Background: Accurate and timely segmentation of ischemic stroke lesions from diffusion-weighted imaging (DWI) is crucial for diagnosis and treatment planning. Manual segmentation is labor-intensive, time-consuming, and prone to inter-observer variability. This study aims to develop and validate a novel deep learning framework that overcomes the common trade-off between high segmentation accuracy and the computational efficiency required for practical clinical use. Methods: We developed FasterNet and Attention-Gated UNet (FA-UNet), a hybrid U-Net-based architecture. The model’s design features two key innovations: a computationally efficient FasterNet block at the bottleneck to capture global lesion context and multi-scale attention gates (MSAGs) on the skip connections to adaptively refine features and suppress noise. The model was trained and validated on the public Ischemic Stroke Lesion Segmentation (ISLES) 2022 dataset (n = 250 patients) and its performance was assessed on an independent, private test set of 600 DWI scans from 80 patients. FA-UNet’s performance was benchmarked against several state-of-the-art U-Net variants using the Dice coefficient, Intersection over Union (IoU), sensitivity, and precision as primary outcome measures. Results: On the independent test set (n = 80), the proposed FA-UNet model achieved a Dice coefficient of 0.8676 and an IoU of 0.7584. This performance surpassed all benchmarked architectures, including U-Net, U-Net3plus, and CMU-Net. Compared with the next best performing model, this represents a relative improvement of approximately 1.64% in the Dice score and 1.42% in IoU. Conclusion: The FA-UNet architecture establishes a new state-of-the-art performance benchmark for automated ischemic stroke segmentation. By effectively balancing high accuracy with computational efficiency, it offers a robust, reliable, and clinically viable tool.

Background Phishing attacks pose a significant threat to Industrial Internet of Things (IoT) networks, with a documented 350% increase in IoT-targeted phishing attempts since 2020. This alarming trend necessitates robust security measures to protect critical infrastructure. Methods We propose a novel methodology integrating a CNN-LSTM model for initial phishing detection and a sophisticated LLM for comprehensive analysis. Our approach combines spatial feature extraction via CNN layers with temporal pattern recognition through LSTM, enhanced by explainable AI techniques for transparency. Results Our CNN-LSTM model achieves 93.26% accuracy on the Web URL dataset and the DistilBERT model achieves an accuracy of 98.64% on text and email dataset. The addition of explainable AI improves transparency by 41% compared to black-box approaches, which is crucial for IoT security operations. Conclusion This approach not only improves phishing detection accuracy but also contributes to adaptive security measures tailored to evolving cybersecurity threats. By reducing false positives by 41% compared to traditional methods and providing human-interpretable security insights through LLM integration, our framework ensures the resilience of IoT networks while maintaining operational efficiency.

Background and Objectives: Liver cancer ranks among the leading causes of cancer-related mortality, necessitating the development of novel diagnostic methods. Deregulated lipid metabolism, a hallmark of hepatocarcinogenesis, offers compelling prospects for biomarker identification. This study aims to employ explainable artificial intelligence (XAI) to identify lipidomic biomarkers for liver cancer and to develop a robust predictive model for early diagnosis. Materials and Methods: This study included 219 patients diagnosed with liver cancer and 219 healthy controls. Serum samples underwent untargeted lipidomic analysis with LC-QTOF-MS. Lipidomic data underwent univariate and multivariate analyses, including fold change (FC), t-tests, PLS-DA, and Elastic Network feature selection, to identify significant biomarker candidate lipids. Machine learning models (AdaBoost, Random Forest, Gradient Boosting) were developed and evaluated utilizing these biomarkers to differentiate liver cancer. The AUC metric was employed to identify the optimal predictive model, whereas SHAP was utilized to achieve interpretability of the model’s predictive decisions. Results: Notable alterations in lipid profiles were observed: decreased sphingomyelins (SM d39:2, SM d41:2) and increased fatty acids (FA 14:1, FA 22:2) and phosphatidylcholines (PC 34:1, PC 32:1). AdaBoost exhibited a superior classification performance, achieving an AUC of 0.875. SHAP identified PC 40:4 as the most efficacious lipid for model predictions. The SM d41:2 and SM d36:3 lipids were specifically associated with an increased risk of low-onset cancer and elevated levels of the PC 40:4 lipid. Conclusions: This study demonstrates that untargeted lipidomics, in conjunction with explainable artificial intelligence (XAI) and machine learning, may effectively identify biomarkers for the early detection of liver cancer. The results suggest that alterations in lipid metabolism are crucial to the progression of liver cancer and provide valuable insights for incorporating lipidomics into precision oncology.

