Search Results (1,103)

Artificial Intelligence (AI) has achieved immense progress in recent years across a wide array of application domains, with biomedical imaging and sensing emerging as particularly impactful areas. However, the integration of AI in safety-critical fields, particularly biomedical domains, continues to face a major challenge of explainability arising from the opacity of complex prediction models. Overcoming this obstacle falls within the realm of eXplainable Artificial Intelligence (XAI), which is widely acknowledged as an essential aspect for successfully implementing and accepting AI techniques in practical applications to ensure transparency, fairness, and accountability in the decision-making processes and mitigate potential biases. This article provides a systematic cross-domain review of XAI techniques applied to quantitative prediction tasks, with a focus on their methodological relevance and potential adaptation to biomedical imaging and sensing. To achieve this, following PRISMA guidelines, we conducted an analysis of 44 Q1 journal articles that utilised XAI techniques for prediction applications across different fields where quantitative databases were used, and their contributions to explaining the predictions were studied. As a result, 13 XAI techniques were identified for prediction tasks. Shapley Additive eXPlanations (SHAP) was identified in 35 out of 44 articles, reflecting its frequent computational use for feature-importance ranking and model interpretation. Local Interpretable Model-Agnostic Explanations (LIME), Partial Dependence Plots (PDPs), and Permutation Feature Index (PFI) ranked second, third, and fourth in popularity, respectively. The study also recognises theoretical limitations of SHAP and related model-agnostic methods, such as their additive and causal assumptions, which are particularly critical in heterogeneous biomedical data. Furthermore, a synthesis of the reviewed studies reveals that while many provide computational evaluation of explanations, none include structured human–subject usability validation, underscoring an important research gap for clinical translation. Overall, this study offers an integrated understanding of quantitative XAI techniques, identifies methodological and usability gaps for biomedical adaptation, and provides guidance for future research aimed at safe and interpretable AI deployment in biomedical imaging and sensing. Full article

Planetary exploration missions have acquired a growing amount of remote sensing data, offering a reliable basis for studying the geological evolution of planetary bodies such as Mars. In recent years, machine learning models have emerged as powerful tools for remote sensing by providing scalable and adaptive solutions for planetary science. We present a machine learning approach to map the spatial distribution of minerals on Mars, representing a step toward large-scale automated mineral mapping. Although existing CRISM dimensionality reduction methods are useful, the feature space remains high-dimensional, and relying on RGB overlays limits the ability to preserve and detect complex relationships, increasing the risk of missing important spectral patterns. Our framework utilises the Self-Organising Map (SOM) model and k-means clustering to identify clusters of spectral signatures, which may correspond to distinct minerals. It reduces dimensionality to a two-dimensional grid while preserving key high-dimensional patterns and relationships, providing a more reliable and interpretable basis for semi-automated analysis than RGB overlays. Although the clusters can be labelled by referencing a spectral library, our framework does not require labelled data and can operate in an unsupervised manner. The framework retains full spectral dimensionality of input features. The results indicate that our framework can identify the spatial distribution of minerals on Mars, even in complex spectral environments with overlapping features. Moreover, the SOM model output is interpretable rather than a black box, providing intuitive guidance for mineral exploration when applied in a semi-automated workflow. Full article

Ovarian cancer’s high mortality is primarily due to late-stage diagnosis, underscoring the critical need for improved early detection tools. This study develops and validates explainable artificial intelligence (XAI) models to discriminate malignant from benign ovarian masses using readily available demographic and laboratory data. A dataset of 309 patients (140 malignant, 169 benign) with 47 clinical parameters was analyzed. The Boruta algorithm selected 19 significant features, including tumor markers (CA125, HE4, CEA, CA19-9, AFP), hematological indices, liver function tests, and electrolytes. Five ensemble machine learning algorithms were optimized and evaluated using repeated stratified 5-fold cross-validation. The Gradient Boosting model achieved the highest performance with 88.99% (±3.2%) accuracy, 0.934 AUC-ROC, and 0.782 Matthews correlation coefficient. SHAP analysis identified HE4, CEA, globulin, CA125, and age as the most globally important features. Unlike black-box approaches, our XAI framework provides clinically interpretable decision pathways through LIME and SHAP visualizations, revealing how feature values push predictions toward malignancy or benignity. Partial dependence plots illustrated non-linear risk relationships, such as a sharp increase in malignancy probability with CA125 > 35 U/mL. This explainable approach demonstrates that ensemble models can achieve high diagnostic accuracy using routine lab data alone, performing comparably to established clinical indices while ensuring transparency and clinical plausibility. The integration of state-of-the-art XAI techniques highlights established biomarkers and reveals potential novel contributors like inflammatory and hepatic indices, offering a pragmatic, scalable triage tool to augment existing diagnostic pathways, particularly in resource-constrained settings. Full article

