Abstract
Automated lung CT segmentation is a prerequisite for quantitative pulmonary analysis. However, the standard 'accuracy-first' development cycle often neglects clinical safety, as traditional metrics like the Dice Similarity Coefficient do not differentiate between benign boundary errors and serious omissions such as missed malignant nodules. This study proposes a human-centric framework to enhance the trustworthiness of lung segmentation by integrating annotation governance, predictive uncertainty quantification, and a risk-based evaluation taxonomy. A ResNet34 + U-Net hybrid model was developed. A specialized annotation harmonization protocol was implemented to ensure the consistent inclusion of juxta-pleural nodules. Predictive uncertainty was quantified via Monte Carlo Dropout to generate voxel-wise predictive entropy maps. The system was evaluated using a proposed risk-based error taxonomy and simulated a human-AI triage workflow. Result obtained showed model M3, which incorporated annotation correction and data augmentation, achieved 0.66% Dice improvement over baseline M1. The framework reduced moderate-risk errors associated with non-lung inclusion by 30% (0.396% ? 0.276%). From the human-AI triage simulation, 37.2%, 16.9% and 10.8% high-risk cases were recalled at operating thresholds (t) of 0.0005,0.003 and 0.01, respectively. Our framework provides a robust foundation for semiautonomous clinical workflows by prioritizing error transparency and risk-aware evaluation. Thus, shifting the paradigm from opaque automation to trustworthy human-AI collaboration. © 2026 IEEE.

Abstract

Abstract
The rapid expansion of solar photovoltaic (SPV) systems poses critical challenges to material supply security and waste management. Addressing these challenges require integrating circular economy strategies. This study develops a dynamic and probabilistic material flow analysis (MFA) to quantify the lifecycle material flows of crystalline silicon (c-Si) modules from 1998 to 2050, with waste projections extended to 2099. Three circular economy scenarios are evaluated, integrating the European Union Directive targets and strategies for reducing, reusing, and recycling. Uncertainty is explicitly addressed through Monte Carlo simulation, capturing variability in installed capacity projections, Weibull lifetime parameters, material composition, pre-operational losses, and recycling efficiencies. Portugal is used as a national-scale case study to demonstrate the applicability of the proposed methodology. Results indicate a cumulative material requirement of approximately 1.46 Mt by 2050 without circular strategies. Across low-, medium-, and high-circularity scenarios, both total material demand and the share of primary versus secondary raw materials vary substantially. Notably, scenarios incorporating reuse may increase primary material extraction due to reduced availability of secondary materials for manufacturing. Deterministic analysis suggests that full c-Si loop closure can be achieved between 2039 and 2041, depending on the scenario. However, probabilistic results reveal substantial uncertainty, with the probability of 100% Circular Material Use Rate (CMUR) in the period 2030-2050 among 53.7%, 43.6% and 68.6% under low, medium, and high circularity respectively. Sensitivity analysis identifies future c-Si's deployment and lifetimes as the dominant drivers of circularity outcomes. This probabilistic MFA contributes with robust evidence to support circular economy policy design and infrastructure planning while opening avenues for further research.

Abstract

Abstract
Demand Responsive Transport (DRT) systems provide versatile transport operations and are capable of quickly adjusting to fluctuating passenger demand. Unlike traditional public transport (PT), which operates with fixed routes and schedules, DRTs offer flexibility in vehicle routes, fleet sizes, and schedules. This flexibility is an intrinsic characteristic of DRT systems and a key attribute in their design and associated decision-making processes. However, flexibility also presents significant design challenges, due to the multitude of potential configurations and the unique characteristics of each service area. As practice shows, the effectiveness of DRT configurations is heavily influenced by demand levels. Highly flexible operations are typically suited for low-demand areas, whereas higher demand may require reduced flexibility to maintain system efficiency. Furthermore, demand may grow to a point where the operators may question whether to continue operating as a DRT or shift to traditional regular public transport, with predefined routes and schedules, and more efficient operation. This work studies how demand levels and characteristics can be used in the decision to adopt a DRT system, instead of PT. The problem was addressed through the simulation of various demand scenarios in a virtual environment, thus comparing the performance of the different transport systems. In the scenario analysed, it was possible to identify a demand threshold where the DRT system is more efficient, while higher demand favours the fixed-route system. However, it is important to note that this threshold may be significantly influenced by the specific characteristics of the service area where the system will operate. Copyright © 2025. Published by Elsevier B.V.

Abstract
Ambient intelligence (AmI) refers to environments where smart devices, sensors, and AI-driven systems work seamlessly to enhance human interactions with their surroundings. Through the combination of real-time data, context-awareness, and adaptive learning, AmI enables environments to respond proactively to user needs, improving efficiency, comfort, and decision-making. However, since AmI systems are inherently human-centric and often operate autonomously, they must be designed with robust ethical, privacy, and safety considerations. Ensuring that these systems function reliably, fairly, and without harm is crucial, especially in sensitive domains like healthcare, security, and smart infrastructure. This work introduces a novel tool, conceptualized as an AmI Digital Twin, which allows developers to simulate or monitor AmI data streams, and develop and thoroughly test AmI applications before and during their real use. Built on a modular architecture leveraging technologies like React.js, Node.js, Kafka, Faust, MongoDB, InfluxDB, Grafana, and Docker, the platform ensures adaptability to different application environments, scalability, and ease of deployment. Besides the description of the tool itself, we provide some early validation results in common AmI tasks such as anomaly and concept drift detection. The tool is available in a public repository, and comes pre-packaged with a set of applications for AmI use-cases. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.