Abstract: Based on literature data in recent years the incidence of invasive fungal infection (IFI) caused by molds is on the rise. Diagnostics of those infections can sometimes be inefficient; they require a longer period of time in laboratory procedures and sometimes may lead to late diagnosis or misdiagnosis, which can result in patient’s critical condition or even mortality. Goal of this research is to train a classifier, a neural network model that can be used for classification of molds from patient materials, which can drastically improve the process of diagnosis and prevent fatal cases. Using a ResNet-50 deep convolutional neural network (CNN) and images obtained from Department of Microbiology and Immunology, Medical faculty, University of Niš, Serbia, archives, a classifier has been developed, displaying promising results, which show that with wider dataset it will be possible to train a model that can be used in diagnostics.

Abstract: The use of modern technologies in all spheres of life offers the possibility of collecting and processing citizens' data in smart cities to improve the quality of their lives through creating smart health services. The collected data should be protected from misuse. This paper describes e-health services in smart cities, as well as the problems of ensuring durability and reliability of patients' data. As a solution to this problem, the use of blockchain technology has been proposed and ways for its application have been given.

Abstract: Modern web applications, as a necessity, require fast, efficient, and accurate data exchange, which all participants in communication can uniformly interpret. Consistency in interpreting exchanged data can be achieved using text, self-describing and platform-independent file formats such as XML and, recently, the even more popular JSON. JSON represents a common model for NoSQL databases. Also, it has been supported for years in relational databases from leading manufacturers, such as Oracle. This paper systematises, analyses, and compares the functionalities for working with JSON across different versions of Oracle DBMS, starting with version 12c R1, which introduced support for native JSON, concluding with the recently introduced version. For selected use cases, experimental tests were performed, which demonstrated the impact on performance achieved by the supported JSON functionalities from version 12c R1 to 19c of Oracle relational database management systems (DBMSs).

Abstract: Nowadays, the modern world strives to live life healthier. With the influence of the virus COVID-19 on everyday life, it is difficult to maintain motivation and training routine. By using technology, overcoming those difficulties can be easier. This paper presents the implementation of the system for helping users during exercise, filling the role of a digital personal trainer. The system mimics the reasoning of a human expert in proposing a suitable exercise plan based on the current user's condition. Also, the system supports the process of monitoring the current user's status during exercise, acts on changes, and shows certain messages concerning the displayed status. The system uses rules to determine which exercises are most suitable for a particular user, as well as the messages that will be shown during the exercise. The system is implemented using a system based on rules.

Abstract: Technological innovations such as M-Learning, automatic graded systems, virtual and augmented reality, online classes, and the gamification of learning are rapidly being integrated in the field of contemporary education. One of these technologies is Blockchain, which has already proved to be a disruptive in several fields, such as finance, healthcare, human resources and advertising. Multiple organizations from different industries are developing blockchain-based systems, but the application of Blockchain in education is still novel. The goal of this paper is to present an overview of existing literature in this field, and the potential of application of Blockchain technology in education. We present a simple solution of an eLearning system for learning programming on the basis of automatic grading developed in the language Python.

Abstract: Recommender systems are used for predicting user preferences mostly for commercial purposes. As the need for recommender systems has increased significantly, due to the rapid lifestyle we all lead, machine learning has also found its wide application in the development and improvement of these systems. Reinforcement learning emerged as one of the newer machine learning techniques, which can make recommender systems more flexible using user interactions. This paper focuses on analyzing and comparing existing implementations of recommender systems based on reinforcement learning, and drawing final conclusions when combining these two techniques, as well as defining possible further upgrades.

Abstract: In data analysis, outlier detection is very important for tasks such as filtering data or finding points of special interest. Finding outliers can be quite challenging task, especially when data set represents human behavior or certain action, which can be hard to describe or predict. This is also case in OpenClick project, a platform which provides various tests where human behavior can be analyzed through human-computer interaction. For a single test type on OpenClick project, an individual test can be invalid for various reasons, e.g., a disturbance occurred during a test (such as ringing of a phone) or perhaps a test subject cheated in some way. One way how could we distinguish invalid test from valid is through detection of outliers in data set using outlier detection methods. For this purpose, we chose five outlier detection methods, which we use to detect suspicious data points and to evaluate tests in terms of how much outliers bring disturbance in test results. In addition, we give some observations on the form of the data set of a single test and propose possible way on how to approach a task of labeling tests as valid or not.

