Machine Learning in Electronic and Biomedical Engineering, Part 2

1. Introduction

The fast evolution of machine learning methods in both electronic and biomedical engineering continues to transform how data is interpreted, validated, and translated into actionable support systems. Following the significant scientific impact of the first edition of this Special Issue [1], the second volume brings together high-quality contributions that demonstrate the maturity of the field and its growing relevance in real clinical and biomedical applications.

Machine learning now supports an impressive spectrum of tasks, including disease risk prediction, medical imaging interpretation, speech-based diagnostics, physiological signal modeling, and intelligent patient monitoring. As healthcare systems move toward personalized medicine and remote decision support, the need for reliable computational approaches has never been greater.

The papers selected for this volume illustrate the current state of the art in this area, demonstrating deep learning innovations, evaluation studies, and methodological advancements with relevance to real-world deployment.

This Special Issue invited contributions on the following topics:

• Machine learning for biomedical signal and image processing;
• Decision support systems and diagnostic automation;
• Computational modeling for patient risk assessment and monitoring;
• Embedded processing and reliable clinical deployment of algorithms;
• Interpretable AI and robust model evaluation.

The papers present in this Special Issue reflect progress across these themes and highlight promising trends for future research and application.

2. Overview of the Contributed Papers

2.1. Predictive Models for Disease Risk Assessment

Stroke prediction is explored by Mia et al. (Contribution 1) through a comparative machine learning study that evaluates several algorithmic approaches to identify cerebral stroke risk factors, reporting significant improvements over conventional scoring methods. A similar focus on clinical prediction is found in a deep learning framework developed by Goretti et al. (Contribution 2) for forecasting congestive heart failure progression, offering a refined probabilistic understanding of individual patient trajectories. These works emphasize how machine learning can complement clinical judgement through early detection and continuous monitoring.

2.2. Advances in Medical Imaging and Diagnostic Support

Several contributions demonstrate sophisticated deep learning solutions for medical imaging.

Thomas et al. (Contribution 3) propose an ensemble of deep meta-learners for COVID-19 classification from CT scans, achieving robust performance in spite of variations in imaging quality. Park et al. (Contribution 4) analyze cervical spine radiographs using a deep learning model to classify foraminal stenosis, presenting a tool that may support improved triage and radiological interpretation.

A convolutional framework is presented by Pasini et al. (Contribution 5) for the segmentation of 18F-FDG PET images in support of Alzheimer’s disease diagnosis, reducing subjective variability in image interpretation.

In dermatology, Foahom Gouabouan et al. (Contribution 6) demonstrate that an end-to-end decoupled training approach mitigates the challenges imposed by highly unbalanced lesion datasets, improving classifier robustness for automated skin lesion evaluation.

A systematic evaluation of surgical tool recognition across heterogeneous laparoscopic recordings demonstrates how convolutional neural networks behave under differing clinical acquisition conditions, providing insights essential for robust computer-assisted surgery systems, as described by Abdulbaki Alshirbaji et al. (Contribution 7).

2.3. Synthetic Data and Augmentation for Limited Biomedical Datasets

Data scarcity remains a core limitation in specialized medical domains. A solution is introduced in Contribution 8 by Mohanty et al., where generative modeling is used to produce realistic synthetic facial images for rosacea cases, substantially enhancing dataset diversity and supporting the development of trained classifiers when clinical datasets are limited.

2.4. Speech and Voice-Based Diagnostics

The use of speech for minimally invasive diagnostic evaluation is covered in a comprehensive systematic review summarizing machine learning methods employed for detecting neurological, laryngeal, and mental disorders through vocal biomarkers. In this work, Sayadi et al. (Contribution 9) provide a structured overview of methods, datasets, performance metrics, and open challenges, forming a solid reference for future investigations.

2.5. Machine Learning for Patient Monitoring and Personalized Therapy

A machine learning-based decision support system for children and adolescents with type 1 diabetes, developed by Campanella et al. (Contribution 10), using hybrid closed-loop insulin systems demonstrates how automated analysis can support personalized therapeutic strategies and improved management outcomes.

2.6. Methodological and Algorithmic Foundations

Beyond application-oriented studies, this Special Issue includes contributions addressing fundamental computational modeling. A framework using Particle–Bernstein polynomials is proposed by Alessandrini et al. (Contribution 11) for nonlinear dynamic system identification in the spectral domain, offering promising applicability to complex physiological systems and biomedical instrumentation.

2.7. Conclusions

This second volume of the Special Issue underscores a research community progressing steadily toward validated and clinically meaningful innovation. Across the contributions, we observe increasing methodological rigor, heightened focus on robustness and interpretability, and continued interest in bringing machine learning systems closer to routine clinical use.

Future developments are expected to expand on not only performance improvements but also transparency, safety, regulatory integration, and real-world deployment within clinical workflows.

We extend our sincere thanks to all authors for their high-quality submissions, the reviewers for their constructive evaluations, and the editorial team of Electronics for supporting this second volume of the Special Issue.