Assessing Africa’s position in the development of AI-enabled ECG devices

A. Basic principle

The electrocardiogram (ECG) is an essential diagnostic device that records the heart’s electrical activity over time using electrodes positioned on the skin and precisely around the heart. It is widely used in medical practice to assess the condition of the heart, identify any irregularities and guide the treatment of various heart diseases. It provides a non-invasive, rapid and reliable method of monitoring heart rhythm and detecting abnormalities that may indicate underlying heart problems.^17,18

The heart’s electrical activity, generated by the spontaneous depolarization and repolarization of the heart muscle, is detected by electrodes placed on the surface of the body. These electrodes are usually attached to the patient’s limbs and chest. A standard ECG records this activity from 12 different perspectives or leads, each offering a unique view of the heart’s electrical function.¹⁹

The fundamental components of a typical ECG waveform are illustrated in Figure 2, detailing the various intervals and segments essential for cardiac analysis.¹⁹

Figure 2. ECG waves and intervals.^17,18

B. Clinical applications of ECG

The ECG is essential for the diagnosis and management of a wide range of cardiac pathologies²⁰:

C. Conventional ECG system

To provide a comprehensive understanding of the components and interactions within a conventional ECG system, Figure 3 offers a clear and structured visualization. It highlights the flow of data and the roles of each component in the process²¹:

Figure 3. Overview of a conventional ECG system.

D. ECG artifacts and their impact on ai-based analysis

One of the main challenges associated with interpreting ECG signals, particularly in the context of AI-based analysis, is the presence of artifacts. These artifacts are distortions that mask or alter the actual cardiac signals.

They can be classified as physiological artifacts (e.g., muscle tremors, breathing) and non-physiological artifacts (e.g., electrode movement, poor contact, baseline drift, and interference from power lines). In conventional clinical settings, expert cardiologists can often identify and ignore these distortions. However, in AI-based ECG analysis, where models rely on the fidelity of input data, artifacts pose a significant risk of false arrhythmia detection, classification error, and overfitting on irrelevant features.^22,23

Many studies have shown the sensitivity of ECG processing algorithms to these types of noise. For example, Gupta et al.²² demonstrated that even advanced methods such as chaos-based R-peak detection can be significantly affected by signal corruption. Similarly, Huque et al.²⁴ reported that HMM-based supervised learning frameworks are particularly susceptible to distorted input signals. The study by Sahoo et al.²³ also highlighted that most machine learning models require significant preprocessing or exhibit reduced reliability when exposed to artifact-laden ECG data.

To solve this problem, researchers have developed various signal denoising and preprocessing techniques. For example, wavelet transforms and principal component analysis (PCA) have been used to clean ECG signals and extract relevant features in noisy environments.^24,25 These approaches aim to reduce the influence of high-frequency noise or baseline drift without compromising waveform features important for diagnosis. Gupta et al.²⁶ applied an FrWT-based arrhythmia detection method enhanced by PCA and demonstrated improved robustness in real-world conditions.

A. Evolution of publication years

Annual scientific output in the field of ECG devices and AI in recent years shows a significant upward trend, accelerating in particular during the COVID-19 pandemic ( Figure 4).

Figure 4. Annual scientific production.

Initially, publication growth between 2019 and 2020 was steady, with 754 articles published in 2019. An increase is observed from 2020 onwards, with publications rising from 754 in 2019 to 923 in 2020, and significantly to 1255 in 2021. This increase is attributed to the influence of the pandemic, which demonstrated the need for advanced remote diagnostic systems in the face of restricted access to healthcare and the increased burden on medical facilities.

The global health crisis underscored the need for advanced, non-invasive monitoring technologies, such as AI-enhanced ECG devices, which attracted growing research interest as the medical community sought new ways to manage patient care remotely. This increase reflects the urgent need to develop reliable, scalable solutions in a context of restrictions on in-person medical visits and increased burden on healthcare systems.

In 2022, the number of articles continued to rise, reaching 1445 publications, then 1547 in 2023, confirming the sustained interest and ongoing development in this field, driven by the persistent challenges posed by COVID-19 and the recognition of potential future pandemics. However, in 2024, a significant drop in publications to 742 is recorded. This reduction could be explained by the fact that we are probably still at the beginning of 2024, which means that data for this year are incomplete. Other factors could also explain this drop. It’s possible that the intense interest generated by the pandemic is reaching a saturation point, with many solutions and innovations already in place. Researchers can now turn their attention to new challenges or further develop technologies already developed, rather than initiating new publications. This trend underlines the importance of considering the temporal and contextual aspects of publications when analyzing research trends, especially in dynamic fields such as medical technology and artificial intelligence.

