Abstract: Lung cancer remains a formidable challenge in modern medicine, characterized by low survival rates and high risks for patients. Improving the medicine by automating the part of the job is crucial for medical employees so they can focus on patient treatment. This paper introduces a clinical decision support system designed to match lung cancer symptoms with potential therapies by employing a knowledge base of rules to analyze patient symptoms and recommend personalized treatment options. The system continuously monitors patient data for symptoms before surgery, enabling early detection of critical conditions. This enables medical professionals to intervene promptly and make necessary adjustments to the treatment plan. In some cases, if the patient's condition deteriorates, the system can suggest postponing or canceling scheduled procedures to prevent complications. Moreover, the rule-based architecture ensures that the system remains flexible and up to date. New medical findings and treatment protocols can be easily integrated, allowing healthcare providers to access the most relevant and accurate information. This adaptability enhances decision-making and improves patient outcomes.

Abstract: This paper assesses the potential of Sentinel-2 remote sensing data to support logistics planning and enhance intelligent transport systems (ITS) at border crossings in Serbia. Leveraging Serbia's strategic transit position, which connects Western and Central Europe with the Middle East, the study addresses the challenges of cross-border delays that arise from Serbia's status as a non-EU country. These delays often lead to frequent traffic congestion, and increased logistics costs. The study develops two algorithms based on Sentinel-2 images. The first algorithm detects moving trucks on roads leading to border crossings using spectral band analysis and temporal displacement effects, achieving an accuracy of 74%. The second algorithm identifies the end of freight vehicle queues, enhanced through super-resolution techniques, and reaches an accuracy of 91% in detecting queue endpoints. The results demonstrate that Sentinel-2 data has the potential to provide cost-effective, scalable solutions for traffic monitoring and optimizing cross-border logistics operations. Nevertheless, it is necessary to address the limitation of its 5-day temporal resolution to enhance its applicability for continuous traffic monitoring.

Abstract: This paper presents the design and development of a novel aquatic haptic interface aimed at providing real-time biomechanical feedback to swimmers. Traditional coaching methods rely on visual observation and post-activity analysis, which limits immediate corrections during swimming. To address these challenges, a wearable haptic feedback system was developed, integrating multiple actuators and kinematic sensors to deliver haptic information in real-time. The system architecture was designed to be modular and adaptable, with a communication framework ensuring efficient data exchange despite aquatic constraints. Three user studies were conducted to evaluate haptic perception, usability, and the impact of feedback on swimming technique refinement. Results demonstrate that swimmers successfully interpreted and responded to meaningful haptic cues, leading to immediate movement adjustments and enhanced technique precision. The findings highlight the potential of haptic feedback in aquatic sports training, providing a foundation for future advancements in swimmer performance optimization and rehabilitation applications.

Abstract: This article investigates the technical possibilities for low-latency transmission of inertial sensor data using consumer-grade software-defined radio (SDR) platforms. Motivated by the need for real-time feedback in applications such as biomechanical analysis in sports, where delays must be kept below 100 milliseconds, we explored a custom transmission protocol designed to handle high sampling rates and precise data payloads. The study examines three transmission strategies with the ADALM-Pluto SDR: using GNU Radio’s default scheduler settings, customizing its buffer sizes, and bypassing the scheduler entirely by employing the pySDR library. GNU Radio’s block-based model, while powerful, introduces latency through its buffering and scheduling mechanisms. Our experiments, conducted with simulated 1 kHz inertial sensor data encapsulated in 30-byte packets and modulated with GMSK, reveal that the GNU Radio configuration incurs significant delays (250–500 ms). Adjusting buffer sizes offers improvements but fails to eliminate inefficiencies. In contrast, the pySDRbased approach achieves substantially lower delays (10–40 ms) despite some challenges with packet loss, primarily due to USB. These results underscore the potential of custom SDR solutions for low-latency applications and suggest directions for further optimization and hardware integration.

Abstract: Buildings are significant consumers of energy, and reducing their energy use is a very important aspect of lowering the environmental impact of the energy sector. Accurate forecasts of the building demands for various forms of energy are essential for the operation of modern and sophisticated energy systems. Data-driven machine learning methods are among the key technologies that provide such predictions today. This paper presents an approach to onehour-ahead forecast of the heat demand for a group of multi-story residential buildings connected to a district heating system. The methodology presented in this paper uses the gradient boosting regressor and defines four prediction models for various parts of a day. The obtained coefficient of determination is greater than 0.99, the root mean squared error is 25.85 kWh, the mean absolute error is 11.44 kWh, and the median absolute error is 6.33 kWh, for the hourly demand data that range from zero to 2179 kWh. This approach is compared to the approach that uses only one model and appears to have 13–23% better performance on the test data. The remaining issue is the existence of relatively high and infrequent errors that appear for transient operation during mornings and evenings.

Abstract: Heat pumps are very important for improving the energy efficiency of heating, cooling, and hot water generation, but also for increasing the utilization of renewable energy and low-temperature heat sources. The design and optimization of energy systems with heat pumps might lead to better results in terms of their energy efficiency and cost attractiveness. This paper extends a previous research and analyzes primary energy consumption of a hybrid heating system that consists of heat pumps, heat storage, and a connection to the district heating network. It assumes that the system is running in a cost-optimal operating regime and considers four cases of electricity pricing: (1) flat tariff, (2) time-of-use tariff, (3) progressive tariff, and (4) combination of time-of-use and progressive tariff. The cost-optimal operation regimes are obtained with mixed integer linear programming. The results indicate that the primary energy consumption considerably depends on the electricity pricing scheme, mainly because the time-of-use tariffs offer electricity for very low prices, but also because progressive tariffs might limit the application of a heat pump with high electricity prices in the most expensive tier. It is also important that the increase in the price of district heat makes heat pumps more attractive and contributes to the primary energy reduction, under the examined set of conditions.