal	Embase	2013	SP+ER	diagnosis	Case study	0	1	1	1	0.5	2	5.5
S131	[173]	Journal	Embase	2013	ER	prognosis	Case study	0	1	1	1	0.5	2	5.5
S132	[174]	Journal	Embase	2009	SP+ER	screening	HBE	1	1	1	1	0.5	1	5.5
S133	[175]	Journal	Embase	2008	SP+ER	diagnosis	HBE	1	1	1	1	0.5	2	6.5
S134	[176]	Journal	Embase	2008	SP+ER	diagnosis	HBE	1	1	1	1	0	1.5	5.5
S135	[177]	Journal	Embase	2007	SP+ER	monitoring	Case study	0	1	1	1	0	1	4
S136	[178]	Journal	Embase	2006	SP+ER	prognosis	HBE	1	1	1	1	0.5	2	6.5
S137	[179]	Journal	Embase	2015	SP+ER	diagnosis	HBE	1	1	1	1	0.5	1	5.5
S138	[180]	Journal	Embase	2011	SP+ER	diagnosis	HBE	1	1	1	1	0	2	6
S139	[181]	Conference	Embase	2010	SP+ER	screening	HBE	1	1	1	1	0	0.5	4.5
S140	[182]	Journal	Embase	2012	SP+ER	diagnosis	Case study	0	1	1	1	0	1.5	4.5
S141	[127]	Journal	Embase	2012	SP+ER	screening	Case study	0	1	1	1	0	1.5	4.5
S142	[13]	Journal	Embase	2011	SP+ER	diagnosis	HBE	1	1	1	1	0	1	5

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This research is part of the mPHR project in Morocco financed by the Ministry of Higher Education and Scientific Research in Morocco 2015–2018 (PPR1-15-17) and part of the project GINSENG (TIN2015-70259-C2-2-R) supported by the Spanish Ministry of Economy and Competitiveness and European FEDER funds.

References

Gaziano

Reddy

Paccaud

, et al. Chapter 33. Cardiovascular disease. In: Jamison

Breman

Measham

, et al. (eds) Disease control priorities in developing countries. 2nd ed. Washington, DC: World Bank, 2006.

Fayyad

Piatetsky-Shapiro

Smyth

From data mining to knowledge discovery in databases. AI Mag 1996; 17: 37–54.

Esfandiari

Babavalian

Moghadam

AME

, et al. Knowledge discovery in medicine: current issue and future trend. Expert Syst Appl 2014; 41: 4434–4463.

Romero

Ventura

Educational data mining: a review of the state of the art. IEEE T Syst Man Cy C 2010; 40: 601–618.

Romero

Ventura

Garcia

Data mining in course management systems: Moodle case study and tutorial. Comput Educ 2008; 51: 368–384.

Bellazzi

Zupan . Predictive data mining in clinical medicine: current issues and guidelines. Int J Med Inform 2008; 77: 81–97.

Kirkos

Spathis

Manolopoulos

Data mining techniques for the detection of fraudulent financial statements. Expert Syst Appl 2007; 32: 995–1003.

Ngai

EWT

Wong

, et al. The application of data mining techniques in financial fraud detection: a classification framework and an academic review of literature. Decis Support Syst 2011; 50: 559–569.

Jinyan

Qiang

Strong compound-risk factors: efficient discovery through emerging patterns and contrast sets. IEEE T Inf Technol B 2007; 11: 544–552.

10.

Kumari

Godara

Review of data mining classification models in cardiovascular disease diagnosis. Int J Comput Appl T 2011; 2: 304–307.

11.

Shouman

Turner

Stocker

Using data mining techniques in heart disease diagnosis and treatment. In: Proceedings of the Japan-Egypt conference on electronics, communications and computers, Alexandria, 6–9 March 2012, pp.173–177. New York: IEEE.

12.

Sivagowry

Durairaj

Persia

. An empirical study on applying data mining techniques for the analysis and prediction of heart disease. In: Proceedings of the international conference on information communication and embedded systems, Chennai, India, 21–22 February 2013, pp.265–270. New York: IEEE.

13.

Bharti

Singh

. Analytical study of heart disease prediction comparing with different algorithms. In: Proceedings of the international conference on computing, communication & automation, Noida, India, 15–16 May 2015, pp.78–82. New York: IEEE.

14.

Feng

Wang

Guo

, et al. Applications of data mining methods in the integrative medical studies of coronary heart disease: progress and prospect. Evid Based Complement Alternat Med 2014; 2014: 791841.

15.

Petersen

Feldt

Mujtaba

, et al. Systematic mapping studies in software engineering. In: Proceedings of the 12th international conference on evaluation and assessment in software engineering, Bari, 26–27 July 2008. New York: ACM.

16.

Brereton

Kitchenham

Budgen

, et al. Lessons from applying the systematic literature review process within the software engineering domain. J Syst Software 2007; 80: 571–583.

17.

