VOSviewer software was utilized to create the national collaboration network diagram (Figure 2A), while Bibliometrix software was employed to generate the country production diagram showcasing the top three countries in terms of publication volume (Figure 2B). In Figure 2A, countries are represented by circles, with larger circles indicating a higher publication volume. The connecting lines signify collaborations between countries, and distinct clusters are represented by different colors. The analysis of national collaboration reveals the emergence of clusters in the fields of ML and MI, with China and the United States as prominent representatives. China primarily collaborates with Asian and European countries such as Japan, South Korea, the United Kingdom, and Germany, whereas the United States mainly collaborates with Asian and European countries, including the United Kingdom, Germany, and India. Analysis of spatiotemporal development indicates that before 2015, there was a comparable level of research output among countries in MI and ML research. However, after 2015, both China and the United States experienced rapid development (a significant increase in research output and quality over a short period) in this field, starting from 2022, China’s research in this area will surpass that of the United States (Figure 2B).

VOSviewer software was used to create the institutional collaboration network diagram (Figure 2C). In this figure, institutions are represented by circles, with larger circles indicating a higher number of publications. Different colors represent distinct clusters, while the lines connecting the circles signify collaboration between institutions. Analysis of Figure 2C reveals that in the fields of MI and ML, a cluster of institutional collaborations has formed, represented by prominent institutions such as Harvard Medical School in the United States and Capital Medical University in China. In particular, there is relatively little collaboration observed between domestic and foreign institutions.

The leading position of the United States and China can be attributed to their large scientific populations, substantial research funding, and supportive policies for healthcare innovation. To gain a more comprehensive understanding, future analyses should delve into the specific research outputs, funding allocations, and policy frameworks that have contributed to the emergence of these countries as leaders in ML and MI research.

VOSviewer software was employed to generate an author collaboration network diagram (Figure 3A). Within this figure, distinct authors and their respective teams are represented by circles of varying colors, while the connecting lines depict their collaborative ties. The analysis underscores the emergence of a prominent cooperative team in the MI and ML fields, led by Dey, Damini, and Berman, Daniel S. from the United States. This team has developed and validated an advanced deep learning system capable of swiftly quantifying plaque volume and stenosis severity through CCTA. Accuracy of this system closely aligns with expert readings and intravascular ultrasound results, suggesting potential prognostic value for future myocardial infarction cases (17). This was a comprehensive multicenter study, with the research paper boasting an impact factor of 30.8 in 2023. Conversely, within the domestic landscape, collaborative author teams have yet to materialize in significant numbers. One standout figure is Chin-Sheng Lin, who has garnered attention with his four publications. Lin has developed an innovative deep learning model (DLM) that predicts biological age based on electrocardiogram data, thereby shedding light on its potential implications for future cardiovascular disease (CVD) risk assessment (18). In another study, a DLM (deep learning model) grounded on 12-lead electrocardiograms was presented as an innovative diagnostic support tool, offering a fresh perspective on clinical diagnosis (19). In 2023, the impact factors associated with these publications varied significantly, with one recording a score of 3.6, while the other had an impact factor of 0.

VOSviewer software was utilized to generate an analysis chart showcasing cited authors (Figure 3B). In this chart, distinct clusters are represented by different colors, while the size of the circles corresponds to the number of citations. In particular, THYGESEN, K, and ACHARYA, UR exhibit the highest citation counts. ACHARYA, UR, et al., published an extensive review that delved into artificial intelligence techniques, specifically leveraging ECG and other biophysical signals for MI diagnosis. This review also highlighted the application of deep convolutional neural networks (DCNN) in MI diagnosis (11). Thygesen, K.’s seminal work, titled “Fourth Universal Definition of Myocardial Infarction (2018),” was published in the prestigious Journal of the American College of Cardiology in 2018. By 2023, this article had acquired an impact factor of 24, attesting to its significance in the field. It represents an expert consensus on myocardial infarction, encompassing contributions from esteemed scholars across various nations, including the United States, Denmark, and New Zealand (17).

Using CiteSpace 6.3.R1 to create an overlay of the Periodicals double-map superposition of journals (Figure 4A) allows for an understanding of research directions. This overlay presents the cited journals on the left and the citing journals on the right. Journals in the fields of Medicine, Medical Sciences, and Clinical Science have been cited by journals in Molecular Biology, Genetics (Z = 4.2199888), Health, Nursing, and Medical Science (Z = 7.7724767).

