Reliable transmission and analysis of biomedical signals over wireless networks remain challenging due to packet loss, bandwidth limitations, and temporal distortions that can significantly impact artificial intelligence (AI) performance. This paper proposes an AI-driven, communication-aware framework for robust electroencephalography (EEG) monitoring in wireless telemedicine environments. The proposed system integrates lightweight Transformer-based temporal representation learning with a neuro-fuzzy decision support module to enhance resilience under realistic network impairments. EEG signals undergo pre-processing and sliding-window segmentation before being processed by a compact self-attention encoder designed for low-complexity edge deployment. To emulate practical communication scenarios, the training pipeline incorporates channelaware perturbation strategies, including temporal masking and simulated packet loss. The resulting temporal embeddings are analyzed using an Adaptive Neuro-Fuzzy Inference System (ANFIS) to provide interpretable and stable classification decisions. Experimental evaluations demonstrate improved robustness compared to conventional ANN and CNN approaches, particularly under degraded wireless conditions. The proposed framework highlights the effectiveness of communication-aware AI architectures for dependable EEG monitoring across next-generation wireless telemedicine networks.

Cardiovascular diseases remain a leading global cause of mortality, highlighting the need for accurate and scalable electrocardiogram (ECG) analysis systems that can support clinicians without increasing interpretation burden. This paper presents an inter-patient ECG arrhythmia classification framework based on a hybrid neuro-symbolic architecture. The system is evaluated under the clinically realistic AAMI DS1/DS2 inter-patient protocol using the MIT-BIH Arrhythmia Database, while additional beats from the MIT-BIH Supraventricular Arrhythmia Database (SVDB) and the St. Petersburg INCART database are incorporated during training to reduce class imbalance and improve robustness. The proposed framework combines beat-centered segmentation and z-score normalization, a lightweight 1D CNN–BiGRU backbone, temporal RR-interval context features, augmentation-based oversampling for minority classes, and a neuro-symbolic arbitration module that integrates neural outputs with physiologically motivated rules. Experimental results show that temporal context substantially improves supraventricular beat detection, while the neuro-symbolic stage improves overall decision consistency and interpretability. Statistical validation using bootstrap confidence intervals, McNemar’s test, and Cochran’s Q test confirms that the observed gains are significant. The framework is lightweight, interpretable, and suitable for real-time ECG monitoring and human-in-the-loop clinical decision support.

Obstructive Sleep Apnea (OSA) is a prevalent sleep disorder defined by repeated disruptions in breathing while sleeping. OSA leads to numerous complications, including poor sleep quality. Polysomnography (PSG) is the gold standard for OSA diagnosis; however, it is also costly, time-consuming, and performed in a clinical setting. In this paper, we provide a portable, inexpensive and light system for OSA detection based on Electrocardiogram (ECG) signals of the patient, representing an accessible alternative to PSG. The proposed method involves a wearable belt embedded with a wireless ECG sensor that captures real-time heart activity. The ECG signals are communicated via Bluetooth, visualized live if desired, and subsequently stored in the Firebase database. The ECG data are processed using an AI based model for automatic detection of apneas. The model is a hybrid CNN-LSTM approach, using no feature extraction, just raw ECG data. CNN creates spatial features from time steps in the ECG signals, while the LSTM learns temporal dependencies in the time steps. The model uses the PhysioNet Apnea-ECG dataset for training, validation and testing. The model obtained exceptional classification results: $\mathbf{9 9. 6 \%}$ Accuracy; 99.4% Sensitivity; and 99.8% Specificity. In addition to accuracy evaluation, real time testing was performed for home-based monitoring purposes in further validation. The overall findings offer evidence that deep learning applied to ECG signals is informative enough to identify an apnea event and act as quantitative. The device draws upon all the advantages associated with smart, wearable technology while providing safety in human health monitoring.