Kitchenham

Mendes

Travassos

. A systematic review of cross vs. within company cost estimation studies. In: Proceedings of the conference empirical assessment in software engineering, UK, 10–11 April 2006, pp.89–98. New York: IEEE.

18.

Ouhbi

Idri

Alema

, et al. Requirements engineering education: a systematic mapping study. Requir Eng 2013; 20(2): 119–138.

19.

Idri

Amazala

Abran

Analogy-based software development effort estimation: a systematic mapping and review. Inform Software Tech 2015; 58: 206–230.

20.

Zapata

Fernández-Alemán

Idri

, et al. Empirical studies on usability of mHealth apps: a systematic literature review. J Med Syst 2015; 39(2): 1.

21.

The CORE conference ranking exercise—CORE portal, http://www.core.edu.au/conference-portal

22.

Journal citation reports, https://error.incites.thomsonreuters.com/error/Error?DestApp=IC2JCR&Alias=IC2&Domain=.thomsonreuters.com&Src=IP&Params=DestApp%3DIC2JCR&RouterURL=https%3A%2F%2Flogin.incites.thomsonreuters.com%2F&Error=IPError

23.

Bowyer

KW.

Mentoring advice on conferences versus journals for CSE faculty, 2012, https://pdfs.semanticscholar.org/ef9f/26a41ed66e39eaa6deb98dffb99312204be3.pdf

24.

Condori-Fernandez

Daneva

Sikkel

, et al. A systematic mapping study on empirical evaluation of software requirements specifications techniques. In: Proceedings of the 3rd international symposium on empirical software engineering and measurement, Lake Buena Vista, FL, 15–16 October 2009. New York: IEEE.

25.

Tonella

Torchiano

DuBois

, et al. Empirical studies in reverse engineering: state of the art and future trends. Empir Softw Eng 2007; 12(5): 551–571.

26.

Mark

Moody

GB.

MIT-BIH arrhythmia database, 1997, http://ecg.mit.edu/dbinfo.html (accessed 11 May 2016).

27.

Newman

Hettich

Blake

CLS

, et al. UCI repository of machine learning database. Irvine, CA: University of California, 1998.

28.

https://archive.ics.uci.edu/ml/datasets/SPECT+Heart

29.

http://archive.ics.uci.edu/ml/datasets/arrhythmia

30.

Karaolis

Moutiris

Papaconstantinou

, et al. Association rule analysis for the assessment of the risk of coronary heart events. In: Proceedings of the annual international conference of the IEEE Engineering in Medicine and Biology Society, Minneapolis, MN, 3–6 September 2009, pp.6238–624. New York: IEEE.

31.

Tsipouras

Exarchos

Fotiadis

, et al. Automated diagnosis of coronary artery disease based on data mining and fuzzy modeling. IEEE Trans Inf Technol Biomed 2008; 12(4): 447–458.

32.

Pecchia

Melillo

Sansone

, et al. Discrimination power of short-term heart rate variability measures for CHF assessment. IEEE Trans Inf Technol Biomed 2011; 15(1): 40–46.

33.

Soria

Martin

Caravaca

, et al. Survival prediction in patients undergoing ischemic cardiopathy. In: Proceedings of the international joint conference on neural networks, Atlanta, GA, 14–19 June 2009, pp.2979–2982. New York: IEEE.

34.

Dalakleidi

Zarkogianni

Karamanos

, et al. A hybrid genetic algorithm for the selection of the critical features for risk prediction of cardiovascular complications in Type 2 diabetes patients. In: Proceedings of the IEEE 13th international conference on bioinformatics and bioengineering, Chania, 10–13 November 2013, pp.1–4. New York: IEEE.

35.

Amin

Agarwal

Beg

. Genetic neural network based data mining in prediction of heart disease using risk factors. In: Proceedings of the IEEE conference on information & communication technologies, Thuckalay, India, 11–12 April 2013, pp. 1227–1231. New York: IEEE.

36.

Khairudin

Mohd

Hamid

. Predictive models of prolonged stay after Coronary Artery Bypass surgery. In: Proceedings of the international conference on statistics in science, business, and engineering, Langkawi, Malaysia, 10–12 September 2012, pp.1–5. New York: IEEE.

37.

Pal

Chakraborty

Mandana

. Data mining approach for coronary artery disease screening. In: Proceedings of the international conference on image information processing, Shimla, India, 3–5 November 2011, pp.1–6. New York: IEEE.

38.

Xing

Wang

Zhao

, et al. Combination data mining methods with new medical data to predicting outcome of coronary heart disease. In: Proceedings of the international conference on convergence information technology, Gyeongju, South Korea, 21–23 November 2007, pp.868–872. New York: IEEE.

39.

Candelieri

Conforti

Perticone

, et al. Early detection of decompensation conditions in heart failure patients by knowledge discovery: the HEARTFAID approaches. In: Proceedings of the computers in cardiology, Bologna, 14–7 September 2008, pp.893–896. New York: IEEE.

40.

