Harnessing artificial intelligence of things for cardiac sensing: current advances and network-based perspectives

Hao Ren, Fengshi Jing, Yuan Ma, Ruining Wang, Chaocheng He, Yufan Wang, Jian Zhou, Yu SunFrontiers in Public Health 2025 · DOI

TL;DR

This paper provides a comprehensive bibliometric analysis of the rapidly growing field of AI-powered IoT for cardiac sensing. It highlights current advancements, identifies key research trends and collaboration networks, and points towards future directions, emphasizing the critical need for data privacy and explainable AI in this domain.

What Problem Does This Paper Address?

Cardiovascular diseases (CVDs) are a leading cause of mortality worldwide, necessitating effective early detection and continuous monitoring. Traditional methods like ECG and Holter monitoring have limitations. While Artificial Intelligence (AI) and the Internet of Things (IoT) offer promising solutions, the integration of these technologies (AIoT) for cardiac sensing is a complex, interdisciplinary field with fragmented research. This paper aims to map the landscape of AIoT for cardiac sensing, identify key advancements, and outline future challenges and opportunities.

Methods

The authors employed a bibliometric approach using data from the Web of Science Core Collection (SCIE and ESCI indices) from January 1, 2018, to December 26, 2024. They conducted a comprehensive visualization analysis using CiteSpace to map countries/regions, academic publications, organizations, authors, citations, and key terminologies. This involved analyzing co-authorship, co-citation, and keyword co-occurrence networks to understand research trends, collaboration patterns, and thematic clusters.

Datasets

The paper discusses several benchmark datasets crucial for AIoT cardiac sensing research:

  • PPG-DaLiA: Wrist-worn PPG and tri-axial accelerometer data for heart rate estimation during daily activities.
  • TROIKA Dataset: Wrist-based PPG and accelerometer data for heart rate estimation during high-intensity exercise.
  • WESAD (Wearable Stress and Affect Detection): Multimodal data (ECG, PPG, EDA, skin temperature) for stress and emotion detection.
  • CapnoBase Dataset: Synchronized PPG and clinical-grade ECG recordings.
  • MIT-BIH Arrhythmia Database: A classic dataset of annotated ECG beats from Holter monitors, considered the gold standard for arrhythmia classification.

Key Insights and Ideas

  • Explosive Growth: The field of AIoT for cardiac sensing has seen a significant surge in publications, with a 500% increase from 2018 to 2024, indicating intense research activity.
  • Dominance of Deep Learning: Keywords like "machine learning," "deep learning," and "neural network" are central, highlighting the reliance on advanced AI algorithms for processing complex cardiac signals.
  • Global Collaboration Patterns: While the US and China lead in research output, international collaboration is still somewhat limited, with a greater concentration within national networks. European countries and Japan show stronger cross-border collaboration.
  • Academia's Leading Role: Academia plays a dominant role in research, providing fundamental insights and theoretical support, while industry contributes to technological applications and commercialization.
  • Thematic Clusters: Key research areas include atrial fibrillation detection, heart rate variability analysis, wearable sensor development, and deep learning architectures for cardiac sensing.
  • Data Privacy is Paramount: As AIoT systems become more integrated, ensuring patient data security and privacy, especially with the rise of large language models, is a critical emerging challenge.
  • Explainable AI for Trust: The "black-box" nature of AI models poses a challenge for clinical adoption. Developing interpretable or explainable AI is crucial for building trust among clinicians and patients.

Key Results

  • Publication Trends: Annual output in AI-driven IoT-cardiology research was modest from 2018-2020, with a slight dip in 2022, potentially due to the COVID-19 pandemic. However, a significant acceleration in publication volume was observed post-2019, with over 300% growth between 2020 and 2024.
  • Top Research Countries: The United States leads in publication volume, followed by China and India.
  • Leading Institutions: The Chinese Academy of Sciences and the Egyptian Knowledge Bank are among the top institutions publishing in this field.
  • Prolific Authors: Professor U Rajendra Acharya is identified as a highly prolific author, with 41 publications since 2018.
  • Key Journals: IEEE Transactions on Biomedical Engineering, IEEE Access, and Biomedical Signal Processing and Control are among the most cited journals.
  • Machine Learning Performance: Various ML models achieve high accuracies, with deep learning approaches like CNNs and LSTMs demonstrating excellent performance. For instance, one study reported an overall classification accuracy of 99.13% using a CNN-LSTM system. Another achieved 97.3% accuracy for ventricular ectopic beat classification using a bCNN.

Limitations and Future Work

  • Data Heterogeneity: The paper acknowledges the need to address challenges in data heterogeneity, on-device inference efficiency, and model explainability for robust, real-time cardiac monitoring.
  • Database Limitations: The reliance on the Web of Science database may exclude significant contributions from conference proceedings, smaller journals, or patent literature.
  • Future Directions:
    • Prioritizing data privacy and security, especially with the advent of large language models.
  • Developing more interpretable and explainable AI models to foster clinical trust.
  • Integrating AIoT into home-based cardiac kits and telehealth platforms.
  • Enhancing the design of wearable and implantable sensors and refining data collection protocols.
  • Leveraging RAG-based pipelines and similar approaches to augment LLM outputs with dynamically retrieved medical knowledge.
  • Focusing on public health strategies for early detection of heart diseases using AIoT-powered cardiac sensing.

Summary updated

Abstract

From our perspective, current advancements in AI-powered IoT cardiac sensors primarily focus on optimizing AI algorithms, such as deep learning techniques, and enhancing the functionality of smart wearable devices for precision medicine. Looking ahead, we anticipate that this field will increasingly prioritize data privacy protection, particularly in the era of large language models, to address emerging challenges and ensure sustainable growth. In summary, we need to continue harnessing the power of AI-powered IoT for cardiac sensing as part of public health strategies to enable early detection of heart diseases.
BibTeX
@inproceedings{ren2025harnessing,
  author = {Hao Ren and Fengshi Jing and Yuan Ma and Ruining Wang and Chaocheng He and Yufan Wang and Jian Zhou and Yu Sun},
  title = {{Harnessing artificial intelligence of things for cardiac sensing: current advances and network-based perspectives}},
  booktitle = {Frontiers in Public Health},
  year = {2025},
  doi = {10.3389/fpubh.2025.1569887},
}
Figure