Using biblioshiny, a list of the most globally cited documents was constructed based on total citations as presented by Table 1. The first-ranked paper is by Attia ZI, published in 2019, which has received a total of 744 citations, averaging 124 citations per year, and has a normalized citation count of 24.20. This work uses an AI-based ECG algorithm to identify patients with atrial fibrillation.²⁷ This article demonstrates how AI improves cardiovascular diagnostics and helps with better patient management. Following this, Huang J’s paper (2019), ranks second with 384 total citations, an average of 64 citations per year, and a normalized citation count of 12.49.²⁸ The research uses neural networks for the classification of arrhythmias and myocardial infarction. The third-ranked paper by Baloglu UB, published in 2019 with 314 total citations, 52.33 citations per year on average, and a normalized count of 10.21.²⁹ The research demonstrates the importance of deep learning techniques for improving the accuracy of ECG diagnoses. On the fourth position is Andersen RS’s paper, published in 2019 with 286 total citations, 47.67 citations per year, and a normalized citation count of 9.30.³⁰ Their document shows the importance of rapid diagnostics for cardiovascular disease management. Finally, Yao Q’s paper, published in 2020, ranks fifth with 281 total citations, averaging 56.20 citations per year, and a normalized citation count of 11.83.³¹ Their research uses deep learning for real-time arrhythmia detection showing the importance of rapid diagnostics for cardiovascular disease management.

These documents reveal the growing impact of AI in ECG technologies. They show how AI enables more accurate diagnoses, continuous monitoring and opens up new avenues for medical research. AI is emerging as an essential tool for improving cardiovascular healthcare, as evidenced by the numerous citations and lasting influence of this work.

B. Main research keywords in AI for ECG technology

The analysis of the main keywords in this field also shows several distinct and interconnected clusters. Each represents a key aspect of AI research and application in this field.

The green cluster in Figure 5, which focuses on terms such as ‘machine learning’, ‘signal processing’ and ‘neural networks’, highlights the importance of machine learning techniques and signal processing in ECG analysis. Researchers in this cluster are focusing on the development of algorithms to improve the detection and classification of cardiac arrhythmias. They stress the need to ensure that these technologies are robust and applicable in real clinical environments in order to maximize their usefulness and impact.

Figure 5. Co-occurence keywords network.

The blue cluster includes keywords such as “deep learning”, “convolutional neural network” and “arrhythmia classification”. This cluster is strongly focused on the use of deep neural networks and other advanced AI methods. They are also used to diagnose cardiac arrhythmias. The aim is to provide more accurate and faster diagnoses, but the practical application of these technologies still needs to be validated on a large scale to ensure their reliability and effectiveness in various clinical contexts.

The red cluster focused on terms such as ‘major clinical study’, ‘sensitivity and specificity’ and ‘data analysis’. This highlighted the importance of clinical studies and data analysis to validate AI technologies in healthcare. This cluster shows efforts to evaluate the clinical effectiveness of new technologies and their integration into routine medical practices. However, it is important to standardize evaluation methods to ensure the comparability of results between different clinical studies.

The yellow cluster with keywords such as “wearable sensor”, “photoplethysmography” and “heart rate variability” shows the integration of wearable sensors and non-invasive technologies in the monitoring and analysis of ECG signals. Research in this area aims to develop wearable devices capable of continuously monitoring cardiac parameters. Although these technologies are promising, questions persist about their accuracy, patient comfort and ability to be integrated into existing healthcare systems.

Overall, Figure 5 provides a good illustration of the interdisciplinary nature of AI-enabled ECG device research.

However, some areas, such as wearable sensors and machine learning methods are more densely explored than others. This suggests potential areas for future research. The concentration of certain keywords around specific themes may indicate research silos, where teams could benefit from closer interdisciplinary collaboration to fill gaps and strengthen innovation. This keyword analysis provides an overview of current trends and research priorities, while highlighting opportunities for collaboration and future development to improve ECG devices using artificial intelligence.

C. Author influence and collaboration

Analysis of the influence of authors in the field of ECG devices using AI shows that Acharya UR has the highest h_index of 29. This indicates that his work is widely cited and recognised. He also has the highest number of total citations (2668) and a g_index of 51. This underlines the breadth and impact of his publications. LI Y follows with an h_index of 25 and a g_index of 42. LI Y has published the largest number of articles (127) and has accumulated 2086 citations. This shows his strong research activity and the importance of his contributions. LIU C, with an h_index of 23 and a g_index of 43, also shows notable influence. He has published 79 articles with 1995 citations.