Effective spam detection systems are essential in online social media networks (OSNs) and cybersecurity, and they directly influence the quality of decision-making pertaining to security. With today’s digital communications, unsolicited spam degrades user experiences and threatens platform security. Machine learning-based spam detection systems offer an automated defense. Despite their effectiveness, such methods are frequently hindered by the “black box” problem, an interpretability deficiency that constrains their deployment in security applications, which, in order to comprehend the rationale of classification processes, is crucial for efficient threat evaluation and response strategies. However, their effectiveness hinges on selecting an optimal feature subset. To address these issues, we propose a lightweight, explainable spam detection model that integrates a nature-inspired optimizer. The approach employs clean data with data preprocessing and feature selection using a swarm-based, nature-inspired meta-heuristic Rüppell’s Fox Optimization (RFO) algorithm. To the best of our knowledge, this is the first time the algorithm has been adapted to the field of cybersecurity. The resulting minimal feature set is used to train a supervised classifier that achieves high detection rates and accuracy with respect to spam accounts. For the interpretation of model predictions, Shapley values are computed and illustrated through swarm and summary charts. The proposed system was empirically assessed using two datasets, achieving accuracies of 99.10%, 98.77%, 96.57%, and 92.24% on Dataset 1 using RFO with DT, KNN, AdaBoost, and LR and 98.94%, 98.67%, 95.04%, and 94.52% on Dataset 2, respectively. The results validate the efficacy of the suggested approach, providing an accurate and understandable model for spam account identification. This study represents notable progress in the field, offering a thorough and dependable resolution for spam account detection issues. Full article

Understanding internal resonance phenomena in transformer windings is essential for evaluating insulation performance and preventing equipment failure under transient conditions. This study presents a measurement-based modeling approach to assess internal voltage distributions in a high-voltage transformer winding of a dry-type distribution transformer. Frequency-domain admittance and voltage transfer functions were experimentally obtained and approximated using vector fitting. The resulting models were employed to simulate time-domain responses through a two-step procedure that integrates electromagnetic transient simulations of the terminal circuit with frequency-derived internal voltage models. The validation was performed using a sinusoidal excitation at 51 kHz, corresponding to the first-mode resonance frequency. Simulated internal voltages and input currents showed close agreement with experimental measurements, confirming the model’s accuracy. The study identified two critical resonance frequencies at 51 kHz and 59 kHz, at which voltage amplification can become severe. At 51 kHz, the maximum overvoltage reached nearly seven times the applied voltage at the winding midpoint, indicating a substantial risk of dielectric failure. These findings highlight the importance of accurately characterizing internal resonances in transformer windings, especially during insulation coordination studies. The proposed methodology offers an effective tool for analyzing internal overvoltages and contributes to the development of more robust transformer design and protection strategies. Full article

