Special Issue: Emerging E-Health Applications and Medical Information Systems

Over the last few decades, research on e-Health applications and medical information systems has significantly grown over time due to the need to reinforce health applications’ governance and access potential as well as medical information systems’ interoperability, satisfaction, performance, and usefulness. Under the prism of the society’s advancement and the increasing demand for well-being, health information technology constitutes an evolving research field that introduces several challenges and opportunities for the development of innovative healthcare services [1].

Healthcare Information Technology (HIT) is transforming the healthcare domain and offers innovative services that span from chronic disease monitoring [2] and patient engagement [3] to telemedicine solutions [4] and personalized medicine [5].

In the current era of Industry 4.0, cutting-edge digital technologies and devices are extensively utilized for innovation and value creation. The healthcare sector is no different. Hospitals and healthcare providers worldwide are actively implementing digital technologies like artificial intelligence (AI), smart sensors and robots, and the Internet of Things (IoT) to enhance the quality of care and boost operational efficiency.

Artificial intelligence (AI) has attracted the attention of stakeholders in the healthcare industry because of its potential for diagnosis, prediction, and decision-making support [6,7]. As data generation is constantly increasing and computational resources are capable of processing big data, AI technologies become more accurate and efficient.

Autonomous systems, robots, and wearable sensors provide solutions for innovative healthcare services [8] that allow for telesurgery, patient monitoring, and independent living. When combined with machine learning techniques, these technologies can prove to be efficient and safe for wide adoption in the healthcare industry [9,10].

As digital technologies evolve and become affordable, Internet of Things (IoT) applications are integrated into medical devices [11], which allows for novel medical services due to their ability to capture real-time physiological data. Current research efforts on IoT technologies focus on several aspects of evaluation metrics, such as scalability, interoperability, and security, to name a few.

The main objective of this Special Issue was to provide an interdisciplinary platform where researchers share knowledge and success stories, present ambitious system designs, and demonstrate state-of-the-art developments in emerging eHealth applications and medical information systems. Submissions were expected to include information and assessment metrics for the quantitative or qualitative evaluation of various health information technologies that depict their impact.

We feel obliged to thank the scientific community that responded with appreciable efforts, as a significant number of original research articles were submitted and considered for publication. Among the twenty-one (21) originally submitted papers, ten (10) were finally accepted as full papers (acceptance rate = 47.62%) as a result of a careful blind peer-review process with diligent editorial input by the handling Guest Editors. All accepted papers depict significant contributions and cover multidisciplinary application domains.

The first contribution was authored by Siddiqui et al. (Contribution 1): “Respiration-Based COPD Detection Using UWB Radar Incorporation with Machine Learning”. In this work, the authors used ultra-wideband (UWB) radar technology combined with machine learning algorithms for non-invasive detection of chronic obstructive pulmonary disease (COPD). COPD is a severe respiratory condition that can lead to death if not diagnosed and treated early. These researchers collected respiration data from COPD patients and healthy subjects in a hospital setting, taking all safety measures that are crucial during pandemics like the COVID-19 pandemic. Signal processing techniques were applied to the raw respiration data, and additional patient information such as age, gender, and smoking history was incorporated to enhance detection accuracy. Various machine learning models such as Naïve Bayes, Support Vector Machine (SVM), Random Forest (RF), k-Nearest Neighbor (KNN), Adaptive boosting (Adaboost), and deep learning models such as Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks were employed for classification. Their experimental results demonstrated that the LSTM model achieved the highest accuracy of 93%. This study highlights the effectiveness of UWB radar technology for COPD detection and outlines the benefits of combining respiratory data with demographic information to improve diagnostic accuracy. This approach offers a promising, non-invasive alternative for monitoring and diagnosing COPD.