Heart disease remains the leading cause of mortality globally, and early diagnosis is critical. Electrocardiography (ECG) is a primary diagnostic tool for heart conditions, and recent advances in machine learning (ML) and deep learning (DL) have greatly improved automated ECG-based heart disease classification. This review provides a comprehensive overview of state-of-the-art ML and DL techniques for classifying heart diseases from ECG signals, with focus on arrhythmia and myocardial infarction (MI) detection. A summary of ECG fundamentals, commonly used datasets, and preprocessing methods for noise reduction and feature extraction are provided. Traditional ML models (SVM, k-NN, random forests) and their performance are compared with DL approaches (including convolutional neural networks CNNs, recurrent networks like LSTM/GRU). Recent innovations such as attention mechanisms, transformer-based models, and hybrid architectures (CNN combined with RNN or Transformer) are highlighted for their superior ability to capture temporal and morphological patterns in ECG data. A discussion on the evaluation metrics and the typical performance achieved on benchmark datasets (MIT-BIH arrhythmia, PTB-XL) by various methods is reported. Finally, a comparative discussion, outlining current challenges including data imbalance, noise, interpretability, and generalization is provided, and the future research directions such as more robust hybrid models, attention-driven architectures, and improved generalization via transfer learning are highlighted. The findings indicate that DL approaches, especially CNN-based and hybrid models with attention, currently achieve the highest accuracy (often less 98% on arrhythmia classification), while ongoing improvements aim to enhance reliability and clinical interpretability for real-world deployment.

Artificial Intelligence (AI) and machine learning analyze strabismus via Electro-oculography (EOG) signals, addressing eye muscle irregularities. Detecting this disorder's manifestation in both eyes is challenging, impacting the vision and the 3-dimensional perception. To this purpose, the aim is to develop an ophthalmologist-assisting system that reduces costly interventions like eyewear, lasers, and drugs. EOG signals from electrodes on the face reveal insights, focusing on the Center of Mass (CoM) of the ocular movements. The CoM determines the ocular misalignment extent due to muscle irregularities. Using the same device, ophthalmologists guide patient recovery through gamified exercises. Results highlighted the system's efficacy in differentiating between normal and aberrant conditions, quantifying deviation, and treating strabismus. In essence, AI and machine learning with EOG signal analysis offer a transformative solution for strabismus, providing early detection, accurate classification, and tailored rehabilitation for this ocular condition. The results of this study also suggested that the K-Nearest Neighbors (KNN) and wavelet decomposition tree methods may be effective for the classification and feature extraction of EOG data in the detection and diagnosis of eye diseases. These methods showed high levels of accuracy, with 98.3 and 98.5% respectively. However, further research is needed to confirm and extend these findings, and to explore the potential of these methods in the context of EOG and eye disease detection.

This paper presents a new methodology for developing a low-cost wireless ECG transmission and monitoring system based on IoT technology, designed for real-time detection and classification of heart diseases. The study focuses on using ECG data for heart disease classification, which is an area of growing interest in recent years. The study collected data from 1000 subjects using our designed system to collect the normal data (300 patients) and a Biopac MP160 data acquisition system for the collection of 10 diseases abnormal data, where all the data are acquired from lead 1. The aim of this study is to develop an accurate and reliable classification model for heart diseases using ECG data. Pre-processing steps were taken to prepare the data for feature extraction, including the use of Empirical Mode Decomposition (EMD) and digital filters such as low pass, high pass, and derivative pass filters. A new feature extraction steps based on a new ECG peak detection, segmentation, and wave modeling for each segment is also presented. Two classification methods were used: Multi-Layer Perceptron (MLP) and Radial Basis Function (RBF). The results showed that MLP had a much higher accuracy of 99.1% compared to RBF, which had an accuracy of 97.4%. The study emphasizes the potential of using ECG data for accurate classification of heart diseases. The results demonstrate that proper pre-processing and feature extraction techniques are crucial for improving accuracy. This study is significant for remote patient monitoring and telemedicine applications, as it provides a low-cost, non-invasive method for detecting and classifying heart diseases using ECG data.

