Traditional diagnostic methods may miss subtle yet critical signs, leading to delayed or inaccurate diagnoses. As the global community grapples with these alarming statistics, there is an urgent need for innovative solutions to address the challenges in CVD diagnosis and management⁴.

Over the past few years, the healthcare industry has witnessed substantial technological advancement, especially in the domains of Artificial Intelligence (AI) and Machine Learning (ML). AI is expanding its footprint in medical sciences and offering algorithmic solutions to clinical professionals in their practice, assisting in diagnosis, prognosis, and therapeutic strategies⁵. The recent development of AI and ML has created transformative prospects in research, especially in the realm of CVD. AI technologies such as ML, Robotics, Imaging, and Natural Language Processing (NLP) have now been integrated in cardiovascular medicine⁶.

Role of AI and ML in Cardiovascular Medicine

• AI and ML in Cardiovascular Imaging

• Predictive Analytics in Cardiology

The capabilities of AI and ML go beyond imaging. Today, there is an abundance of data accessibility due to the growth of wearable health devices and the digitalization of health records. ML algorithms can filter through this data, detecting patterns and forecasting results. For example, ML models may estimate the chance of a patient having a heart attack in the following year by zanalyzing their electronic health record, allowing for early treatment⁸.

• AI-driven Decision Support Systems

With its multifaceted conditions and treatments, cardiology often presents doctors with difficult decision-making circumstances. AI-powered decision support systems can help physicians by making real-time suggestions based on patient data. For example, when a patient complains of chest discomfort, these systems may analyze the clinical data, imaging, and lab findings and provide probable diagnosis and therapy courses⁹.

• Drug Discovery and Development

AI algorithms can analyze vast datasets to identify potential drug candidates for cardiovascular diseases. They can predict how different compounds can affect human biology and speed up the drug discovery process, which traditionally takes years and significant resources¹⁰.

• Risk Stratification

AI can analyze electronic health records, genetic data, and other relevant information to stratify patients based on their risk of developing cardiovascular diseases. High-risk individuals can then be targeted for preventive measures, screenings, or more frequent check-ups¹¹.

• Virtual Health Assistants

AI-driven virtual assistants can provide patients with information about their conditions, answer questions, and even remind them to take medications or perform exercises. These assistants can act as an extension of the healthcare team, ensuring that patients are well-informed and adhering to their treatment plans. This highly advanced technology is continuously evolving rapidly, and it is conceivable that there have been even more breakthroughs in this realm^12,13.

• Other applications

In cardiovascular medicine, other notable uses encompass AI-driven personalized treatment strategies designed to cater to each patient's unique needs, guaranteeing optimal therapeutic results. Cardiac rehabilitation programs, enhanced by AI, deliver tailored exercise and therapy plans, ensuring the best recovery results. The realm of rehabilitation and aftercare is transformed by AI's precision in assisting patients throughout their recuperation. Moreover, incorporating AI into remote monitoring tools offers uninterrupted supervision, swiftly identifying irregularities and assuring prompt patient care. Together, these innovations signify a shift towards a more personalized and forward-thinking approach in cardiac care¹⁴.

The present article uses bibliometric tools to assess the status, trends, and frontiers of research activities on the use of AI-ML in diagnosing and managing CVD, such as most referenced publications, countries, journals, authors, and funding agencies involved in AI-ML CVD research. Moreover, it is a valuable resource for clinicians, researchers, and stakeholders seeking guidance and a deeper understanding of the current AI-ML landscape in CVD research.

Methods

Search Protocol

The search algorithm gathered data from PubMed (MEDLINE), which has over 33 million citations ranging from biomedical literature and life sciences. PubMed stands as a premier repository for biomedical literature. PubMed contains citations and abstracts from peer-reviewed journals, scholarly journals, books, and conference proceedings.

Table 1: PubMed Search query terms

Inclusion Criteria

We researched peer-reviewed articles focusing on the role and use of AI in CVD research. Literature published from 1/1/2013 to 15/6/2023 was included in the search. The data was acquired on 1/7/2023 from the PubMed database. This study's timeframe was between 2013 and 2023 due to the significant advancements and rapid progression of AI and ML applications, particularly in the healthcare sector.

Exclusion Criteria

Grey literature, preprints, book chapters and novels were excluded from the review. Additionally, publications not written in English were also not considered.

Data Extraction

All data was sourced from PubMed in .txt (text) format, encompassing 11 fields and containing details like 1) author, 2) title, 3) journal name, 4) publication year, and 5) citation. This data was then inputted into a Microsoft Excel spreadsheet for error verification. Subsequently, the consolidated dataset was imported into VosViewer for bibliometric analysis.

Visualization and Bibliometric Indicators

The VOSviewer (version 1.6.19 developed by the Center for Science and Technology, Leiden University, the Netherlands) was used to analyze and visualize author keywords. VOSviewer is a software application for creating and visualizing bibliometric networks. These networks can be built via patterns like keywords, co-occurrence, co-citation, bibliographic coupling, or co-authorship. This tool is popular for mapping scientific research and helping users understand how different studies or topics are connected within a specific field¹⁵.

