We present here the interim results from a large prospective, multicentre, open-label, non-randomized clinical trial evaluating the breast cancer detection capabilities of MammoWave microwave imaging device on an asymptomatic population across five European countries. This study will conduct a comparative analysis between reference standard, defined as the outcome of the conventional breast examination pathway, and MammoWave’s AI model classifications. The reference standard will be classified as ‘positive’ for histology confirmed breast cancer, and ‘negative’ otherwise. The algorithm will assign each breast to one of two categories: ‘With suspicious finding’ or ‘No suspicious finding’. We report results from the first 3,000 volunteers enrolled in nine hospitals. This interim analysis allowed us to evaluate the performance of the AI model, with a sensitivity of 42% and a specificity of 75%. Subsequently, as indicated in the clinical protocol, we updated the AI model by fusing the information from machine learning and statistical models. The updated AI model allowed us to retrospectively reach a sensitivity of 67% (80% in dense breast) and a specificity of 81% (80% in dense breast).

Remote patient monitoring has become a vital component in healthcare services. It facilitates continuous and unobtrusive patient observations outside a traditional clinical setup. While there are numerous sensing technologies utilised for remote monitoring, ultrawideband (UWB) radar-based systems are gaining significant popularity due to their higher spatiotemporal resolution, non-disruptive and safe operation, and robustness towards background noise. However, these monitoring systems require proper infrastructure to ensure connectivity, efficiency, accuracy, reliability and security for remote monitoring. 5G/6G technologies provide infrastructures, facilitating low latency, massive connectivity, enhanced bandwidth, and network slicing capabilities, making them best suited for demanding remote monitoring applications. This paper proposes a novel non-contact UWB remote monitoring system designed for 5G and 6G networks. It comprises distributed UWB radar nodes for monitoring, cellular interfaces for communication, and CPU and GPU clusters for artificial intelligence algorithms. The proposed architecture is implemented on a digital-twin framework of an open radio access network, 5G core, software-defined radios and CPU and GPU clusters for real-time processing. Moreover, a generative adversarial network is implemented in GPU clusters to reconstruct missing radar frames. Experimental results demonstrate that the system achieved a consistent end-to-end synchronised radar frame rate of 15 fps , ensuring smooth and reliable operation. The findings indicate that the proposed 5G/6G-enabled UWB radar system has significant potential for long-term remote gait monitoring in real-world environments.

Photovoltaic (PV) systems are widely deployed renewable sources, yet their performance and reliability are often undermined by operational faults. This paper presents a novel weather-driven fault classification framework for PV predictive maintenance (PdM), addressing a key limitation of existing studies that analyse faults under fixed, short-term or simplified weather assumptions. The major contribution of this work lies in explicitly modelling variability in both solar irradiance and temperature across four distinct weather conditions, enabling the separation of weather-induced variations from true fault signatures. Features are extracted from IV characteristics under varying weather scenarios and processed through a regression-based residual analysis with a pre-feature assessment stage, generating deviation patterns and feature contribution signatures. These enhanced feature vectors are then classified using a Bayesian-optimised gradient boosting model within a two-stage architecture that (1) distinguishes normal vs. abnormal operation and (2) identifies either the prevailing weather condition (if normal) or classifies the fault type together with its corresponding weather scenario (if abnormal). The robustness and generalisability of the framework are validated using cross-validation, high-fidelity simulations, and independent laboratory experiments, achieving 99% classification accuracy. By directly linking each detected fault to its prevailing weather condition, the proposed approach provides a robust pathway for weather-aware PdM scheduling, improving diagnostic reliability and operational resilience across diverse environments.