Wang

Simon

Bonde

, et al. Decision tree for adjuvant right ventricular support in patients receiving a left ventricular assist device. J Heart Lung Transplant 2012; 31(2): 140–149.

41.

Melillo

Jovic

De Luca

, et al. Automatic classifier based on heart rate variability to identify fallers among hypertensive subjects. Healthc Technol Lett 2015; 2: 89–94.

42.

Hachesu

Ahmadi

Alizadeh

, et al. Use of data mining techniques to determine and predict length of stay of cardiac patients. Healthc Inform Res 2013; 19: 121–129.

43.

Guidi

Pettenati

Melillo

, et al. A machine learning system to improve heart failure patient assistance. IEEE J Biomed Health Inf 2014; 18: 1750–1756.

44.

RazvanAndonie

ID.

How to write a good paper in computer science and how will it be measured by ISI web of knowledge. Int J Comput V 2010; 5: 432–446.

45.

Babaei

Geranmayeh

Heart sound reproduction based on neural network classification of cardiac valve disorders using wavelet transforms of PCG signals. Comput Biol Med 2009; 39: 8–15.

46.

Cuesta-Frau

Biagetti

Quinteiro

, et al. Unsupervised classification of ventricular extrasystoles using bounded clustering algorithms and morphology matching. Med Biol Eng Comput 2007; 45: 229–239.

47.

Lee

Yoon

. Classification of ventricular arrhythmia using a support vector machine based on morphological features. In: Proceedings of the 35th annual international conference of the IEEE Engineering in Medicine and Biology Society, Osaka, Japan, 3–7 July 2013, pp.5785–5788. New York: IEEE.

48.

Podgorelec

Kokol

Stiglic

, et al. Knowledge discovery with classification rules in a cardiovascular dataset. Comput Methods Programs Biomed 2005; 80: S39–S49.

49.

Melillo

DeLuca

Bracale

, et al. Classification tree for risk assessment in patients suffering from congestive heart failure via long-term heart rate variability. IEEE J Biomed Health Inform 2013; 17(3): 727–733.

50.

Kurt

Ture

Kurum

AT.

Comparing performances of logistic regression, classification and regression tree, and neural networks for predicting coronary artery disease. Expert Syst Appl 2008; 34: 366–374.

51.

Muthukaruppan

MJ.

A hybrid particle swarm optimization based fuzzy expert system for the diagnosis of coronary artery disease. Expert Syst Appl 2012; 39(14): 11657–11665.

52.

Exarchos

Papaloukas

Fotiadis

, et al. An association rule mining-based methodology for automated detection of ischemic ECG beats. IEEE T Biomed Eng 2006; 53: 1531–1540.

53.

Kampouraki

Manis

Nikou

Heartbeat time series classification with support vector machines. IEEE Trans Inf Technol Biomed 2009; 13: 512–518.

54.

Carrión Señor

Fernández-Alemán

Toval

. Personal health records: new means to safely handle health data. IEEE Comput 2012; 45(11): 27–33.

55.

Carrión Señor

Fernández-Alemán

Toval

. Are personal health records safe? A review of free web-accessible personal health record privacy policies. J Med Internet Res 2012; 14(4): e114.

56.

Fernández-Alemán

Seva-Llor

Toval

, et al. Free web-based personal health records: an analysis of functionality. J Med Syst 2013; 37(6): 9990.

57.

Asl

Setarehdan

Mohebbi

Support vector machine-based arrhythmia classification using reduced features of heart rate variability signal. Artif Intell Med 2008; 44: 51–64.

58.

Sim

Gorman

Greenes

, et al. Clinical decision support systems for the practice of evidence-based medicine. J Am Med Inform Assoc 2001; 8(6): 527–534.

59.

Ravindranath

KR.

Clinical Decision Support System for heart diseases using Extended sub tree. In: Proceedings of the international conference on pervasive computing, Pune, India, 8–10 January 2015, pp.1–5. New York: IEEE.

60.

Bashir

Qamar

Khan

FH.

BagMOOV: a novel ensemble for heart disease prediction bootstrap aggregation with multi-objective optimized voting. Australas Phys Eng Sci Med 2015; 38(2): 305–323.

61.

Idri

Kadi

Evaluating a decision tree-based approach for cardiovascular dysautonomias diagnosis. SpringerPlus 2015; 5: 81.

62.

Kutlu

Kuntalp

A multi-stage automatic arrhythmia recognition and classification system. Comput Biol Med 2011; 41: 37–45.

63.

Dangare

Apte

SS.

A data mining approach for prediction of heart disease using neural networks. Int J Comput Eng Techn 2012; 3: 30–40.

64.

Liu

Tang

Zeng

, et al. Artificial neural network models for prediction of cardiovascular autonomic dysfunction in general Chinese population. BMC Med Inform Decis 2013; 13: 80.

65.

Karaolis

Moutiris

Hadjipanayi

, et al. Assessment of the risk factors of coronary heart events based on data mining with decision trees. IEEE T Inf Technol B 2010; 14(3): 559–566.