Aim: Breast cancer (BC) is the most common type of cancer in women, accounting for more than 30% of new female cancers each year. Although various treatments are available for BC, most cancer-related deaths are due to incurable metastases. Therefore, the early diagnosis and treatment of BC are crucial before metastasis. Mammography and ultrasonography are primarily used in the clinic for the initial identification and staging of BC; these methods are useful for general screening but have limitations in terms of sensitivity and specificity. Omics-based biomarkers, like metabolomics, can make early diagnosis much more accurate, make tracking the disease’s progression more accurate, and help make personalized treatment plans that are tailored to each tumor’s specific molecular profile. Metabolomics technology is a feasible and comprehensive method for early disease detection and biomarker identification at the molecular level. This research aimed to establish an interpretable predictive artificial intelligence (AI) model using plasma-based metabolomics panel data to identify potential biomarkers that distinguish BC individuals from healthy controls. Methods: A cohort of 138 BC patients and 76 healthy controls were studied. Plasma metabolites were examined using LC-TOFMS and GC-TOFMS techniques. Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), Adaptive Boosting (AdaBoost), and Random Forest (RF) were evaluated using performance metrics such as Receiver Operating Characteristic-Area Under the Curve (ROC AUC), accuracy, sensitivity, specificity, and F1 score. ROC and Precision-Recall (PR) curves were generated for comparative analysis. The SHapley Additive Descriptions (SHAP) analysis evaluated the optimal prediction model for interpretability. Results: The RF algorithm showed improved accuracy (0.963 ± 0.043) and sensitivity (0.977 ± 0.051); however, LightGBM achieved the highest ROC AUC (0.983 ± 0.028). RF also achieved the best Precision-Recall Area under the Curve (PR AUC) at 0.989. SHAP search found glycerophosphocholine and pentosidine as the most significant discriminatory metabolites. Uracil, glutamine, and butyrylcarnitine were also among the significant metabolites. Conclusions: Metabolomics biomarkers and an explainable AI (XAI)-based prediction model showed significant diagnostic accuracy and sensitivity in the detection of BC. The proposed XAI system using interpretable metabolite data can serve as a clinical decision support tool to improve early diagnosis processes.

Background and Objectives: Rheumatoid arthritis (RA) is a chronic autoimmune disease characterised by joint inflammation and pain. Metabolomics approaches, which are high-throughput profiling of small molecule metabolites in plasma or serum in RA patients, have so far provided biomarker discovery in the literature for clinical subgroups, risk factors, and predictors of treatment response using classical statistical approaches or machine learning models. Despite these recent developments, an explainable artificial intelligence (XAI)-based methodology has not been used to identify RA metabolomic biomarkers and distinguish patients with RA. This study constructed a XAI-based EBM model using global plasma metabolomics profiling to identify metabolites predictive of RA patients and to develop a classification model that can distinguish RA patients from healthy controls. Materials and Methods: Global plasma metabolomics data were analysed from RA patients (49 samples) and healthy individuals (10 samples). SMOTE technique was used for class imbalance in data preprocessing. EBM, LightGBM, and AdaBoost algorithms were applied to generate a discriminatory model between RA and controls. Comprehensive performance metrics were calculated, and the interpretability of the optimal model was assessed using global and local feature descriptions. Results: A total of 59 samples were analysed, 49 from RA patients, and 10 from healthy subjects. The EBM generated better results than LightGBM and AdaBoost by attaining an AUC of 0.901 (95% CI: 0.847–0.955) with 87.8% sensitivity which helps prevent false negative early RA diagnosis. The primary biomarkers EBM-based XAI identified were N-acetyleucine, pyruvic acid, and glycerol-3-phosphate. EBM global explanation analysis indicated that elevated pyruvic acid levels were significantly correlated with RA, whereas N-acetyleucine exhibited a nonlinear relationship, implying possible protective effects at specific concentrations. Conclusions: This study underscores the promise of XAI and evidence-based medicine methodology in developing biomarkers for RA through metabolomics. The discovered metabolites offer significant insights into RA pathophysiology and may function as diagnostic biomarkers or therapeutic targets. Incorporating EBM methodologies integrated with XAI improves model transparency and increases the therapeutic applicability of predictive models for RA diagnosis/management. Furthermore, the transparent structure of the EBM model empowers clinicians to understand and verify the reasoning behind each prediction, thereby fostering trust in AI-assisted decision-making and facilitating the integration of metabolomic insights into routine clinical practice.