The second contribution, titled “Evaluation of the Predictive Ability and User Acceptance of Panoramix 2.0, an AI-Based E-Health Tool for the Detection of Cognitive Impairment”, was authored by Valladares-Rodríguez et al. (Contribution 2). The authors of this study validated the predictive ability and user acceptance of Panoramix 2.0, an artificial intelligence application, for the early detection of cognitive impairment in older adults. This tool creates a non-intrusive assessment environment by leveraging machine learning algorithms and gamification. The pilot study involved 30 subjects categorized into three groups: healthy controls, individuals with mild cognitive impairment (MCI), and early-stage Alzheimer’s patients. Participants played three serious games in two separate time periods, with data from these interactions used to assess cognitive status. The results revealed that Panoramix 2.0 effectively differentiated between healthy and cognitively impaired individuals, and the best classification performance was achieved using a multilayer perceptron classifier, a SVM, and RF algorithms, with a high weighted recall of ≥0.9784 ± 0.0265 and an F1 score = 0.9764 ± 0.0291. This study indicated no significant concerns from participants. It highlighted the tool’s user acceptance, with an average rating of 4.55/5 from the HC group, 4.13/5 from the MCI group, and 3.99/5 from people with AD. The ease of use was also positively rated, ranging from 4.62/5 for the HC group to 4.19/5 for those with AD. Furthermore, all cognitive groups expressed a willingness to use the digital tool again, with ratings of at least 4.23/5. According to these researchers, Panoramix 2.0 is a promising self-administered tool for cognitive assessment, particularly useful in scenarios with limited access to healthcare professionals, such as during pandemics. Further research focuses on the refinement of predictive patterns.

The third contribution was authored by G. Nokas and T. Kotsilieris (Contribution 3), titled “Preventing Keratoconus through Eye Rubbing Activity Detection: A Machine Learning Approach”. This study proposed an early detection system for eye rubbing to prevent keratoconus, which is a prevalent eye disease characterized by the thinning and cone-like deformation of the cornea, with eye rubbing identified as a significant risk factor. These researchers utilized an Inertial Measurement Unit (IMU) to collect hand motion data, processed through four machine learning classifiers: SVM, Decision Trees (DTs), RF, and eXtreme Gradient Boosting (XGBoost). This study revealed that SVM, RF, and XGBoost outperformed DTs, all achieving high accuracy in detecting eye rubbing activities. This paper concludes with recommendations for future research to refine the detection models and expand their applications in wearable technology, offering a non-intrusive method for monitoring and alerting users, thereby reducing the risk of keratoconus progression.

The fourth contribution was titled “InSEption: A Robust Mechanism for Predicting FoG Episodes in PD Patients”, and it was authored by Dimoudis et al. (Contribution 4). These authors present the InSEption model, an advanced mechanism for predicting Freezing of Gait (FoG) episodes in Parkinson’s Disease (PD) patients. This study focuses on the integration of Internet of Things (IoT) and deep learning technologies to enhance continuous monitoring and personalized treatment of PD. Two inception-based models, LN-Inception and InSEption, were introduced and tested against the well-known Daphnet dataset and a novel dataset collected via IMU sensors. The InSEption model showed superior performance, achieving a 6% increase in macro F1 score on the Daphnet dataset compared to traditional inception-based models. Additionally, on the new IMU dataset, InSEption achieved F1 and AUC scores of 97.2% and 98.6%, respectively. The improved performance is attributed to the inclusion of squeeze and excitation blocks and domain-specific oversampling methods. This study highlights the benefits of using deep learning for signal data analysis and suggests potential integration into wearable IoT devices for real-time monitoring. This could significantly aid in improving the quality of life for PD patients by providing timely alerts and personalized treatment adjustments. This study concludes with suggestions for further research to validate the findings across larger and more diverse datasets.