Development of Brain Computer Interface (BCI) has been rapid since the mid 1990's. There are three criteria for BCI, (i) comfortability and possession of a suitable signal acquisition device, (ii) system validation and dissemination, and (iii) reliability and potentiality. As there are no BCI possessing the optimal criteria, it was essential to consider building a new one. Thereby, the paper investigates building BCI based on the utilization of EEG signals to translate brainwave patterns into actionable commands. The primary objective is to enhance communication capabilities for individuals afflicted with neurological disorders, empowering them to command external devices and engage more effectively with their surroundings. We built our model on EEG online dataset for the purpose of feature extraction and classification. Statistical features and Discrete Wavelet Transform (DWT) have been applied for feature selection. Multi-Layer Perceptron (MLP) and Radial Basis Function (RBF) were the classifiers involved. Results showed that the proposed architecture of MLP and RBF were able to classify the EEG signals into two classes (open eye and closed eye). Results also showed that the proposed approach, which is based on the combination of statistical features and DWT for features selection using AF3 and AF4 channels by the application of MLP, has 98% succession rate. BCI system based on Arduino circuit has been built after the classification Further algorithms and system evaluation need to be considered as future work.

Osteoporosis is a common disease characterized by low bone density and structural deterioration of bone tissue. For a successful course of treatment and fracture avoidance, early diagnosis of this disease is essential. The aim was to provide a novel method for osteoporosis prediction using Artificial Intelligence (AI)-based framework. The purpose was to predict the likelihood of osteoporosis based on the Bone Mineral Density (BMD), along with other characteristics such as age, weight, height, gender, and Body Mass Index (BMI) extracted from medical reports and images collected from a comprehensive medical center in Lebanon. Three machine-learning algorithms were implemented and tested, Logistic Regression (LR), Support Vector Machines (SVM), and Decision Trees (DT). Variety of quantitative statistical metrics were used to evaluate the performance of our framework, upon training and testing our algorithms. The metrics that were employed to evaluate our results included accuracy, precision, sensitivity, and F-score, in addition to the Receiver Operating Characteristic (ROC) Curve and the Area Under the Curve (AUC). Experimental Results demonstrated that both the SVM and LR algorithms achieved the highest accuracy of detection of osteoporosis as compared to existing algorithms applied in this field, with an accuracy of 89%. The sensitivity of diagnosis obtained was 98% by LR and 97% by SVM and surpassed the sensitivity obtained by DT. As such LR showed the best performance. The output of the algorithms could help medical doctors assess patients automatically. These findings demonstrated the potential of AI in osteoporosis prediction and thus prevention, highlighting the significance of early diagnosis. Thereby as a future prospect, choosing carefully the framework is crucial and additional algorithms have to be considered and tested.

System of Systems (SoS) engineering is an essential concept in engineering education. It provides students with the skills and knowledge to design, build, and manage complex systems that involve multiple interdependent subsystems. By incorporating the concept of SoS into the engineering curriculum, engineering educators can better prepare students to address the challenges of the 21st century. Improvement of teaching methodologies and technologies plays an important role in the enhancement of engineering education. In this paper, we will discuss and focus on how to improve the engineering education program as a larger system, based on the relation between all the constituted elements (subsystems) of the engineering curriculum, and how their work together for its development and enhancement. Subsystems of the engineering education program are divided into three main parts: course design, instructor role model, and modern teaching methodologies. To achieve our goal, we will discuss and review, the design of an engineering course based on Bloom's taxonomy technique, the role of the instructor/educator as the best role model for the students, and the different modern teaching methodologies in higher education. This paper is a theoretical study concerning the engineering education program enhancement. As results and findings, we present the challenges and difficulties that the professors and engineering teachers should be aware of to develop and support the engineering education system.

The Islamic University of Lebanon (IUL) is committed to providing faith and knowledge as a source of inspiration for the Lebanese nation, citizen, state and society as a whole. IUL has paid special attention to the environment and to the green metrics when establishing the new campus in Wardanieh, the rules related to green buildings and the preservation of the environment were taken into account. The COVID-19 pandemic has created the largest disruption of education systems in human history, affecting nearly 1.6 billion learners in more than 200 countries. In this paper we will describe how the setting and infrastructure of Wardanieh campus has helped in managing the Covid-19 crisis at the university through different levels: the physical distancing, the effect of the large campus buildings area according to the campus population, moreover the large forest spaces, the natural ventilation of buildings, the sanitization and sterilization procedures with setting guidelines for Covid-19 such as the obligation of wearing masks. Furthermore, we will describe the influence of ICT technologies into education in many different ways for distance learning, and how the university has evolved its ICT infrastructure to support the increase in demand on the internet capacity and on university servers.Keyword: IUL university, Green, SDGs, Covid-19 prevention, education, settings and infrastructure