Table 2 ranks the top 10 authors according to their contributions since 2019. The m_indexes show that these researchers maintain a high level of productivity throughout their career. Acharya UR and LI Y’s m_index scores, above 4, reflect remarkable productivity. This trend highlights the growing interest in AI-based healthcare technologies and the significant impact of these authors in the field of ECG devices.

The model of collaboration between these authors in this field shows interesting dynamics and significant trends (Figure 6).

Figure 6. Author collaboration network.

Firstly, it is clear that certain authors, such as Acharya UR play an important role in the network of collaborations. This type of centralization can be beneficial, as it enables knowledge to be disseminated quickly and research projects to be coordinated effectively. The close collaboration between Acharya UR, Tan RS, and Pachori RB shows strong cooperation, which is often essential for the rapid development of new technologies. However, this centralization can also pose problems, such as over-reliance on a limited number of influential researchers. This can limit the diversity of perspectives and ideas, and potentially hinder innovation in the long term.

In parallel, LI Y and his collaborations with Liu C and Zhang H show significant but less dense connections. This may indicate more specialized research and less diffusion of knowledge between groups. Less dense clusters may be a sign of fragmentation in the field, where research focuses on specific niches without sufficient integration with other subfields. This fragmentation could limit the potential synergies between different approaches and technologies, thus slowing down overall progress.

The cluster around Noseworthy PA and Friedman PA illustrates a well-connected network, but it is questionable whether these collaborations are sufficient to cover all the research needed in this field. Intense collaborations within small groups can lead to significant advances, but they also run the risk of creating knowledge silos, where innovations remain confined to a small group of researchers.

It is also notable that some authors seem isolated or less connected. These authors could represent innovative approaches or important research niches that are not yet integrated into the main network. This raises the question of how these researchers can be better integrated into the collaborative network to maximize their contribution and avoid duplication of effort.

D. Influential journals

The analysis highlights the main journals contributing to AI-enhanced ECG device research, presenting their performance. IEEE Access stands out with the highest h_index of 44, a g_index of 69 and an m_index of 7,333. It has a total of 6,562 citations from 284 publications. This indicates its substantial impact and wide recognition in the research community. The journal’s high metrics reflect its role as a leading outlet for AI and biomedical engineering research. It provides a platform for the dissemination of influential studies. Biomedical Signal Processing and Control follows with an h_index of 34, a g_index of 55 and an m_index of 5,667, accumulating 5,504 citations from 401 articles. The journal’s focus on signal processing and control systems in biomedical applications making it an essential source of knowledge for advances in AI-based ECG technologies. Its high number of citations and significant number of publications indicates its importance in this niche field. Sensors (Switzerland) also ranks highly, with an h_index of 34, a g_index of 48 and an m_index of 5,667. It received 3,677 citations from 137 articles. This journal plays a central role in the field of sensor technology, particularly in the development of portable, non-invasive monitoring systems, which are an integral part of modern ECG devices.

Overall, Table 3 illustrates the key journals playing in the advancement of AI-enhanced ECG device research. The high citation counts and robust indices of these journals reflect their importance and influence in the academic community. These journals serve as essential platforms for publishing innovative research. This facilitates the dissemination of new ideas and technological advances in ECG devices and AI applications. Measurements suggest that researchers frequently refer to these publications. They underline their relevance and the quality of the research they attract.

E. Institutions and their contributions

An analysis of the most active institutions in the publication of articles on ECG devices using AI shows some significant trends ( Table 4). The Mayo Clinic in Rochester in the United States stands out with a high number of articles published. This institution is world-renowned for its expertise in cardiovascular medicine and plays an important role in the development of AI ECG technologies. The participation of several Mayo Clinic departments in the publications reinforces this observation. It also highlights a solid research infrastructure and effective internal collaboration.

The University of Wuhan in China also shows a notable ability to integrate the physical sciences and technology into medical research. Its School of Physics and Technology is a major contributor, with numerous articles published. This institution is demonstrating the importance of an interdisciplinary approach to progress in the field of ECG devices using AI.

Seoul, South Korea is another example of the growing importance of Asian research. The contribution from this region highlights the effective integration of technological innovations into medicine.

These institutions show how leaders in the field are playing an important role in the development and advancement of AI-enabled ECG technologies. Their collaborative efforts foster rapid innovation and continuous improvement of medical devices.