The transition to 5G networks brings unprecedented speed, ultra-low latency, and massive connectivity. Nevertheless, it introduces complex traffic patterns and broader attack surfaces that render traditional intrusion detection systems (IDSs) ineffective. Existing rule-based methods and classical machine learning approaches struggle to capture the temporal and dynamic characteristics of 5G traffic, while many deep learning models lack interpretability, making them unsuitable for high-stakes security environments. To address these challenges, we propose Bidirectional Temporal Anomaly Detector (BiTAD), a deep temporal learning architecture for anomaly detection in 5G networks. BiTAD leverages dual-direction temporal sequence modelling with attention to encode both past and future dependencies while focusing on critical segments within network sequences. Like many deep models, BiTAD’s faces interpretability challenges. To resolve its “black-box” nature, a dual-perspective explainability module, coined TwinLens, is proposed. This module integrates SHAP and TimeSHAP to provide global feature attribution and temporal relevance, delivering dual-perspective interpretability. Evaluated on the public 5G-NIDD dataset, BiTAD demonstrates superior detection performance compared to existing models. TwinLens enables transparent insights by identifying which features and when they were most influential to anomaly predictions. By jointly addressing the limitations in temporal modelling and interpretability, our work contributes a practical IDS framework tailored to the demands of next-generation mobile networks. Full article

Making time-critical decisions with serious consequences is a daily aspect of work environments. To support the process of finding optimal actions, data-driven approaches are increasingly being used. The most advanced form of data-driven analytics is prescriptive analytics, which prescribes actionable recommendations for users. However, the produced recommendations rely on complex models and optimization techniques that are difficult to understand or justify to non-expert users. Currently, there is a lack of platforms that offer easy integration of domain-specific prescriptive analytics workflows into production environments. In particular, there is no centralized environment and standardized approach for implementing such prescriptive workflows. To address these challenges, large language models (LLMs) can be leveraged to improve interpretability by translating complex recommendations into clear, context-specific explanations, enabling non-experts to grasp the rationale behind the suggested actions. Nevertheless, we acknowledge the inherent black-box nature of LLMs, which may introduce limitations in transparency. To mitigate these limitations and to provide interpretable recommendations based on real user knowledge, a knowledge graph is integrated. In this paper, we present and validate a prescriptive analytics platform that integrates ontology-based graph retrieval-augmented generation (GraphRAG) to enhance decision making by delivering actionable and context-aware recommendations. For this purpose, a knowledge graph is created through a fully automated workflow based on an ontology, which serves as the backbone of the prescriptive platform. Data sources for the knowledge graph are standardized and classified according to the ontology by employing a zero-shot classifier. For user-friendly presentation, we critically examine the usability of GraphRAG in prescriptive analytics platforms. We validate our prescriptive platform in a customer clinic with industry experts in our IoT-Factory, a dedicated research environment. Full article

Accurate assessment of sleep architecture is critical for diagnosing and managing sleep disorders, which significantly impact global health and well-being. While polysomnography (PSG) remains the clinical gold standard, its inherent intrusiveness, high cost, and logistical complexity limit its utility for routine or home-based monitoring. Recent advances highlight that subtle variations in respiratory dynamics, such as respiratory rate and cycle regularity, exhibit meaningful correlations with distinct sleep stages and could serve as valuable non-invasive biomarkers. In this work, we propose a framework for estimating sleep stage distribution—specifically Wake, Light (N1+N2), Deep (N3), and REM—based on respiratory audio captured over a single sleep episode. The framework comprises three principal components: (1) a segmentation module that identifies distinct respiratory cycles in respiratory sounds using a fine-tuned Transformer-based architecture; (2) a feature extraction module that derives a suite of statistical, spectral, and distributional descriptors from these segmented respiratory patterns; and (3) stage-specific regression models that predict the proportion of time spent in each sleep stage. Experiments on the public PSG-Audio dataset (287 subjects; mean 5.3 h per subject), using subject-wise cross-validation, demonstrate the efficacy of the proposed approach. The segmentation model achieved lower RMSE and MAE in predicting respiratory rate and cycle duration, outperforming classical signal-processing baselines. For sleep stage proportion prediction, the proposed method yielded favorable RMSE and MAE across all stages, with the TabPFN model consistently delivering the best results. By quantifying interpretable respiratory features and intentionally avoiding black-box end-to-end modeling, our system may support transparent, contact-free sleep monitoring using passive audio. Full article

