Machine Learning in Electronic and Biomedical Engineering, Part 3

1. Introduction

The success of the first two Special Issues on Machine Learning in Electronic and Biomedical Engineering [1,2] has demonstrated the growing relevance of data-driven methods across healthcare technologies, smart sensing systems, signal processing, and industrial automation. This third volume continues in the same direction, presenting mature examples of how artificial intelligence is now deeply embedded into the development cycle of electronic systems—from hardware design and optimization to deployment on edge devices and medical environments.

As in the previous editions, the contributions collected here span multiple application domains, confirming the increasingly transversal role of machine learning across engineering. Below, we introduce the main themes represented and summarize each included contribution.

2. Overview of the Contributed Papers

2.1. Embedded and Edge AI Systems

The consolidation of real-time machine learning at the edge is one of the clearest trends emerging from this Special Issue. A prominent study conducted by Calì et al. (Contribution 1) demonstrates the optimized implementation of YOLOv3-Tiny on FPGA, achieving high-speed object detection with low power consumption, confirming the critical importance of hardware-aware design in embedded computer vision. Similarly, Hobbs et al. (Contribution 2) demonstrate that hand gesture recognition entirely executed on a Raspberry Pi 5 enables intuitive human–machine interaction without external processing, showing how edge platforms are now capable of hosting responsive AI services.

Wearable and mobile systems also benefit from this evolution. The work realized by Kim et al. (Contribution 3) presents a true wireless stereo device equipped with a photoplethysmographic (PPG) sensor that integrates onboard inference to detect improper device usage in real time, eliminating bandwidth and latency concerns typical of remote computation.

Together, these works reflect a broader shift: machine learning is no longer confined to cloud computing, but is moving toward efficient execution directly where data are generated.

2.2. Medical and Healthcare Applications

Healthcare remains one of the most active fronts in the evolution of machine learning in electronic systems. An explainable AI framework for personalized diabetes risk prediction, presented by Maimaitijiang et al. (Contribution 4), provides transparent decision support and connects to patients through a chatbot interface, emphasizing usability and trust in clinical environments. Salvi et al. (Contribution 5), using machine learning methods, analyze factors associated with tooth loss in U.S. adults, offering data-driven insights from large-scale epidemiological datasets.

Machine learning also supports medical training. A multi-modal fusion network employing multi-head self-attention is proposed by Li et al. (Contribution 6) for the objective assessment of injection procedures, demonstrating how AI tools can enhance medical education by providing structured and quantitative performance evaluation. These contributions confirm the central role of AI not only in diagnosis, but in education, prevention, and health management.

2.3. Industrial Applications

Industrial environments continue to benefit from intelligent monitoring and diagnostic systems. Jeon et al. (Contribution 7) propose an explainable artificial intelligence pipeline for predictive maintenance in wafer transfer robotics, combining data-driven methods with interpretability—a key requirement for deployment in high-risk, mission-critical settings. Attention-based LSTM networks are also employed by Kim et al. (Contribution 8) to improve endpoint detection in plasma etching processes, contributing to greater precision in semiconductor manufacturing.

Another contribution, by Avanzato et al. (Contribution 9), presents a biometric verification approach based on phonocardiogram fingerprinting combined with multilayer perceptron classification, demonstrating the feasibility of robust and non-invasive identity authentication in electronic systems. These advances illustrate how automated inference is increasingly integrated into the operational logic of industrial systems, supporting safety, efficiency, and continuous monitoring.

Another interesting contribution by Kabashkin et al. (Contribution 10) proposes a hybrid model combining AI, blockchain, and analytics for aircraft health monitoring, in the field of aviation industry. This work exemplifies cross-domain innovation, integrating secure data management and predictive maintenance in aerospace engineering through ML-enabled decision frameworks. This interdisciplinary paper demonstrates how ML-based decision systems can be applied to aerospace maintenance and safety.

2.4. AI Development Frameworks and Optimization Pipelines

Beyond applications, several contributions focus on the evolution of tools and methodologies enabling the design and deployment of advanced machine learning systems. The ADNA framework proposed by Lane et al. (Contribution 11) provides automatic generation of neural network accelerators for ASICs, a significant step toward the standardization and acceleration of dedicated AI hardware development. The use of automated hyperparameter optimization with Optuna for disease prediction further illustrates how systematic model exploration can substantially improve predictive performance, as demonstrated by Lai et al. (Contribution 12).

Grailoo et al. (Contribution 13) conducted a study on heterogeneous edge computing for molecular property prediction that combines CPUs with specialized accelerators through graph convolutional networks, confirming that distributed hardware architectures can accelerate demanding inference tasks.

Additionally, El Barkani et al. (Contribution 14) show how intelligent sensing can be implemented even under extremely limited computational budgets presenting a compact tiny machine learning system for gas leakage detection.

These works collectively highlight the rapid maturation of AI development environments, from high-level design automation to constrained hardware deployment.

2.5. Review Contribution

Finally, the Special Issue includes a comprehensive review by Dumachi et al. (Contribution 15) examining the use of machine learning in cancer imaging across six major cancer types, synthesizing methods, challenges, and diagnostic performance trends. This contribution serves as a valuable reference for researchers entering the rapidly expanding intersection of oncology and AI.

3. Conclusions

The papers published in this third Special Issue provide a clear picture of the current state of machine learning in electronic and biomedical engineering. Across embedded systems, clinical decision support, industrial monitoring, and automated development pipelines, the contributions show how AI is moving toward higher performance, greater interpretability, reduced computational cost, and deeper integration with electronic platforms.

We expect the advances presented here to stimulate further research on efficient deployment, hardware–software co-design, and trustworthy AI models capable of operating in real-world environments.

We would like to thank all the authors for their valuable contributions and all the reviewers for their fruitful comments and feedback to help improve the quality of this Special Issue, and we hope that this volume will serve the community as a reference and an inspiration for future developments. Special thanks also go to the Editorial Board of MDPI’s Electronics journal for the opportunity to Guest Edit this Special Issue, and to the Editorial Office staff for their valuable and timely support.