66.

Mitraa

Samanta

RK.

Cardiac arrhythmia classification using neural networks with selected features (also published in 1st international conference on computational intelligence: modeling techniques and applications). Proced Technol 2013; 10: 76–84.

67.

Acharya

Kumar

Bhat

, et al. Classification of cardiac abnormalities using heart rate signals. Med Biol Eng Comput 2004; 42: 288–293.

68.

Jabbar

Deekshatulu

Chandra

Classification of heart disease using K-nearest neighbor and genetic algorithm (also published in 1st international conference on computational intelligence: modeling techniques and applications). Proced Technol 2013; 10: 85–94.

69.

Anooj

PK.

Clinical decision support system: risk level prediction of heart disease using weighted fuzzy rules. Cent Eur J Comput Sci 2011; 1: 482–498.

70.

Fatima

Basharat

Khan

, et al. Biomedical (cardiac) data mining: extraction of significant patterns for predicting heart condition. In: Proceedings of the IEEE conference on computational intelligence in bioinformatics and computational biology, Honolulu, HI, 21–24 May 2014, pp.1–7. New York: IEEE.

71.

Son

Kim

, et al. Decision-making model for early diagnosis of congestive heart failure using rough set and decision tree approaches. J Biomed Inform 2012; 45: 999–1008.

72.

Jabbar

Chandra

Deekshatulu

BL.

Heart disease prediction system using associative classification and genetic algorithm. In: Proceedings of the international conference on emerging trends in electrical, electronics and communication technologies, Andhra Pradesh, India, 21–23 December 2012, vol. 1, pp.183–192. Elsevier.

73.

Dessai

ISF

. Intelligent heart disease prediction system using probabilistic neural network. Int J Adv Comput Theory Eng 2013; 2: 38–44.

74.

Bachler

Hörtenhuber

Mayer

, et al. Entropy-based data mining on the example of cardiac arrhythmia suppression. In: Proceedings of the international conference on brain informatics and health, Warsaw, 11–14 August 2014, pp.574–585. Springer.

75.

Mohebbi

Ghassemian

Prediction of paroxysmal atrial fibrillation based on non-linear analysis and spectrum and bispectrum features of the heart rate variability signal. Comput Meth Prog Bio 2012; 105: 40–49.

76.

Çomak

Arslan

Türkoğlu

A decision support system based on support vector machines for diagnosis of the heart valve diseases. Comput Biol Med 2007; 37: 21–27.

77.

Özbaya

Tezel

A new method for classification of ECG arrhythmias using neural network with adaptive activation function. Digit Signal Process 2010; 20(4): 1040–1049.

78.

Sufi

Khalil

Diagnosis of cardiovascular abnormalities from compressed ECG: a data mining-based approach. IEEE Trans Inf Technol Biomed 2011; 15(1): 33–39.

79.

Huang

Zhu

Sparse representation-based heartbeat classification using independent component analysis. J Med Syst 2012; 36: 1235–1247.

80.

Choi

Jiang

Kim

, et al. Cardiovascular abnormality detection method using cardiac sound characteristic waveform with data clustering technique. In: Proceedings of the international conference on control, automation and systems, Seoul, South Korea, 17–20 October 2007, pp.1596–1600. New York: IEEE.

81.

Rouhani

Soleymani

. Neural networks based diagnosis of heart arrhythmias using chaotic and nonlinear features of HRV signals. In: Proceedings of the international association of computer science and information technology, Singapore, 17–20 April 2009, pp.545–549. New York: IEEE.

82.

Liu

Hsu

Hung

. Development of evolutionary data mining algorithms and their applications to cardiac disease diagnosis. In: Proceedings of the IEEE congress on evolutionary computation, Brisbane, QLD, Australia, 10–15 June 2012, pp.1–8. New York: IEEE.

83.

Duan

. Association rule discovery with fuzzy decreasing support on Syndrome Differentiation and medication in coronary heart disease. In: Proceedings of the 3rd international conference on biomedical engineering and informatics, Yantai, China, 16–18 October 2010, pp.2297–2301. New York: IEEE.

84.

Karabulut

İbrikçi

Effective diagnosis of coronary artery disease using the rotation forest ensemble method. J Med Syst 2012; 36: 3011–3018.

85.

Lahsasna

Ainon

Zainuddin

, et al. Design of a fuzzy-based decision support system for coronary heart disease diagnosis. J Med Syst 2012; 36: 3293–3306.

86.

Khemphila

Boonjing

Heart disease classification using neural network and feature selection. In: Proceedings of the 21st international conference on systems engineering, Las Vegas, NV, 16–18 August 2011, pp.406–409. New York: IEEE.

87.

Das

Turkoglu

Sengur

Effective diagnosis of heart disease through neural networks ensembles. Expert Syst Appl 2009; 36: 7675–7680.

88.

Zadeh

Khazaee

Ranaee

Classification of the electrocardiogram signals using supervised classifiers and efficient features. Comput Meth Prog Bio 2010; 99(2): 179–194.