Sepsis remains a critical medical emergency caused by a dysregulated immune response to infection, with timely detection and intervention being essential for improving survival rates. Traditional methods often rely on clinician intuition and structured scoring systems, which may be time-intensive and prone to variability. To address these limitations, Machine Learning (ML) offers a powerful alternative, bringing precision and efficiency to sepsis detection. This study investigates both white-box and complex black-box ML models applied to patient data collected across the continuum of care, including monitoring at the urgent care, en route in ambulances, and diagnostics conducted within hospital emergency department settings themselves. White-box models, such as logistic regression and decision trees, are valued for their interpretability, allowing healthcare providers to understand and trust the reasoning behind predictions. Meanwhile, black-box models like deep neural networks and support vector machines deliver superior accuracy but pose challenges in clinical transparency. This trade-off between explainability and performance is explored in detail, supported by experimental results aimed at identifying the most effective computational strategies for early sepsis recognition across diverse healthcare environments. Full article

Online customer reviews contain rich sentimental expressions of customer preferences on products, which is valuable information for analyzing customer preferences in product design. The adaptive neuro fuzzy inference system (ANFIS) was applied to the establishment of customer preference models based on online reviews, which can address the fuzziness of customers’ emotional responses in comments and the nonlinearity of modeling. However, due to the black box problem in ANFIS, the nonlinearity of the modeling cannot be shown explicitly. To solve the above problems, a chaos-driven ANFIS approach is proposed to develop customer preference models using online comments. The model’s nonlinear relationships are represented transparently through the fuzzy rules obtained, which provide human-readable equations. In the proposed approach, online reviews are analyzed using sentiment analysis to extract the information that will be used as the data sets for modeling. After that, the chaos optimization algorithm (COA) is applied to determine the polynomial structure of the fuzzy rules in ANFIS to model the customer preferences. Using laptop products as a case study, several approaches are evaluated for validation, including fuzzy regression, fuzzy least-squares regression, ANFIS, ANFIS with subtractive cluster, and ANFIS with K-means. Compared to the other five approaches, the values of mean relative error, variance of error, and confidence interval of validation error are improved based on the proposed approach. Full article

Recent advancements in large language models (LLMs) have increased the necessity of alignment and safety mechanisms. Despite these efforts, jailbreak attacks remain a significant threat, exploiting vulnerabilities to elicit harmful responses. While white-box attacks, such as the Greedy Coordinate Gradient (GCG) method, have demonstrated promise, their efficacy is often limited by non-smooth optimization landscapes and a tendency to converge to local minima. To mitigate these issues, we propose Spatial Momentum GCG (SM-GCG), a novel method that incorporates spatial momentum. This technique aggregates gradient information across multiple transformation spaces—including text, token, one-hot, and embedding spaces—to stabilize the optimization process and enhance the estimation of update directions, thereby more effectively exploiting model vulnerabilities to elicit harmful responses. Experimental results on models including Vicuna-7B, Guanaco-7B, and Llama2-7B-Chat demonstrate that SM-GCG significantly enhances the attack success rate in white-box settings. The method achieves a 10–15% improvement in attack success rate over baseline methods against robust models such as Llama2, while also exhibiting enhanced transferability to black-box models. These findings indicate that spatial momentum effectively mitigates the problem of local optima in discrete prompt optimization, thereby offering a more powerful and generalizable approach for red-team assessments of LLM safety. Warning: This paper contains potentially offensive and harmful text. Full article

This study advances the unmanned aerial vehicle (UAV) localization technology within the framework of a low-altitude economy, with particular emphasis on the accurate and real-time identification and tracking of unauthorized (“black-flying”) drones. Conventional YOLOv8s-based target detection algorithms often suffer from missed detections due to their reliance on single-frame features. To address this limitation, this paper proposes an improved detection algorithm that integrates a long-short-term memory (LSTM) network into the YOLOv8s framework. By incorporating time-series modeling, the LSTM module enables the retention of historical features and dynamic prediction of UAV trajectories. The loss function combines bounding box regression loss with binary cross-entropy and is optimized using the Adam algorithm to enhance training convergence. The training data distribution is validated through Monte Carlo random sampling, which improves the model’s generalization to complex scenes. Simulation results demonstrate that the proposed method significantly enhances UAV detection performance. In addition, when deployed on the RK3588-based embedded system, the method achieves a low false negative rate and exhibits robust detection capabilities, indicating strong potential for practical applications in airspace management and counter-UAV operations. Full article