Biometric system has gained more importance in providing high security in individual identification as it uses a network of blood vessels underneath the palm skin. This paper proposes a new algorithm for palm vein identification using a histogram of gradient t(HOG). Raw images of palm hand vein are taken from a public dataset named VP base dataset. Region of interest was extracted after preprocessing stage, after that essential features were extracted by following different steps of the HOG algorithm, which captures edge information inside the images, where they are the vital features containing valuable information. For the purpose of classification Support Vector Machine was used. By experimental work, our algorithm gives promising results than different existing feature descriptors.

Premature Ventricular Contractions (PVCs), a common type of cardiac arrhythmia, can be identified by analyzing electrocardiogram (ECG) signals. If not treated on time, PVCs become life-threatening. In this paper, a high-performance approach is proposed for detecting PVCs in an unsupervised manner. The main objective is to perform an automatic PVCs detection in ECG without prior knowledge. Ten different statistical features are extracted to represent various characteristics of the signal. Thereafter, the proposed approach explores PVCs detection by two different strategies. Performance evaluation results over the MIT-BIH Arrhythmia Database (MIT-BIH-AD) show that the strategy based on Agglomerative Hierarchical Clustering (AHC) Method outperforms K-means Clustering Method with an average Accuracy (ACC), Specificity (SPE), Sensitivity (SEN), and Positive Predictive Value (PPV) of 98.43%, 99.23%, 94.47%, and 96.67%, respectively. With less complexity and computation load, AHC can be an accurate candidate for PVCs detection to be used in clinical applications.

Nowadays, the accurate analysis of energy consumption has become vital for the development of efficient energy projects as well as, for demonstrating the consumptive behavior of the energy consumers in the system. The importance of this analysis comes from many reasons, one of them is that it leads to a better understanding of the system components. This paper presents a clustering algorithm for residential energy consumption using the K-Means algorithm in two different approaches. The dataset utilized in this article contains energy consumption features selected from 25 houses over a period of two years. Firstly, data cleaning has been used to remove and eliminate the inconsistent data, secondly the Elbow method has been applied to determine the optimal number of clusters before using the K-means approach for the purpose of clustering. In K-means, the data have been clustered into two different approaches. The first one is clustering the daily mean consumption in each season in each year. The second one is clustering the monthly mean consumption over the two years. Finally, data visualization has been applied in order to present the result of our proposed method. The paper finds that the households have different consumption behaviors in different seasons, days, and months and that it is due to the change of the average temperature in each season as well as the different appliances and consumptive patters of each house. The results are representative and match the aim of the paper. Further, they are significant for the further development of the energy system and efficient for tracking the consumption of the houses. Finally, the results of this paper are going to be used after running the algorithm again with a different number of clusters to compare the results and find new insights in the data that might affect the decision.

In the light of the rapidly growing COVID-19 pandemic, the need for an expeditious diagnosis of COVID-19 infection became essential. The immediate diagnosis will allow the initiation of the isolation process and adequate treatment as well. While the standard test used for the diagnosis of COVID-19 disease (RT-PCR) is usually time consuming (6 hours up to days in some centers); the need for a highly sensitive test became essential. Many studies have illustrated the utility of chest CT scan in the diagnoses of COVID-19. This paper evaluates the value of classical machine learning techniques and the convolutional neural networks in aiding physicians to further classify patients into either COVID-19 positive or negative according to their chest CT findings, and thus facilitating their work. To address this problem, this paper proposes classical neural networks using statistical features and deep CNN models to further classify a dataset of preprocessed chest CT images, using several classifiers and to evaluate the results. This latter showed that the best proposed method was a four layers CNN with SVM classifier with 99.6% accuracy. This demonstrates the potential of the proposed technique in computer-aided diagnosis for healthcare applications, especially for COVID-19 classification.

Raman spectral imaging is applied to human skin tissue fixed with formalin and embedded in paraffin. However, some restrictions may appear due to the high intensity of the paraffin signal. Extended Multiplicative Signal Correction (EMSC) is applied to correct the paraffin contribution of the Raman spectral image. For this purpose, the paraffin signal must be modeled as precisely as possible. In this paper, we propose to apply the matrix decomposition method such as Principal Component Analysis (PCA) to model the components of pure paraffin in the digital dewaxing technique. Then, the corrected spectra obtained using EMSC and PCA are classified using K-means, and compared with spectra obtained using classical EMSC result.