89.

Shen

Kao

Yang

, et al. Detection of cardiac arrhythmia in electrocardiograms using adaptive feature extraction and modified support vector machines. Expert Syst Appl 2012; 39: 7845–7852.

90.

Gharehchopogh

Khalifelu

. Neural network application in diagnosis of patient: a case study. In: Proceedings of the international conference on computer networks and information technology, Abbottabad, 11–13 July 2011, pp.245–249. New York: IEEE.

91.

Prati

Batista

GEAPA

Monard

MC.

A survey on graphical methods for classification predictive performance evaluation. IEEE T Knowl Data En 2011; 23: 1601–1618.

92.

Han

Kamber

Data mining, concepts and techniques. Burlington, MA: Morgan Kaufmann Publishers, 2001.

93.

Jabbar

Deekshatulu

Chandra

. Computational intelligence technique for early diagnosis of heart disease. In: Proceedings of the IEEE international conference on engineering and technology, Coimbatore, India, 20 March 2015, pp.1–6. New York: IEEE.

94.

Aljaaf

Al-Jumeily

Hussain

, et al. Predicting the likelihood of heart failure with a multi level risk assessment using decision tree. In: Proceedings of the 3rd international conference on technological advances in electrical, electronics and computer engineering, Beirut, Lebanon, 29 April–1 May 2015, pp.101–106. New York: IEEE.

95.

Joshi

Kaur

A review: comparative study of various clustering techniques in data mining. Int J Adv Res Comput Sci Softw Eng 2013; 3: 55–57.

96.

Jianxin

Yanwei

Guangcheng

, et al. A comparison of four data mining models: Bayes, neural network, SVM and decision trees in identifying syndromes in coronary heart disease. In: Proceedings of the 4th international symposium on neural networks, Nanjing, China, 3–7 June 2007, pp.1274–1279. New York: ACM.

97.

Shouman

Turner

Stocker

. Using decision tree for diagnosing heart disease patients. In: Proceedings of the ninth Australasian data mining conference (AusDM ‘11), Ballarat, VIC, Australia, 1–2 December 2011, vol. 121, pp.23–30. New York: ACM.

98.

Lee

Kim

Noh

, et al. Coronary artery disease prediction method using linear and nonlinear feature of heart rate variability in three recumbent postures. Inform Syst Front 2009; 11: 419–431.

99.

Nahar

Tasadduq

Tickle

, et al. Association rule mining to detect factors which contribute to heart disease in males and females. Expert Syst Appl 2013; 40: 1086–1093.

100.

Babaoglu

Baykan

Aygul

, et al. Assessment of exercise stress testing with artificial neural network in determining coronary artery disease and predicting lesion localization. Expert Syst Appl 2009; 36: 2562–2566.

101.

Martis

Acharya

Mandana

, et al. Application of principal component analysis to ECG signals for automated diagnosis of cardiac health. Expert Syst Appl 2012; 39(14): 11792–11800.

102.

Gueguin

Roux

Hernandez

, et al. Exploring time series retrieved from cardiac implantable devices for optimizing patient follow-up. IEEE T Biomed Eng 2008; 55: 2343–2352.

103.

Babaoglu

Findik

Ülker

A comparison of feature selection models utilizing binary particle swarm optimization and genetic algorithm in determining coronary artery disease using support vector machine. Expert Syst Appl 2010; 37: 3177–3183.

104.

Lee

Noh

Ryu

. A data mining approach for coronary heart disease prediction using HRV features and carotid arterial wall thickness. In: Proceedings of the international conference on biomedical engineering and informatics (BMEI 2008), Sanya, China, 27–30 May 2008, pp.200–206. New York: IEEE.

105.

Maglogiannis

Loukis

Zafiropoulos

, et al. Support Vectors Machine-based identification of heart valve diseases using heart sounds. Comput Methods Programs Biomed 2009; 95: 47–61.

106.

Polat

Şahan

Güneş

Automatic detection of heart disease using an artificial immune recognition system (AIRS) with fuzzy resource allocation mechanism and k-nn (nearest neighbour) based weighting preprocessing. Expert Syst Appl 2007; 32: 625–631.

107.

Soni

Ansari

Sharma

, et al. Intelligent and effective heart disease prediction system using weighted associative classifiers. Int J Comput Sci Eng 2011; 3: 2385–2392.

108.

Manikantan

Latha

Predicting the analysis of heart disease symptoms using medicinal data mining methods. Int J Adv Comput Theory Eng 2013; 2: 2319–2526.

109.

Alizadehsani

Habibi

Hosseini

, et al. A data mining approach for diagnosis of coronary artery disease. Comput Methods Programs Biomed 2013; 111: 52–61.

110.

Palaniappan

Awang

Intelligent heart disease prediction system using data mining techniques. In: Proceedings of the IEEE/ACS international conference on computer systems and applications, Doha, Qatar, 31 March–4 April 2008, pp.108–115. New York: IEEE.

111.