The analysis showed significant dynamics in citation rates between institutions in the field of AI-enabled ECG devices. It highlights the main contributing institutions and their collaborative networks. The Department of Cardiovascular Medicine at the Mayo Clinic in Rochester clearly stands out with 41 papers and a total of 1457 citations, as well as a total link strength of 173. This institution plays a central role, indicating substantial scientific production and significant influence through numerous citations. Similarly, the Department of Electronics and Computer Engineering at Ngee Ann Polytechnic in Singapore, with 15 papers and 1361 citations, and a total link strength of 159, is also a key contributor. This demonstrates the successful integration of electronics and computer engineering in medical device research. The high number of citations for these two institutions shows the substantial recognition of their work by other researchers.

The School of Medicine, Faculty of Health and Medical Sciences, despite publishing only 6 papers, receives 919 citations and has a total link strength of 144. These figures show that despite a smaller number of publications, the research carried out by this institution is of high quality and has a significant impact.

Figure 7 provides a visual illustration of the citations between these institutions. It highlights the centrality of certain organizations. The Mayo Clinic’s Department of Cardiovascular Medicine is a central node with numerous links, indicating frequent recognition and collaboration by other researchers. The dense clusters around the Mayo Clinic and Singapore institutions such as Duke-NUS Medical School and the National Heart Centre Singapore show close cooperation and research synergy that facilitate rapid technological advancements.

Figure 7. Collaboration network among top institutions in AI-enhanced ECG research.

However, some institutions appear isolated on the network map, such as the Department of Biomedical Engineering. This isolation may indicate specialization in very specific niches or less collaboration with other institutions.

F. Countries and their contributions

Analysis of the table of countries of corresponding authors in the field of ECG devices using AI highlights varied dynamics in terms of scientific production and international collaboration. China stands out with 1,572 articles, demonstrating its significant scientific output. It has 1256 single-signature papers (SCP) and 316 multiple-collaboration papers (MCP), the MCP ratio of 0.201 indicates a moderate tendency to collaborate with researchers from other countries. Although China produces a large number of articles, a significant proportion of this production is carried out independently.

India follows with 899 articles. Its MCP ratio of 0.097, the lowest among the countries listed, suggests that most of the research is carried out independently, without any significant international collaboration. This low level of collaboration could limit the impact and dissemination of Indian research at global level.

The United States comes third with 643 articles. Their MCP ratio of 0.247, one of the highest, shows a strong inclination to collaborate internationally. This reflects the globalized nature of American research and its integration into world research networks, favoring greater dissemination and a more significant impact of their work.

Overall, Table 5 reveals significant disparities in international collaboration rates. Countries such as the United States and the United Kingdom, with high MCP ratios, show a strong inclination to collaborate beyond their borders, which can amplify the impact of their research. On the other hand, countries such as India and China, despite their high output, show less marked tendencies towards international collaboration, which could limit the dissemination and overall impact of their work. This trend could be explained by cultural factors, research infrastructures or national policies on scientific collaboration.

Another important aspect is collaboration between countries in this field. The analysis shows several potential clusters and interesting collaborative dynamics. The cluster comprising the United States, Canada and several European countries is central. It is characterized by strong transatlantic cooperation. The United States emerges as a central node with numerous connections to other countries, notably Canada, the United Kingdom, Germany, Italy and Belgium. This North American and European collaboration reflects a high level of integration and scientific cooperation which promotes shared innovation. European countries such as Italy, Spain and Belgium are strongly interconnected and form a sub-cluster within this larger entity. This illustrates a solid infrastructure for scientific collaboration within Europe and with North America.

The Asian cluster of China, India, Taiwan and Japan shows close regional collaboration and significant interaction between these countries. China and India, with large node sizes, are leaders in terms of scientific production in this field, but their collaborations seem to be more regional than international. Taiwan stands out for its significant connections with other Asian countries, while Japan, although important, seems less centralized than China and India. This cluster highlights the importance of intra-regional collaboration in Asia.

The Eastern Europe and Middle East cluster, which includes Iran and Turkey. This cluster shows collaboration with European and Asian countries. Turkey, in particular, acts as a bridge between Europe and Asia, encouraging scientific exchanges between these regions. Overall, the diagram illustrates a high density of international collaborations, with countries such as the United States, China and India playing central roles. The United States shows a strong tendency to collaborate with Europe and Canada, while China and India focus more on regional collaborations in Asia.