VOSviewer software was utilized to generate a co-occurrence map of cited journals as depicted in Figure 4B. In this figure, journals are represented by circles, with the size of each circle indicating the number of citations received by the corresponding journal. The connecting lines between the circles represent the frequency of collaborations among the journals, while different colors denote distinct clusters. In particular, the journal with the highest number of citations is Circulation from the United States, which received 1790 citations. It is followed by the Journal of the American College of Cardiology, also from the United States, with 1,249 citations.

VOSviewer software was utilized to generate a journal co-occurrence map as presented in Figure 4C. The analysis conducted revealed that the journal with the highest number of publications is Frontiers in Cardiovascular Medicine from Switzerland. This journal boasted an impact factor of 3.1 in 2023, with a total of 49 publications and 249 citations. On the other hand, the journal with the highest number of citations was Computers in Biology and Medicine from the United States, which garnered 538 citations, 12 publications, and had a latest impact factor of 6.7 in 2023. A comprehensive analysis underscores the prominent position occupied by Computers in Biology and Medicine from the United States in this study.

Keywords in bibliometrics are the embodiment of research hot spots (19). The keyword clustering diagram generated by VOSviewer (Figure 5A) reveals three primary clusters. The first cluster revolves around “myocardial infarction,” with associated keywords such as cells, biomarkers, cardiovascular disease, inflammation, acute myocardial infarction, and risk. These keywords indicate that this cluster primarily focuses on molecular biology research. The second cluster centers on “machine learning,” with related keywords including impact, mortality, prediction, risk factors, hospital mortality, acute coronary syndrome, and in-hospital mortality. This suggests that the research within this cluster primarily adopts a clinical approach. The third cluster is centered on “Deep Learning (DL),” with associated keywords being myocardium, diagnosis, feature extraction, classification, ECG signals, electrocardiogram, and localization. This illustrates that DL research in the context of myocardial infarction primarily involves the use of machine learning techniques to extract ECG features for predicting and diagnosing risks and prognosis.

CiteSpace 6.3.R1 was employed to create an insightful keyword clustering diagram where thematic labels for each cluster were meticulously selected utilizing the log-likelihood ratio (LLR) method ensuring that each cluster’s essence is accurately captured and presented (21). A total of 21 primary clusters were extracted with the system selecting the top 18 clusters labeled as #0-#17 (Figure 5B). The modularity score of a network indicates the clarity of its structure at the level of decomposed clusters whereas the silhouette score of a cluster measures the homogeneity of its members. In this clustering the modularity score (Q) is 0.7686 and the average silhouette score (S) is 0.8959. With an average silhouette score (S) greater than 0.3 and a modularity score (Q) exceeding 0.5 this clustering is deemed credible and effective (21). The top 10 clusters in descending order are: #0 feature extraction #1 classification #2 percutaneous coronary intervention #3 air pollution #4 deep learning #5 cardiac MRI #6 heart failure #7 machine learning #8 diagnostic model and #9 cardiovascular disease (Figure 5B). These 10 clusters represent the primary research hotspots. According to the main cluster distribution table (Table 3) the oldest cluster is #1 classification. Cluster #7 machine learning has an average publication year of 2012. The youngest cluster is #16 random forest. In particular cluster #15 cardiac remodeling holds the highest silhouette score of 0.975 (Figure 5B and Table 3).