Bourdès

Ferrières

Amar

, et al. Prediction of persistence of combined evidence-based cardiovascular medications in patients with acute coronary syndrome after hospital discharge using neural networks. Med Biol Eng Comput 2011; 49: 947–955.

112.

Ordonez

Association rule discovery with the train and test approach for heart disease prediction. IEEE Trans Inf Technol Biomed 2006; 10(2): 334–343.

113.

Maragoudakis

Loukis

Heart sound screening in real-time assistive environments through MCMC Bayesian data mining. Univers Access Inform Soc 2014; 13: 73–88.

114.

Dasa

Turkoglu

Sengur

Diagnosis of valvular heart disease through neural networks ensembles. Comput Methods Programs Biomed 2009; 93: 185–191.

115.

Sudha

Gayathri

Jaisankar

Effective analysis and predictive model of stroke disease using classification methods. Int J Comput Appl 2012; 43(14): 26–31.

116.

Pecchia

Melillo

Bracale

Remote health monitoring of heart failure with data mining via CART method on HRV features. IEEE Trans Biomed Eng 2011; 58(3): 800–804.

117.

Liu

Wang

, et al. Modelling of inquiry diagnosis for coronary heart disease in traditional Chinese medicine by using multi-label learning. BMC Complem Altern M 2010; 10: 37.

118.

Setiawan

Venkatachalam

Hani

AFM

. Missing attribute value prediction based on artificial neural network and rough set theory. In: Proceedings of the international conference on biomedical engineering and informatics, Sanya, China, 27–30 May 2008, pp.306–310. New York: IEEE.

119.

Kotsiantis

SB.

Decision trees: a recent overview. Artif Intell Rev 2013; 39: 261–283.

120.

Shao

Hou

Chiu

CC.

Hybrid intelligent modeling schemes for heart disease classification. Appl Soft Comput 2014; 14: 47–52.

121.

Acharya

Bhat

Iyengar

, et al. Classification of heart rate data using artificial neural network and fuzzy equivalence relation. Pattern Recogn 2003; 36: 61–68.

122.

Vafaie

Ataei

Koofigar

HR.

Heart diseases prediction based on ECG signals classification using a genetic-fuzzy system and dynamical model of ECG signals. Biomed Signal Proces 2014; 14: 291–296.

123.

Zhang

Patuwo

MY.

Forecasting with artificial neural networks: the state of the art. Int J Forecasting 1998; 14: 35–62.

124.

Zhang

Dong

Luo

, et al. Heartbeat classification using disease-specific feature selection. Comput Biol Med 2014; 46: 79–89.

125.

Park

Kang

PcHD: personalized classification of heartbeat types using a decision tree. Comput Biol Med 2014; 54: 79–88.

126.

Sanz

Galar

Jurio

, et al. Medical diagnosis of cardiovascular diseases using an interval-valued fuzzy rule-based classification system. Appl Soft Comput 2014; 20: 103–111.

127.

Saini

Singh

Khosla

Electrocardiogram beat classification using empirical mode decomposition and multiclass directed acyclic graph support vector machine. Comput Electr Eng 2014; 40: 1774–1787.

128.

Qiao

Rajagopalan

Clifford

GD.

Ventricular fibrillation and tachycardia classification using a machine learning approach. IEEE T Biomed Eng 2014; 61(6): 1607–1613.

129.

Chen

Cheng

Liao

, et al. Hybrid classification engine for cardiac arrhythmia cloud service in elderly healthcare management. J Visual Lang Comput 2014; 25: 745–753.

130.

Yuan

Chen

Liu

Association rules mining on heart failure differential treatment based on the improved firefly algorithm. J Comput 2014; 9: 822–829.

131.

Zhu

Ding

Hao

Multiclass maximum margin clustering via immune evolutionary algorithm for automatic diagnosis of electrocardiogram arrhythmias. Appl Math Comput 2014; 227: 428–436.

132.

Roy

Chin

. Prediction and management of readmission risk for congestive heart failure. In: Proceedings of the international conference on health informatics (HEALTHINF 2014), Angers, 3–6 March 2014.

133.

Kim

Lee

Park

, et al. Adaptive mining prediction model for content recommendation to coronary heart disease patients. Cluster Comput 2014; 17: 881–891.

134.

Acharya

Sree

Krishnan

MMR

, et al. Automated classification of patients with coronary artery disease using grayscale features from left ventricle echocardiographic images. Comput Methods Programs Biomed 2013; 112: 624–632.

135.

Zuo

Ouyang

, et al. Identifying functional modules for coronary artery disease by a prior knowledge-based approach. Gene 2014; 537: 260–268.

136.

Purwanto Eswaran

Logeswaran

Rahman

ARA

. Prediction models for early risk detection of cardiovascular event. J Med Syst 2012; 36: 521–531.

137.

Tavares

Paredes

Rocha

, et al. Expert knowledge integration in the data mining process with application to cardiovascular risk assessment. In: Proceedings of the 37th annual international conference of the IEEE Engineering in Medicine and Biology Society, Milan, 25–29 August 2015, pp.2538–2542. New York: IEEE.