However, there is also relative isolation, particularly in some African and South American countries, suggesting less integration into the global research network. To maximize the overall impact of research into AI-enabled ECG devices, it would be beneficial to encourage more inclusive collaborations that integrate these less connected regions. Figure 8 shows a complex but highly interconnected collaboration dynamic, with strong regional clusters and a notable ability to exchange knowledge across borders. Strengthening international collaborations could further improve innovation and the dissemination of advanced technologies in the field of AI-based ECG devices.

Figure 8. Country collaboration network.

A. Africa’s position in electrocardiogram research

The analysis shows a significant under-representation of African countries in AI-enhanced ECG research. It highlights the critical challenges and obstacles facing African scientists in this rapidly evolving field. African researchers and institutions have minimal visibility in the global AI-enhanced ECG research field. This is marked by the absence of African countries and institutions among the main contributors in terms of publications, citations and collaborative networks. Leading institutions from the USA, Europe and Asia dominate the field without significant representation from Africa. The study by Tuyishime et al. (2022)³² shows that publications from African institutions account for less than 1% of articles published in the most influential cardiology journals.

One of the main obstacles is the limited research infrastructure in many African countries. The advanced facilities needed to conduct high-quality research in AI and medical technology are often unavailable. This strongly limits the ability of African researchers to engage in advanced technology related studies. Balogun et al (2023)³³ noted the absence of laboratories equipped with advanced technologies in several African countries. This affects research conducted in Africa as compared to research conducted in developed countries. In addition, AI and medical device research generally requires substantial financial investment which is the main challenge for Africa. This financial constraint limits the scope and scale of research activities that Africa can conduct. Overland et al. (2021)³⁴ analyzed funding flows and found that grants dedicated to research in Africa are largely insufficient compared to those awarded to European and American institutions. The cost and maintenance of AI and ECG technology also present significant obstacles, particularly for widespread adoption and research in resource-limited settings. For example, Bodagh et al. (2023)³⁵ pointed out that advanced ECG devices require regular updates and costly maintenance. This represents a significant barrier for African institutions.

The lack of strong collaborative networks further isolates African researchers. Countries such as the United States and the United Kingdom benefit from high levels of international collaboration. This improves the impact and dissemination of their research. Townsend et al. (2023)³⁶ have shown that American researchers frequently publish with international co-authors, thus improving the reach and impact of their work. In contrast, African researchers often have fewer opportunities to collaborate with leading global institutions. This limits their exposure to new methodologies, technologies and ideas essential to advancing research and innovation. Jackson et al. (2022)³⁷ found that African researchers rarely participate in international collaborative research projects.

Efforts to foster collaborations are essential which necessitates international partnerships and research consortia. This will result into integration of African researchers into the global scientific community. For example, Omosa- Manyonyi et al. (2023)³⁸ described a regional training program in East Africa that significantly improved the skills of local researchers and facilitated international collaborations.

The under-representation of African researchers has important implications. It limits the diversity of perspectives and approaches needed to address global health challenges globally. Diverse research teams bring varied experiences and ideas that can lead to more innovative and inclusive solutions. The absence of African voices means that the specific health challenges and needs of African populations may not be sufficiently addressed. Farrell et al. (2023)³⁹ also noted that healthcare technologies developed without taking local particularities into account often perform less well when applied in African contexts.

This situation perpetuates a cycle of underdevelopment in scientific research on the continent. Without substantial contributions to high-impact research areas, African countries struggle to attract funding, build advanced research facilities and develop the human capital needed to compete on a global scale. This gap limits not only scientific progress, but also the potential health benefits that could grow from inclusive research efforts. Yankam et al. (2023)⁴⁰ have shown that the lack of research funding and infrastructure prevents African researchers from fully participating in international projects. Building local research capacity and investing in education and training programs are essential steps to breaking this cycle and empowering African researchers. Asubiaro & Shaik (2021)⁴¹ have suggested that specific training programs and investment in research infrastructure are essential to improve the participation of African researchers in advanced research projects.

Efforts to close this gap are essential. The creation of AI centers of excellence for global health equity in PRFIs, including African countries has been proposed as a strategic vision to strengthen local research capabilities and address these disparities. Bédair et al. (2023)⁴² proposed the establishment of regional centers of excellence to stimulate local research and foster international collaborations. Google has also committed to investing heavily in Africa’s digital transformation, including establishing a product development center in Nairobi and expanding its AI residency programs across the continent.⁴³ These initiatives are designed to improve research infrastructures, provide financial support and foster international collaborations, ultimately aiming to integrate African researchers more fully into the global scientific community.

Ethics and consent

The authors declare that all ethical guidelines were followed in conducting this research.