Photovoltaic (PV) energy is considered one of the most widespread renewable sources. However, faults frequently compromise their performance, leading to reduced energy efficiency and posing risks to system reliability. While extensive research has focused on fault detection and classification, the quantitative evaluation of fault severity has received relatively little attention. Assessing fault severity is essential for predictive maintenance because it not only identifies faulty conditions but also reveals the degree of performance deterioration, enabling maintenance decisions to be both timely and optimised. This paper presents a Machine learning-based framework dedicated to the severity assessment of PV array faults. Multiple fault scenarios are systematically reproduced through a detailed Simulink model and a laboratory-scale PV array setup, generating diverse datasets for model training and validation. Extensive feature engineering is carried out to extract degradation indicators, enabling the differentiation of faults across multiple severity levels. Gradient boosting served as the primary technique for severity classification, supported by independent runs, grid search optimisation and hyperparameter sensitivity analysis, Probability Density Function, evaluation under measurement noise and cross-validation to ensure accuracy and robustness. The proposed methodology goes beyond traditional fault diagnosis by quantitatively assessing the severity of PV array faults, thus bridging the gap between fault detection and actionable maintenance strategies. Experimental and simulation results confirm that the approach can effectively map faults to their severity levels, providing critical input for predictive maintenance scheduling.

The use of microwave imaging techniques for detection and classification of brain strokes is a growing field of research, motivated by the distinct dielectric properties of hemorrhagic and ischemic strokes relative to their surrounding tissues. In this work we combine our image processing algorithm based on the Huygens' principle (HP) and U-net based deep learning to produce morphological reconstructed images, allowing detection and classification of brain strokes.

Microwave imaging has emerged as a promising alternative due to its non-ionizing, portable, and cost-effective nature. However, reconstructing high-quality brain images from microwave signals remains challenging due to the complexity of brain tissues. In this study, we propose a new brain imaging framework integrating microwave imaging with a diffusion model to generate high-quality morphological brain images. Our approach employs a Huygens’ principle (HP)-based imaging technique to capture brain tissue characteristics, followed by a two-stage deep learning pipeline: a U-Net model for initial image reconstruction and a diffusion model for high-fidelity image enhancement. Moreover, we have demonstrated the process of generating synthetic images for microwave imaging and effectively utilizing them in a diffusion model. Our results demonstrate that the proposed method significantly improves image quality.

Gait, the way people move, provides insights into assessing activity levels, identifying neuromuscular disorders, and predicting falls. Although wearable sensors and camera-based systems are employed to measure gait, they face challenges such as calibration issues, motion artefacts, and environmental constraints. This research presents a novel ultra-wideband (UWB) radar system to capture multiple views of the gait and an algorithm to process the radar signals for generating Detection Intensity Level (DIL) maps, which represent motion in three-dimensional space. These DIL maps are processed to estimate walking speed with 94% accuracy, cadence with 92% accuracy, and step length with accuracy of 92%. Therefore, the proposed system demonstrates the potential for long-term remote gait monitoring in real world environments.

Cervical cancer remains a significant global public health challenge, affecting over half a million women annually, with a mortality rate of approximately 60%, especially in resource-limited regions. This study presents an advanced methodology for cervical cancer diagnosis through deep learning techniques. Utilizing a publicly available cervical cancer image dataset, the research introduces a novel classification framework that integrates a Neural Feature Extractor (NFE) based on a pre-trained VGG16 architecture and an AutoInt model for automatic feature interaction learning. The extracted features are processed through machine learning classifiers such as KNN, LGBM, Extra Trees, and others for classification tasks. Among these classifiers, KNN achieved the highest accuracy of 99.96%, followed closely by LGBM at 99.92%. This study also assesses the computational complexity of various models, demonstrating that simpler models like LDA exhibit faster prediction times, while more complex models, such as KNN and LGBM, provide higher accuracy. These findings highlight the potential of deep learning frameworks in improving cervical cancer classification accuracy, especially in resource-limited environments.

Here, we propose an analytical approach to simulating MammoWave, a novel apparatus for breast cancer detection using microwave imaging. The approach is built upon the theory of cylindrical waves emitted by line sources. The sample is modelled as a cylinder with an inclusion. Our results indicate that when compared with phantom measurements, our approach gives an average relative error (between the image generated through measurement with phantoms and the image generated through the analytical simulation approach) of less than 6% when considering the full frequency band of 1-9 GHz. The procedure permits the simulation of the MammoWave imaging system loaded with multilayered eccentric cylinders; thus, it can be used to obtain an insight into MammoWave's detection capability, without having to perform either time-consuming full-wave simulations or phantom measurements.