138.

Bandyopadhyay

Wolfson

Vock

, et al. Data mining for censored time-to-event data: a Bayesian network model for predicting cardiovascular risk from electronic health record data. Data Min Knowl Disc 2015; 29(4): 1033–1069.

139.

Kotu

Engan

Borhani

, et al. Cardiac magnetic resonance image-based classification of the risk of arrhythmias in post-myocardial infarction patients. Artif Intell Med 2015; 64(3): 205–215.

140.

Sahin

Subas

Classification of the cardiotocogram data for anticipation of fetal risks using machine learning techniques. Appl Soft Comput 2015; 33: 231–238.

141.

Alickovic

Subasi

Effect of multiscale PCA De-noising in ECG beat classification for diagnosis of cardiovascular diseases. Circ Syst Signal Pr 2015; 34(2): 513–533.

142.

Shenzhen

Qin

, et al. A restricted Boltzmann machine based two-lead electrocardiography classification. In: Proceedings of the IEEE 12th international conference on wearable and implantable body sensor networks, Cambridge, MA, 9–12 June 2015, pp.1–9. New York: IEEE.

143.

Nazarahari

Namin

Davaie-Markazi

, et al. A multi-wavelet optimization approach using similarity measures for electrocardiogram signal classification. Biomed Signal Proces 2015; 20: 142–151.

144.

Mendes

Paredes

Rocha

, et al. Assessment of cardiovascular risk based on a data-driven knowledge discovery approach. In: Proceedings of the annual international conference of the IEEE Engineering in Medicine and Biology Society, Milan, 25–29 August 2015, pp.6800–6803. New York: IEEE.

145.

Dobbins

Rawassizadeh

. Clustering of physical activities for quantified self and mHealth applications. In: Proceedings of the IEEE international conference on computer and information technology; ubiquitous computing and communications; dependable, autonomic and secure computing; pervasive intelligence and computing, Liverpool, 26–28 October 2015, pp.1423–1428. New York: IEEE.

146.

Alshraideh

Otoom

Al-Araida

, et al. A web based cardiovascular disease detection system. J Med Syst 2015; 39: 122.

147.

Marateb

Goudarzi

A noninvasive method for coronary artery diseases diagnosis using a clinically-interpretable fuzzy rule-based system. J Res Med Sci 2015; 20: 214–223.

148.

Ghadiri Hedeshi

Saniee Abadeh

. Coronary artery disease detection using a fuzzy-boosting PSO approach. Comput Intell Neurosci 2014; 2014: 783734.

149.

Austin

, et al. Using methods from the data-mining and machine-learning literature for disease classification and prediction: a case study examining classification of heart failure subtypes. J Clin Epidemiol 2013; 66: 398–407.

150.

Austin

Lee

Steyerberg

, et al. Regression trees for predicting mortality in patients with cardiovascular disease: what improvement is achieved by using ensemble-based methods? Biom J 2012; 54: 657–673.

151.

Sacha

Goodenday

Cios

KJ.

Bayesian learning for cardiac SPECT image interpretation. Artif Intell Med 2002; 26: 109–143.

152.

Goletsis

Papaloukas

Fotiadis

, et al. Automated ischemic beat classification using genetic algorithms and multicriteria decision analysis. IEEE T Biomed Eng 2004; 51: 1717–1725.

153.

Balouchestani

Krishnan

. Fast clustering algorithm for large ECG data sets based on CS theory in combination with PCA and K-NN methods. In: Proceedings of the IEEE Engineering in Medicine and Biology Society, Chicago, IL, 26–30 August 2014, pp.98–101. New York: IEEE.

154.

Abawajy

Kelarev

Chowdhury

Multistage approach for clustering and classification of ECG data. Comput Methods Programs Biomed 2013; 112: 720–730.

155.

De Carvalho

Moreno

Pimenta

, et al. A heart disease recognition embedded system with fuzzy cluster algorithm. Comput Methods Programs Biomed 2013; 110: 447–454.

156.

Huang

Lin

Hsu

, et al. Using n-gram analysis to cluster heartbeat signals. BMC Med Inform Decis 2012; 12: 64.

157.

Chudácek

Petrik

Georgoulas

, et al. Comparison of seven approaches for holter ECG clustering and classification. In: Proceedings of the IEEE Engineering in Medicine and Biology Society, Lyon, 22–26 August 2007, pp.3844–3847. New York: IEEE.

158.

Bursa

Lhotska

Modified ant colony clustering method in long-term electrocardiogram processing. In: Proceedings of the IEEE Engineering in Medicine and Biology Society, Lyon, 22–26 August 2007, pp.3249–3252. New York: IEEE.

159.

Georgoulas

Stylios

Groumpos

PP.

Predicting the risk of metabolic acidosis for newborns based on fetal heart rate signal classification using support vector machines. IEEE T Biomed Eng 2006; 53: 875–884.