The retrieved records were thoroughly examined, and the following bibliometric indicators were obtained: publications growth rate, distribution (country, publications, documents citation, funding sources and co-occurrence of authors keywords).

Total number of publications and Growth Rate

The data highlights increasing publication growth in the realm of topics from 2013 to 2023, with an average annual increase rate of 32.6%. Specifically, 2022 saw a peak in publications with a total of 262 articles, constituting 29.2% of the total articles. The preceding year, 2021, contributed 181 articles, contributing to 20.2% publications. Meanwhile, 2020 witnessed the publication of 164 articles, making up 16.3% of the total count. As of the midpoint of the current year, up to 15/6/23, there have been 121 recorded publications, representing 13.5% of the overall dataset.

Figure 1. Distribution of Publications between 2013-2023 (till 15/6/2023).

Distribution by countries

Figure 2 provides insights into the global landscape of AI -ML research in CVD, identifying regions that contributed to the field and potentially uncovering areas where further research collaboration or exploration is needed. The data indicates that the United States of America (USA) has the highest number of publications, with a total of 213 articles. China came second with 183 articles, while the United Kingdom produced 112 publications. Germany had 66 publications, while South Korea had 54. Canada, Italy, Japan, and India also made considerable contributions, with 50, 45, 41, and 32 articles, respectively. Most of the nations on the list are highly developed economies. The varying shades on the QGIS map represent the number of publications from each country, allowing for a visual and spatial understanding of the distribution of scholarly articles across different regions.

Figure 2. Choropleth map depicting the distribution of global publications by country between 2013 and 2023 as recorded by PubMed (https://pubmed.ncbi.nlm.nih.gov/) until June 30. The map was created using QGIS 3.30.1, with the base layer map sourced from ArcGIS Hub

Distribution by academic journals

Figure 3 presents the top journals that have published research on AI-ML in CVD studies. During this time, 332 different journals have published 895 articles on the topic. The Journal of Scientific Reports (Sci Rep) contributed the largest, with 37 out of 895 articles. Following closely were Sensors (Basel) with 33 articles, PLoS One with 32, Stroke with 25, and the Annual International Conference of the IEEE Engineering in Medicine and Biology Society (Annu Int Conf IEEE Eng Med Biol Soc) with 23 articles.

Figure 3. Publications distribution by journals.

Distribution by citations

Among the top listed articles, the top 5 articles were ranked according to the number of times they had been cited by Web of Science. The top-cited article was titled "Artificial Intelligence in Precision Cardiovascular Medicine" by Krittanawong et al. (2017), with a total citation count of 394. The article "Automatic Detection of Cerebral Microbleeds from MR Images via 3D Convolutional Neural Networks" by Dou et al. (2016), with 388 citations, was second on the list. This was followed by "Machine learning in cardiovascular medicine: are we there yet?" by Shameer et al. (2018), which gathered 211 citations, addressing the progress and challenges of machine learning in cardiovascular medicine. The articles "Classification of myocardial infarction with multi-lead ECG signals and deep CNN" by Baloglu et al. (2019) and "Clinical applications of ML in CVD and its relevance to cardiac imaging" by Al'Aref et al. (2019) have also received notable attention, with 207 and 204 citations respectively. These highly cited articles reflect the significance of AI and ML in diagnosing and managing CVD.

Table 2: Top 5 Highly Cited articles.

Funding sponsors

Analysis of keywords

Figure 5. displays the co-occurrence of authors' keywords using the full counting method, where each instance of a keyword in a document is tallied. Out of a total of 1520 keywords, the network visualization was generated by selecting keywords that appeared at least 5 times, resulting in 77 keywords meeting this threshold. These keywords are then grouped into clusters and presented in the network visualization diagram. The selection process focuses on keywords with the highest total link strength. In the figure, coloured and differently-sized circle fields are interconnected at varying distances. The colours signify groupings or clusters, enabling the identification of topics with closely related aspects.

Figure 4: Top funding institutions and the corresponding number of publications.

The circle's size corresponds to the frequency of a keyword's usage in the documents; the more frequent the use, the larger the circle. The distances between keywords in the diagram represent their connections, with shorter distances indicating stronger associations. The lines between terms symbolize their interconnectedness in the studies. The figure highlights that some of the most frequently utilized keywords include "Machine learning" (occurrence=244), "stroke" (occurrence=151), "Artificial Intelligence" (occurrence=105), and "Deep learning" (occurrence=99).

Figure 5. Co-occurrence analysis of authors' keywords

Although PubMed offers a robust platform for bibliometric analysis, incorporating other databases like Web of Science (WOS), Scopus, or IEEE in future research could provide a more extensive collection of papers, enriching the study's depth. Articles included in our study were restricted to peer-reviewed and English-language publications. Future research should consider broadening the search criteria to include diverse sources and languages for a more holistic understanding of the subject.