Power networks are vital to society, yet service outages and faults can have devastating consequences. This study introduces a novel integration of machine learning and data augmentation techniques for fault detection and classification, addressing gaps in data diversity and imbalance. Unlike traditional approaches, the research utilizes an Auxiliary Classifier Generative Adversarial Network (ACGAN) to generate synthetic data representative of underrepresented fault types, enhancing model training and performance. By extracting both spectral and statistical features from the Grid Event Signature Library (GESL) dataset, a comprehensive representation of power system signals is achieved. A comparative evaluation of models including Decision Trees (DT), Random Forest (RF), Extra Tree Classifier (ETC), Gradient Boosting Classifier (GBC), and K-Nearest Neighbors (KNN) revealed the Extra Tree Classifier achieved the highest testing accuracy of 93.85%. The methodologies demonstrated scalability by using a dataset augmented to 9,000 samples and validated robustness through 10-fold cross-validation with a standard deviation of 0.00659. These results highlight the proposed framework’s potential for real-world implementation in modern power grids, offering enhanced fault prediction and resilience. This research establishes a pathway for integrating advanced data augmentation and machine learning techniques into operational power grid systems, ensuring stability and reliability.

Gender classification plays a vital role in various applications, particularly in security and healthcare. While several biometric methods such as facial recognition, voice analysis, activity monitoring, and gait recognition are commonly used, their accuracy and reliability often suffer due to challenges like body part occlusion, high computational costs, and recognition errors. This study investigates gender classification using gait data captured by Ultra-Wideband radar, offering a non-intrusive and occlusion-resilient alternative to traditional biometric methods. A dataset comprising 163 participants was collected, and the radar signals underwent preprocessing, including clutter suppression and peak detection, to isolate meaningful gait cycles. Spectral features extracted from these cycles were transformed using a novel integration of Feedforward Artificial Neural Networks and Random Forests , enhancing discriminative power. Among the models evaluated, the Random Forest classifier demonstrated superior performance, achieving 94.68% accuracy and a cross-validation score of 0.93. The study highlights the effectiveness of Ultra-wideband radar and the proposed transformation framework in advancing robust gender classification.

Variations in knee angles are important parameters used for clinical gait analysis. They are measured using gold-standard marker-based motion capture systems in clinical research. However, these systems are difficult to operate for longer durations and are limited to laboratory environments. Although camera-based approaches are available, they require a direct line of sight, adequate ambient lighting, and a clear background. Ultra-wideband radar technology (UWB) has the potential to address these limitations and support moving away from strict laboratory settings. This paper presents a UWB radar-based gait analysis system to capture walking gait, a novel algorithm, a unique detection intensity level algorithm (DILA), registering detections in a 3D space, and a PointNet-inspired neural network (DIL-PointNet) for knee angle estimation. The proposed system is validated against knee angles calculated with the Qualisys marker-based motion capture system. It achieved a mean absolute error (MAE) of 7.9° and a root mean squared error (RMSE) of 11.1°. Thus, the proposed system has the potential for long-term remote gait monitoring in real world environments.

Peripheral Sensory Neuropathy (PSN) affects a large proportion of individuals suffering from type 2 diabetes. To avoid ulceration and other damage to the patient’s feet, regular PSN testing, and assessment must be undertaken. Currently, the Semmes-Weinstein Monofilament Examination (SWME) is one of the most widely accepted techniques for PSN assessment. This process is time-consuming, requires special training, and is prone to errors. The number of type 2 diabetes sufferers globally is growing at alarming rates with healthcare workers under enormous pressure to continue to provide one-to-one regular care. In order to reduce the burden on existing services whilst providing the necessary care to patients, automated approaches for PSN detection provide many advantages. Importantly, with respect to an automated SWME method, there will be areas on the plantar surface where the SWM probe should not be applied i.e., areas with lesions or suspect regions. The research presented in this manuscript conducted a comprehensive analysis of different feature sets and classifiers for the task of lesion classification. Three distinct feature sets Local Binary Pattern (LBP), Mel Frequency Cepstral Coefficients (MFCC), and Scale-Invariant Feature Transform (SIFT) were evaluated across various classifiers, including Support Vector Machine (SVM), Multi-layer Perceptron (MLP), Random Forest (RF), Naïve Bayes (NB), and XGBoost. The results revealed nuanced performances across the combinations of feature sets and classifiers. While each feature set demonstrated strengths, the NB classifier applied to the LBP feature set emerged as the most notable performer with an accuracy score of 100%. This combination achieved perfect accuracy, precision, recall, and F1-score metrics, showcasing its robustness in accurately classifying lesion instances. The 5-fold cross-validation results underscored the stability of NB on the LBP feature set, with a negligible standard deviation, affirming its consistent performance across different data subsets. Additionally, the computational time complexity of 0.91 seconds highlighted its efficiency, making NB on the LBP feature set a practical and reliable choice for real-world applications. Statistical analysis using the one-way ANOVA test revealed significant differences in classifier performance across feature sets, with MFCC resulting in significantly lower accuracy compared to LBP and SIFT, which showed similar performance. The Tukey HSD post-hoc test confirmed these findings, highlighting the crucial role of feature set selection in classifier effectiveness.