The fifth contribution is titled “A Projected AR Serious Game for Shoulder Rehabilitation Using Hand-Finger Tracking and Performance Metrics: A Preliminary Study on Healthy Subjects”, and it is authored by Viglialoro et al. (Contribution 5). The authors of this study explored the development and preliminary evaluation of a projected augmented reality (AR) serious game designed for shoulder rehabilitation. It utilizes hand/finger tracking technology and integrates performance metrics to assess patient progress. This innovative approach aims to increase patient motivation, providing an engaging alternative to conventional physiotherapy. The latest version of the system offers two visualization modes—standard on-screen and projected AR—and tracks key movement parameters such as velocity, acceleration, and normalized jerk. Sixteen healthy volunteers, including technical and rehabilitation experts, tested the prototype and revealed that the game is engaging, ergonomically sound, and potentially effective for shoulder rehabilitation. However, further clinical validation is required to confirm its therapeutic efficacy. This study paves the way for future research focused on diagnosing shoulder movement abnormalities using hand/finger tracking technology.

The sixth contribution is authored by Androutsou et al. (Contribution 6), and its title is “Automated Multimodal Stress Detection in Computer Office Workspace”. This study presents a system that automatically detects stress in office environments using a combination of physiological and behavioral measurements. The system employs IoT technologies and data analysis advances to unobtrusively monitor stress levels. Key components of the system include sensors embedded in a custom-made smart computer mouse to measure physiological parameters like PhotoPlethysmoGraphy (PPG) and Galvanic Skin Response (GSR), as well as behavioral data derived from computer keyboard and mouse dynamics. The research involved an experimental protocol simulating an office environment with common work stressors to validate the system. Different classifiers and data labeling methods were applied to the collected data, achieving high-performance metrics. The feature-level fusion analysis of physiological and behavioral parameters detected stress with an accuracy of 90.06% and an F1 score of 0.90. Decision-level fusion, combining features from both the keyboard and mouse, yielded an average accuracy of 66% and an F1 score of 0.56.

Pintelas et al. (Contribution 7) authored the seventh contribution, which is titled “Explainable Feature Extraction and Prediction Framework for 3D Image Recognition Applied to Pneumonia Detection”. These authors explored a novel approach to creating interpretable machine learning models for critical applications, focusing on pneumonia detection using 3D CT images. Traditional deep learning models, while accurate, are often considered “black boxes” because their decision-making processes are not transparent or understandable to humans. This lack of interpretability is a significant drawback in applications where trust and understanding of the model’s decisions are crucial. To address this, these authors proposed a white-box (WB) model that relies on explainable features derived from transparent, mathematical, and geometric concepts. These features include lines, vertices, contours, and area sizes, which are extracted from the contours of each 3D image slice. The proposed method emphasizes creating features that are both interpretable and rotation-invariant, meaning their extraction and representation are not affected by the rotation of the input images. The framework was applied to the pneumonia detection problem, achieving performance comparable to or better than state-of-the-art black-box 3D-CNN models. The results highlight the potential of the proposed approach for making accurate and interpretable predictions, which is particularly significant in medical and other critical real-world applications where understanding the rationale behind decisions is essential. More specifically, the proposed framework exhibited better performance compared to other WB approaches, while it managed to outperform most of the other BB approaches as it achieved the best geometric mean score of 0.883. Among the ML models tested, the SVM classifier achieved, on average, the best results for all feature representation approaches.

The eighth contribution is titled “VaccineHero: An Extended Reality System That Reduces Toddlers’ Discomfort during Vaccination” and is authored by Antonopoulos et al. (Contribution 8). This study introduced an innovative system designed to alleviate the discomfort and fear experienced by children during vaccination procedures. The system employs eXtended Reality (XR) technology to distract children during the vaccination process. The narrative, synchronized with the vaccination, includes a hero whose experiences mirror the real-world actions performed by the healthcare professional. A clinical trial with 16 children and two doctors was conducted, showing that VaccineHero reduced discomfort by 40% and eliminated extreme discomfort. It is a cost-effective solution that is accessible for widespread use in pediatric practices. This application immerses children in a 3D world through a VR headset, where they take on the role of a hero, with a magician character interacting with them in ways that correspond to the vaccination process. This synchronization helps blur the boundaries between the virtual and physical worlds, keeping children calm and distracted. This system uses the Unity game engine and includes the VR content, an Android phone, and a headset suitable for children, with the VR content also displayed on an external monitor for the healthcare professional to follow along. This research demonstrated the potential of XR technology in reducing vaccination-related discomfort and offering a practical and scalable solution for improving the vaccination experience for young children.

