With the recent advancements in the field of wearable technologies, the opportunity to monitor stress continuously using different physiological variables has gained significant interest. The early detection of stress can help improve healthcare and minimizes the negative impact of long-term stress. This paper reports outcomes of a pilot study and associated stress-monitoring dataset, named the “Stress-Predict Dataset”, created by collecting physiological signals from healthy subjects using wrist-worn watches with a photoplethysmogram (PPG) sensor. While wearing these watches, 35 healthy volunteers underwent a series of tasks (i.e., Stroop color test, Trier Social Stress Test and Hyperventilation Provocation Test), along with a rest period in-between each task. They also answered questionnaires designed to induce stress levels compatible with daily life. The changes in the blood volume pulse (BVP) and heart rate were recorded by the watch and were labelled as occurring during stress-inducing tasks or a rest period (no stress). Additionally, respiratory rate was estimated using the BVP signal. Statistical models and personalised adaptive reference ranges were used to determine the utility of the proposed stressors and the extracted variables (heart rate and respiratory rate). The analysis showed that the interview session was the most significant stress stimulus, causing a significant variation in heart rate of 27 (77%) participants and respiratory rate of 28 (80%) participants out of 35. The outcomes of this study contribute to the understanding the role of stressors and their association with physiological response and provide a dataset to help develop new wearable solutions for more reliable, valid, and sensitive physio-logical stress monitoring.

In certain healthcare settings, such as emergency or critical care units, where quick and accurate real-time analysis and decision-making are required, the healthcare system can leverage the power of artificial intelligence (AI) models to support decision-making and prevent complications. This paper investigates the optimization of healthcare AI models based on time complexity, hyper-parameter tuning, and XAI for a classification task. The paper highlights the significance of a lightweight convolutional neural network (CNN) for analysing and classifying Magnetic Resonance Imaging (MRI) in real-time and is compared with CNN-RandomForest (CNN-RF). The role of hyper-parameter is also examined in finding optimal configurations that enhance the model’s performance while efficiently utilizing the limited computational resources. Finally, the benefits of incorporating the XAI technique (e.g. GradCAM and Layer-wise Relevance Propagation) in providing transparency and interpretable explanations of AI model predictions, fostering trust, and error/bias detection are explored. Our inference time on a MacBook laptop for 323 test images of size 100x100 is only 2.6 sec, which is merely 8 milliseconds per image while providing comparable classification accuracy with the ensemble model of CNN-RF classifiers. Using the proposed model, clinicians/cardiologists can achieve accurate and reliable results while ensuring patients’ safety and answering questions imposed by the General Data Protection Regulation (GDPR). The proposed investigative study will advance the understanding and acceptance of AI systems in connected healthcare settings.

Identifying hidden biases in AI documentation metadata (model, data, and dataspace cards) is essential for responsible AI; yet this domain remains largely unexplored. The proposed work evaluates four Transformer models (XLNet, DistilBERT, RoBERTa, and ELECTRA) for bias detection across publicly available, synthetic, and custom datasets. On the BABE news corpus, all models achieved 77–80% accuracy, with only ELECTRA exceeding 80% on every metric. To address the absence of publicly available AI-card datasets, we generated synthetic metadata for two use cases ( Customer Interaction and Customer Data Uploaded by Organisations ) using ChatGPT. Models trained on this synthetic corpus displayed near-perfect scores, reflecting shared stylistic cues embedded in the generated text. To test real-world robustness, we curated a Hugging Face dataset by scraping documentation comments, filtering for bias-related keywords, and obtaining annotations from four independent labellers in a single-blind setting. Partial fine-tuning (zero-shot) evaluations of models trained only on BABE or synthetic data revealed substantial performance drops on this real-world set. To mitigate this cross-domain loss, we introduce a cascaded, full fine-tuning (few-shot) pipeline in which Transformer models are sequentially fine-tuned on BABE, synthetic text, and a subset of the Hugging Face corpus. Evaluation on the remaining portion achieved over 85% across all performance metrics, enhancing precision and generalisation. This study demonstrates the challenges of bias detection beyond controlled or synthetic data and highlights cascaded fine-tuning as a practical, low-resource strategy. Future directions include leveraging evidence fusion methods, integrating cross-attention with bias taxonomies, and adopting dual-encoder architectures to advance bias detection toward more in-depth, knowledge-guided reasoning.