Marker-based motion-capturing technologies are widely used in clinics to diagnose motor-related pathologies due to their high resolution and accuracy. However, it often requires manual intervention to process the raw marker data. Although previous research has proposed algorithms to automate these processes, they do not address different marker placement models, abnormal gait patterns, or variations in the anthropometric measurements which limits their scalability. Therefore, this research proposes a novel automated algorithm to process the raw marker data and generate a novel 6D skeleton representation. It is used in machine learning classifiers to identify abnormal gait patterns. The proposed algorithm was tested with marker-based gait analysis data and achieved 99.7% accuracy in classifying normal and abnormal gait patterns using multilayer perceptron classifiers.

In this paper, we present preliminary outcomes derived from a prospective multicentric clinical trial focusing on microwave breast imaging involving 336 women. This investigation was conducted as part of the RadioSpin project. Our primary objective was to evaluate the prospective performance of MammoWave, our microwave imaging device, in differentiating between breasts with and without radiological findings, utilizing specific microwave images features' thresholds. Beyond this primary assessment, we explored the use of individual frequency sub-bands to categorize breasts into two groups: healthy (without findings or with benign findings) and non-healthy (malignant findings), drawing on features identified in a previous clinical trial. Our findings reveal a sensitivity of 72% in detecting radiological findings and a noteworthy 78% in identifying non-healthy (cancerous) breasts.

This paper presents the Modified Flower Pollination Algorithm-based Multi-Layer Perceptron Neural Network (MFPA-MLPNN) as an optimization technique for efficient power flow management in a Smart Building Microgrid (SBMG) in-tegrated with solar and wind generation, and Electric Vehicle Batteries (EVBs) within grid connected structure while concurrently reducing optimization processing time. To achieve both technical and economic superiority, two optimization objectives are addressed. Firstly, a Demand Response (DR) framework is harnessed to accommodate the stochastic behavior and forecasting errors associated with intermittent sources. Secondly, the degradation of EVBs is considered, ensuring an economically viable power flow proposed strategy for both EV owners and mi-crogrid (MG) authorities. Power generation of Variable Renew-able Energy Sources (VRES) has been forecasted using MLPNN. Battery degradation and system stability under the action of the proposed topology have been evaluated using a simulation-based environment. Results show a significant decrease in battery degradation and processing time using the proposed MFPA-MLPNN optimization architecture.

The number of patients with blindness and partial sightedness has grown over time. These vision impairments significantly affect the quality of life, and navigation is one of the main challenges these patients face. Although assistive devices were introduced in the past, they have a low acceptance rate due to their limited usability, cost, portability, battery life and higher cost. This paper proposes an on-body area network based on ultra-wideband (UWB) technology and a novel algorithm to detect and classify the obstacles. The on-body radar network captures the backscattered UWB pulses and publishes them to an MQTT network at a rate of 5 frames per second. Then the algorithms unit obtains the UWB radar frame and generates a detection image. This image presents the position and features of the obstacles. At the final stage, detection images are fed to machine learning classifiers to identify the obstacles. The proposal system was experimentally validated and obtained a classification accuracy of 93