Picozzi et al. (Contribution 9) authored the ninth contribution, which is titled “Telemedicine and Robotic Surgery: A Narrative Review to Analyze Advantages, Limitations and Future Developments”. These authors reviewed the integration of telemedicine into surgical practices through robotic systems that enable surgeons to perform procedures remotely. This study highlights the potential of telesurgery to revolutionize surgical care by combining robotic technology with advanced communication systems. A comprehensive narrative review was conducted. The relevant studies were selected from PubMed, Scopus, and Web of Science databases. The inclusion criteria focused on English-language articles published between 2001 and 2023 that detailed long-distance telesurgery involving different hospitals for the surgeon and patient. This review identified several key advantages, such as reduced travel for patients, enhanced precision of surgical procedures, and the potential to address surgical needs in underserved regions. However, this study also highlights significant limitations and challenges, including technical issues related to latency and network stability, the high cost of robotic systems, and the need for specialized training for both surgeons and support staff. Additionally, regulatory and ethical considerations were discussed as hurdles to widespread adoption. These authors concluded by emphasizing the need for further research and development to address existing challenges and optimize the integration of telesurgery into routine clinical practice. They propose future directions, such as improvements in communication technology, cost reduction strategies, and enhanced training programs, to fully realize the potential of telesurgery in transforming healthcare delivery.

The tenth contribution was authored by Sharma et al. (Contribution 10) and is titled “Predicting Gait Parameters of Leg Movement with sEMG and Accelerometer Using CatBoost Machine Learning”. This study investigated the use of surface ElectroMyoGraphy (sEMG) and accelerometer (ACC) data to estimate gait characteristics. These researchers employed a custom wireless multi-channel measurement system combined with GaitUp Physilog 5 inertial sensors to record the walking patterns and muscle activations of 17 participants, resulting in a dataset of 1452 samples. Key contributions of this study include: (i) the collection and analysis of sEMG and ACC signals to extract relevant features for gait analysis; (ii) the prediction of 17 temporospatial gait parameters using machine learning techniques, specifically Categorical Boosting (CatBoost), XGBoost, and DTs; and (iii) the implementation of feature selection methods to identify significant features that enhance prediction accuracy. The experimental results revealed that the CatBoost model achieved nearly equal performance with the XGBoost model and both outperformed compared to DTs. Specifically, the CatBoost Pearson correlation coefficient (PCC) for the left and right legs averaged 0.878 ± 0.169 and 0.921 ± 0.047, respectively, while it achieved a mean squared error (MSE) of 7.65, a root mean squared error (RMSE) of 1.48, a mean absolute error (MAE) of 1.00, a mean absolute percentage error (MAPE) of 0.03, and an R² score of 0.91. The XGBoost model demonstrated a PCC of 0.887 ± 0.174 for the left foot and 0.906 ± 0.052 for the right one, while achieving a mean squared error (MSE) of 7.81, a root mean squared error (RMSE) of 1.53, a mean absolute error (MAE) of 1.00, a mean absolute percentage error (MAPE) of 0.03, and an R² score of 0.81. This study emphasized the potential of integrating sEMG and ACC data with machine learning to provide a comprehensive and accurate method for gait analysis. This approach can significantly impact clinical settings, particularly in facilitating the diagnosis and monitoring of conditions like Parkinson’s disease and in enhancing the functionality of prosthetic limbs.

The purpose and inspiration behind this Special Issue were to make a broad yet timely addition to the current body of literature. It is anticipated that the valuable methodologies featured in this Special Issue will be considered beneficial, engaging, and acknowledged by both the industry and the scientific community. Our aim is that researchers will be motivated by the novel strategies presented, thereby advancing research across various multidisciplinary fields and encouraging further investigation in the area of e-health applications and medical information systems overall. Future efforts may include leveraging cutting-edge techniques and methodologies to enhance disease diagnosis and prognosis and improve user acceptance of digital healthcare services.