Timeline plots provide a visual representation of the duration and historical progress of each group, effectively reflecting the trends mentioned earlier (22). CiteSpace6.3.R 1 Keyword timeline map is drawn after keyword clustering (Figure 5C). A node represents a keyword, and the location of the node corresponds to the year when the keyword occurred. The size of the node is proportional to its frequency of occurrence. In the right keywords of the timeline plot cluster clusters, these clusters are labeled in different colors at different times in the lower left corner. In the keyword timeline, the keyword machine learning first appeared in 2014, and after 2018, the frequency of machine learning keywords doubled (Figure 5E). The cluster deep learning keywords first appeared in 2018, and the frequency of keyword occurrence observed a significant increase in 2021 (Figure 5F). In traditional machine learning methods, researchers usually rely on statistical learning techniques, such as support vector machines, decision trees, and random forests, to process and analyze medical data (20). These methods perform well in feature extraction and classification tasks but require substantial feature engineering work. Feature engineering is a complex and time-consuming process that requires researchers to have rich domain knowledge and experience in order to manually select and construct effective feature sets (23). Moreover, the performance of machine learning methods is often limited by the quality and quantity of selected features. With the emergence of deep learning technology, the field of mental infarction research has ushered in a major breakthrough. Deep learning, especially the convolutional neural network (CNN) and the recurrent neural network (RNN), has the ability to automatically learn and extract data features (21, 24), thus greatly reducing the dependence on feature engineering. This technique can capture complex patterns and relationships from raw medical data, in turn, providing more accurate and comprehensive disease analysis. The introduction of deep learning brings significant improvements to the risk assessment model of mental infarction. Through the training of deep neural networks, the model can more accurately identify the factors associated with the risk of infarction of patients and even predict the patients’ disease progression. Moreover, in addition to risk assessment, deep learning has also played a key role in the diagnosis and treatment strategies of myocardial infarction. For example, in medical imaging analysis, deep learning techniques can accurately identify and segment abnormal regions in cardiac imaging (25) to provide doctors with a more accurate basis for their diagnosis. At the same time, through deep mining of patients’ medical history and physiological data, deep learning helps developing more personalized and effective treatment plans (26). The transition from machine learning to deep learning marks a new era for the study of myocardial infarction. Deep learning not only improves the efficiency and accuracy of data processing but also provides more powerful and accurate tools for the prevention, diagnosis, and treatment of myocardial infarction (2) with the continuous progress of technology and the increasing wealth of medical data, it is reasonable to believe that deep learning will play an increasingly important role in the field of mental infarction research. According to the timeline analysis of keywords and clusters, the latest study is # 16 random deep forest. The latest study shows that the machine learning method of random forest is applied to establish a reliable diagnostic prediction and diagnostic model for the diagnosis of myocardial infarction, which provides potential genetic markers for the screening of early-stage disease (27, 28).

Keyword outbreak refers to keywords that are used more frequently in a short period of time and can predict research trends and research frontiers in the field (29). CiteSpace6.3.R1 Draw the keyword bursts map (Figure 5D). Red figure represents the outbreak time of the keyword, indicating that the keyword is frequently cited during this period, and the blue line represents the appearance and duration of the keyword. The keyword with the longest and strongest outbreak time in 2008–2020 is artificial neural networks (ANN), indicating that the application of ANN in MI and ML research fields is the research trend. The latest outbreak keyword of 2022–2024 is society. According to the relevant literature analysis, the deep learning system can predict mechanical infarction by rapidly measuring plaque volume and stenosis severity by coronary computed tomography angiography (CCTA) (17, 26) is the latest research trend in the field of MI and ML.

Myocardial infarction (MI), as one of the leading causes of death and disability worldwide, requires early and accurate diagnosis for timely intervention and improved patient outcomes. Traditional electrocardiogram (ECG) analysis relies on the experience and subjective judgment of doctors, posing risks of misdiagnosis and missed diagnosis. The introduction of deep learning technology provides a new solution for automatic, rapid, and accurate diagnosis of MI.

Deep learning is very promising in extracting cardiac strain information quickly and accurately (4, 32). Deep learning will be accelerating from 2021. Deep learning in a wide range of visual applications shows a strong ability, to learn effective hierarchy, all kinds of deep learning architecture (such as deep neural networks, deep belief networks, and recurrent neural networks) have been widely used in the medical field, the results are comparable to human experts, even in some cases is superior to human experts (33). One of the advantages of deep learning models is that they can achieve high-precision performance without a large amount of signal processing, and usually, only a small number of signals are needed to effectively classify signals. In recent studies, none of the authors who developed deep models for MI detection explained how these models made predictions, and the lack of interpretation of these model mechanisms also poses challenges, making clinicians lose confidence in using deep models to assist diagnostic decisions in clinical settings (4). In recent years, the application of deep learning in MI research has been emerging. For example, a CNN-based automated ECG analysis system can detect ST-segment elevation myocardial infarction (STEMI) in real time to win valuable time for emergency treatment (34), The LSTM model achieves early warning and risk assessment of MI by continuously monitoring ECG signals in patients (35). In addition, there are studies applying deep learning to cardiac MRI image analysis to help doctors identify myocardial scar tissue after myocardial infarction (36).