An Adaptive Neuro-Fuzzy Inference System (ANFIS) framework for paediatric wrist injury classification (fracture versus sprain) was developed utilising infrared thermography (IRT). ANFIS combines artificial neural network (ANN) learning with interpretable fuzzy rules, mitigating the “black-box” limitation of conventional ANNs through explicit membership functions and Takagi–Sugeno rule consequents. Forty children (19 fractures, 21 sprains, confirmed by X-ray radiograph) provided thermal image sequences from which three statistically discriminative temperature distribution features namely standard deviation, inter-quartile range (IQR) and kurtosis were selected. A five-layer Sugeno ANFIS with Gaussian membership functions were trained using a hybrid least-squares/gradient descent optimisation and evaluated under three premise-parameter initialisation strategies: random seeding, K-means clustering, and fuzzy C-means (FCM) data partitioning. Five-fold cross-validation guided the selection of membership functions standard deviation (σ) and rule count, yielding an optimal nine-rule model. Comparative experiments show K-means initialisation achieved the best balance between convergence speed and generalisation versus slower but highly precise random initialisation and rapidly convergent yet unstable FCM. The proposed K-means–driven ANFIS offered data-efficient decision support, highlighting the potential of thermal feature fusion with neuro-fuzzy modelling to reduce unnecessary radiographs in emergency bone fracture triage. Full article

As a critical technological foundation for electric vehicles, power battery state estimation primarily involves estimating the State of Charge (SOC), the State of Health (SOH) and the Remaining Useful Life (RUL). This paper systematically categorizes battery state estimation methods into three distinct generations, tracing the evolutionary progression from single-state to multi-state cooperative estimation approaches. First-generation methods based on equivalent circuit models offer straightforward implementation but accumulate SOC-SOH estimation errors during battery aging, as they fail to account for the evolution of microscopic parameters such as solid electrolyte interphase film growth, lithium inventory loss, and electrode degradation. Second-generation data-driven approaches, which leverage big data and deep learning, can effectively model highly nonlinear relationships between measurements and battery states. However, they often suffer from poor physical interpretability and generalizability due to the “black-box” nature of deep learning. The emerging third-generation technology establishes transmission mechanisms from microscopic electrode interface parameters via electrochemical impedance spectroscopy to macroscopic SOC, SOH, and RUL states, forming a bidirectional closed-loop system integrating estimation, prediction, and optimization that demonstrates potential to enhance both full-operating-condition adaptability and estimation accuracy. This progress supports the development of high-reliability, long-lifetime electric vehicles. Full article

3D point cloud object tracking plays a critical role in fields such as autonomous driving and robotics, making the security of these models essential. Adversarial attacks are a key approach for studying the robustness and security of tracking models. However, research on the generalization of adversarial attacks for 3D point-cloud-tracking models is limited, and the frequency-domain information of the point cloud’s geometric structure is often overlooked. This frequency information is closely related to the generalization of 3D point-cloud-tracking models. To address these limitations, this paper proposes a novel adversarial method for 3D point cloud object tracking, utilizing frequency-domain attacks based on the importance of frequency bands. The attack operates in the frequency domain, targeting the low-frequency components of the point cloud within the search area. To make the attack more targeted, the paper introduces a frequency band importance saliency map, which reflects the significance of sub-frequency bands for tracking and uses this importance as attack weights to enhance the attack’s effectiveness. The proposed attack method was evaluated on mainstream 3D point-cloud-tracking models, and the adversarial examples generated from white-box attacks were transferred to other black-box tracking models. Experiments show that the proposed attack method reduces both the average success rate and precision of tracking, proving the effectiveness of the proposed adversarial attack. Furthermore, when the white-box adversarial samples were transferred to the black-box model, the tracking metrics also decreased, verifying the transferability of the attack method. Full article