160.

Osowski

Hoai

Markiewicz

Support vector machine-based expert system for reliable heartbeat recognition. IEEE T Biomed Eng 2004; 51: 582–589.

161.

Osowski

Linh

TH.

ECG beat recognition using fuzzy hybrid neural network. IEEE T Biomed Eng 2001; 48: 1265–1271.

162.

Dai

Brisimi

Adams

, et al. Prediction of hospitalization due to heart diseases by supervised learning methods. Int J Med Inform 2015; 84: 189–197.

163.

Gregori

Bigi

Cortigiani

, et al. Non-invasive risk stratification of coronary artery disease: an evaluation of some commonly used statistical classifiers in terms of predictive accuracy and clinical usefulness. J Eval Clin Pract 2009; 15: 777–781.

164.

Chen

Leong

, et al. Predicting coronary artery disease with medical profile and gene polymorphisms data. Stud Health Technol Inform 2007; 129: 1219–1224.

165.

Colombet

Ruelland

Chatellier

, et al. Models to predict cardiovascular risk: comparison of CART, multilayer perceptron and logistic regression. AMIA Symp 2000; 1: 156–160.

166.

Gong

, et al. Combining multiple ECG features does not improve prediction of defibrillation outcome compared to single features in a large population of out-of-hospital cardiac arrests. Crit Care 2015; 19: 425.

167.

Murukesan

Murugappan

Iqbal

, et al. Machine learning approach for sudden cardiac arrest prediction based on optimal heart rate variability features. J Med Imaging Health Inf 2014; 4: 521–532.

168.

Wang

Chiang

Hsu

, et al. ECG arrhythmia classification using a probabilistic neural network with a feature reduction method. Neurocomputing 2013; 116: 38–45.

169.

Tang

Liu

Zeng

, et al. Comparison of prediction model for cardiovascular autonomic dysfunction using artificial neural network and logistic regression analysis. PLoS ONE 2013; 8: e70571.

170.

Arena

Myers

De Souza

, et al. A neural network approach to predicting outcomes in heart failure using cardiopulmonary exercise testing. J Am Coll Cardiol 2013; 61: 10.

171.

Kim

Shin

, et al. Robust algorithm for arrhythmia classification in ECG using extreme learning machine. Biomed Eng Online 2009; 8: 31.

172.

Melgani

Bazi

Classification of electrocardiogram signals with support vector machines and particle swarm optimization. IEEE Trans Inf Technol Biomed 2008; 12: 667–677.

173.

Chesnokov

YV.

Complexity and spectral analysis of the heart rate variability dynamics for distant prediction of paroxysmal atrial fibrillation with artificial intelligence methods. Artif Intell Med 2008; 43: 151–165.

174.

Cevenini

Barbini

Scolletta

, et al. A comparative analysis of predictive models of morbidity in intensive care unit after cardiac surgery—part II: an illustrative example. BMC Med Inform Decis Mak 2007; 7: 35.

175.

Nilsson

Ohlsson

Thulin

, et al. Risk factor identification and mortality prediction in cardiac surgery using artificial neural networks. J Thorac Cardiovasc Surg 2006; 132: 12–19.

176.

Moridani

Setarehdan

Nasrabadi

, et al. New algorithm of mortality risk prediction for cardiovascular patients admitted in intensive care unit. Int J Clin Exp Med 2015; 8: 8916–8926.

177.

Zadeh

Khazaee

High efficient system for automatic classification of the electrocardiogram beats. Ann Biomed Eng 2011; 39: 996–1011.

178.

Coimbra

Vijaya Kumar

. Arrhythmia detection and classification using morphological and dynamic features of ECG signals. In: Proceedings of the IEEE Engineering in Medicine and Biology Society, Buenos Aires, Argentina, 31 August–4 September 2010, pp.1918–1921. New York: IEEE.

179.

Mandal

Sairam

Accurate prediction of coronary artery disease using reliable diagnosis system. J Med Syst 2012; 36: 3353–3373.

Systematic mapping study of data mining–based empirical studies in cardiology

Abstract

Keywords

Introduction

Related work

Research methodology

RQs

Search strategy

Study selection

Quality assessment

Data extraction strategy and synthesis

Threats to validity

Results

Overview of the selected studies

RQ1: In which sources and in which years were selected studies on the application of DM techniques in cardiology published?

RQ2: What types of investigations are carried out and published in cardiac DM studies?

RQ3: What are the disciplines of cardiology in which DM techniques were applied?

RQ4: What types of empirical studies were conducted when applying DM techniques in cardiology?

Discussion

RQ1: In which sources and in which years were the selected studies on the application of DM techniques in cardiology published?

What types of investigations are carried out and published in cardiac DM studies?

What are the disciplines of cardiology where DM techniques were applied?

What types of empirical studies were conducted when applying DM techniques in cardiology?

Implications for researchers

Conclusion

Footnotes

Appendix A. Classification Results

Declaration of conflicting interests

Funding

References