Speech Recognition Software is a computer program that is trained to take the input of human speech, interpret it, and transcribe it into text. Most recently, the field has benefited from advances in deep learning and big data. The advances are evidenced not only by the surge of academic papers published in the field, but more importantly by the worldwide industry adoption of a variety of deep learning methods in designing and deploying speech recognition systems. The objective of this paper is to propose an advanced and accurate end-user software system that is able to recognize specific commands to control a robot to perform specified tasks in a hospital. This model will be based on Deep Learning since it is effective in models having huge data as for the two versions of Google TensorFlow and AIY datasets used in our model. Convolutional neural network will be used since it is able to extract features from the dataset instead of traditional methods of feature extraction, thus saving training time and reducing the complexity of the system. With addition to that, NVIDIA CUDA will be also used to train the model with GPU to decrease the training time. During training, some experiments have been done to see the effect of some parameters on the results of the system, and to make sure that the chosen parameters in our model are the best. The results indicate that the training, validation, and testing accuracies of the proposed approach were high, the training duration reached very low values due to the innovation used (CUDA Toolkit) and the commands were successfully recognized by the model. These results outcome the results of the papers that developed similar work which will be presented in the coming sections.

Abstract—Information decision support systems are becomingmore in use as we are living in the era of digital data andrise of artificial intelligence. Heart disease as one of the mostknown and dangerous is getting very important attention, thisattention is translated into digital and prediction system thatdetects the presence of disease according to the available dataand information. Such systems faced a lot of problems since thefirst rise, but now with the deveolopment of machine learnigfield we are using them in developing new models to detect thepresence of this disease, in addition to algorithms data is veryimportant which also form the heart of the predicton systems,as we know prediction algorithms take decisions and thesedecisions must be based on facts, and these facts are extractedfrom data, as a result data is the starting point of every system.In this paper we propose a Heart Disease Prediction Systemusing Machine Learning Algorithms, in terms of data we usedCleveland dataset, this dataset is normalized then divided intothree scnearios in terms of traning and testing respectively,80%-20%, 50%-50%, 30%-70%. In each case of dataset ifit is normalized or not we will have these three scenarios.We used three machine learning algorithms for every scenarioof the mentioned before which are SVM, SMO and MLP, inthese algorithms we’ve used two different kernels to test theresults upon that. These two types of simulation are added tothe collection of scenarios mentioned above to become as thefollowing we have at the main level two types normalized andunnormalized dataset, then for each one we have three typesaccording to the amount of training and testing dataset, thenfor each of these scenarios we have two scenarios according tothe type of kernel to become 30 scenarios in total, our proposedsystem have shown a dominance in terms of accuracy over theother previous works.

This paper presents a new concept for electrocardiogram (ECG) signals compression based on Fourier series modeling. The goal of the compression is to enable the ECG Holter to record and store ECG data for several days instead of just 24 hours while maintaining all the features of the signals. This proposed method can be used to record up to 26 days when using Fourier series of 4th degree, and 21 days when using Fourier series of 5th degree, whith high accuracy and a mean square error (RMSE) of approximately 0.001, which is considered extremely low and satisfactory.

For a long time, the major problem area for researchers is disease diagnosis and the main interest of the medicine is an accurate diagnosis. Many engineering techniques have been developed in the past to help the medical staff with a diagnosis tool. There are many traditional methods of disease diagnosis, but the application of machine learning techniques has given a new dimension to this area. In this work, two different approaches have been used for the purpose of classification between diabetic and non-diabetic, using Pima Indian Diabetes Dataset. Principal Component Analysis has been used in the purpose of feature dimension reduction before applying any proposed classifier. Support Vector Machine (SVM) and Naïve Bayes (NB) are the two classifiers used in our study. 94.14 % and 93.88% are the accuracies obtained for the SVM and NB approaches.The results obtained are very interesting and show improvement from the previous works. With this accurate learning technique, there is enough scope for improvement considerably in this field.