Driving while drowsy poses significant risks, including reduced cognitive function and the potential for accidents, which can lead to severe consequences such as trauma, economic losses, injuries, or death. The use of artificial intelligence can enable effective detection of driver drowsiness, helping to prevent accidents and enhance driver performance. This research aims to address the crucial need for real-time and accurate drowsiness detection to mitigate the impact of fatigue-related accidents. Leveraging ultra-wideband radar data collected over five minutes, the dataset was segmented into one-minute chunks and transformed into grayscale images. Spatial features are retrieved from the images using a two-dimensional Convolutional Neural Network. Following that, these features were used to train and test multiple machine learning classifiers. The ensemble classifier RF-XGB-SVM, which combines Random Forest, XGBoost, and Support Vector Machine using a hard voting criterion, performed admirably with an accuracy of 96.6%. Additionally, the proposed approach was validated with a robust k-fold score of 97% and a standard deviation of 0.018, demonstrating significant results. The dataset is augmented using Generative Adversarial Networks, resulting in improved accuracies for all models. Among them, the RF-XGB-SVM model outperformed the rest with an accuracy score of 99.58%.

The identification of individuals based on their walking patterns, also known as gait recognition, has garnered considerable interest as a biometric trait. The use of gait patterns for gender classification has emerged as a significant research domain with diverse applications across multiple fields. The present investigation centers on the classification of gender based on gait utilizing data from Ultra-wide band radar. A total of 181 participants were included in the study, and data was gathered using Ultra-wide band radar technology. This study investigates various preprocessing techniques, feature extraction methods, and dimensionality reduction approaches to efficiently process Ultra-wide band radar data. The data quality is improved through the utilization of a two-pulse canceller and discrete wavelet transform. The hybrid feature dataset is generated through the creation of gray-level co-occurrence matrices and subsequent extraction of statistical features. Principal Component Analysis is utilized for dimensionality reduction, and prediction probabilities are incorporated as features for classification optimization. The present study employs k-fold cross-validation to train and assess machine learning classifiers, Decision Tree, Random Forest, Support Vector Machine, Logistic Regression, Multi-Layer Perceptron, K-Nearest Neighbors, and Extra Tree Classifier. The Multilayer Perceptron exhibits superior performance, achieving an accuracy of 0.936. The Support Vector Machine and k-Nearest Neighbors classifiers closely trail behind, both achieving an accuracy of 0.934. This research is of the utmost importance due to its capacity to offer solutions to crucial problems in multiple domains. The findings indicate that the utilization of UWB radar data for gait-based gender classification holds promise in diverse domains, including biometrics, surveillance, and healthcare. The present study makes a valuable contribution to the progress of gender classification systems that rely on gait patterns.

As an effective dimensionality reduction method, Same Degree Distribution (SDD) has been demonstrated to be able to maintain better data structure than other dimensionality reduction methods, including Principal Component Analysis (PCA), Multidimensional Scaling (MDS), Isomap, Locally Linear Embedding (LLE), Laplacian Eigen-maps (LE), Uniform Manifold Approximation and Projection (UMAP) and t-Stochastic Neighbour Embedding (t-SNE). In addition, SDD does not require tuning the number of neighbours or perplexity to scale the structure capturing performance. Instead, it requires tuning the degree of degree-distribution ranging in e certain interval. Hence, tuning the degree of degree-distribution makes SDD a less costly method than other methods that require tuning the number of neighbours or perplexity. Although these advantages, SDD is still an expensive method compared with parameter-free methods such as PCA and MDS. A parameter-free SDD is proposed based on standard SDD, with two main differences: 1) it does not require tuning the degree of degree-distribution in the entire range from 1 to 15, but only uses degree 1; and 2) it re-scales the pairwise distances in the range [0, 2] instead of range [0, 1]. A theoretical analysis is presented to prove the better performance of parameter-free SDD. In addition, the performances of the proposed parameter-free SDD and the standard SDD have been experimentally compared in terms of structure capturing and computational time. This paper also proposes a parametric version of SDD using a deep neural network approach to learn the mapping based on the samples of the original data and their corresponding embedded representations in a low dimensional space. Comparative experiments have been undertaken with SDD and other methods such as Isomap, t-SNE and UMAP to demonstrate the effectiveness of the proposed parametric SDD with several popular synthetic and real datasets such as Churn, SEER Breast Cancer, AVletters (LIPS Reading) and MNIST.