Abstract: Power flow calculation is one of the fundamental functions of the power system management. Traditional power flow calculations have been performed using singlethreaded technique. In recent years, multi-threaded implementations of the power flow calculation procedures have become increasingly popular as they decrease calculation speeds. The multi-threaded power flow calculation is the main subject of this paper and the main objectives are as follows: (i) to present the model of the power flow calculation, (ii) to identify which parts of the proposed algorithm could be implemented in the multithreaded manner, and (iii) to develop the general multithreaded power flow procedure for large-scale weakly-meshed and meshed networks.

Abstract: In this paper we propose an approach for developing digital twins of highly complex, dynamic physical systems that interact with numerous external systems. The platform utilizes microservice architecture that allows separate loosely coupled services for representing and storing Digital Twins and reasoning over them. The current version of the platform supports ontological representations of digital twins. Asynchronous communication is achieved through message queues. The scalability of the system is met by utilizing the cloud solutions. The approach is evaluated by building a prototype platform and a Digital Twin of a smart building.

Abstract: There is a need to develop a disc herniation decision support system, as lumbar disc herniation is one of the most prevalent causes of intervertebral disc diseases, accounting for 90% of all spine surgeries. Image processing techniques and artificial intelligence algorithms can be efficiently used to describe a strong relationship between the MRI of the lumbar spine disc and final diagnosis. This study presents a methodology for automatic segmentation of the region of interest (ROI) - disc area, after which distinct features are extracted and can be used for diagnosing disc herniation on axial MR images. U-net convolutional neural network is used for ROI segmentation, after which features such as moments, eccentricity, equivalent diameter, eigenvalues of the inertia tensor and ratio of bounding rectangle area and disc area are calculated. Additionally, centroid distance function was used to visually differentiate between the herniated and non-herniated discs. The methodology was based on the number of peaks on the graph plot of distance between the disc centroid and points on the edges of the disc. The results achieve dice coefficient of 0.961 and IOU of 0.925 for segmentation on axial test images, while eigenvalues of the inertia tensor proved to be very descriptive in differentiating herniated and non-herniated discs. The results from these tests can be regarded as an early step toward establishing a fully automated system for identifying lumbar spine disc herniation in future work.

Abstract: Court decisions represent outcomes of court cases and their number constantly increases. These decisions are usually published online allowing lawyers and other interested citizens to be informed on the work of courts. The size of published datasets can bring some difficulties to the retrieval and analysis of court decisions on a particular legal issue. Extracting data from court decisions and identification of cited legislation, regulation, and case law would allow better data organization to more efficiently retrieve, browse and further process the data for statistical analysis. These data could also help in detecting potential inconsistencies in decisions delivered by courts. This paper proposes an approach consisting of various NLP techniques for extracting relevant data from New Zealand’s court decisions and using them to construct a system of linked legal documents for better navigation through legal cases.

Abstract: In order to establish safe operation of the railway, it is very important to maintain vegetation on and near the railway. In this paper, a solution based on deep learning for detection of weeds and bushes on the railways and the control of intelligent herbicide sprayers using PLC, is presented. Hardware setup, software implementation, the configurations of the convolutional neural networks, and the datasets used to train the neural networks are described. A video stream with recording of the railway tracks in front of the train, is sent to a PC computer located in the cabin. This video stream is processed using software for detection of weeds and bushes based on multiple convolutional neural networks. This software sends signals to the PLC that controls the opening of the nozzles on the herbicide sprayer. The presented solution is developed for Serbian Railways company.

Abstract: This paper presents an implementation of the System for Software Testing in Real-Time Simulation Environment. The main goal of this system is to allow real-time simulation of software in an isolated environment according to hardware-in-the-loop concept. Due to numerous challenges in development of software that run on devices that process lots of signals and that are supposed to be integrated in large systems, hardware-in-the-loop testing has emerged as standard technique for their validation by simulating their work as if devices were in real environment. Due to the nature of the testing process, only one simulation can be executed at a time and usually the same environment is reused for different setups or different processes in the development of different components of system. This proposed system allows having clean environment with only necessary dependencies for a specific setup and allows running multiple simulations at the same time. In addition, it provides tools for observing reports of finished test sessions and searching executed tests by different parameters.

Abstract: Paper describes a test automation system modeled through the master-agent architecture, where the master node defines tasks as a part of the testing workflow and delegates them to the agent node. Agent node is the one that does the actual work, which is running the tests. After finishing the test session, agent generates descriptive report about the executed test session and sends it back to the master node, where it is available to the end user. The agent part of the system is implemented within containerized and controlled environment. In addition to the implementation details, the observed shortcomings that led to the expansion of the system were highlighted with suggestions for possible future expansion of the system.

Abstract: Cognitive disorder is a condition severe enough to compromise social and/or occupational functioning. It has a huge social significance to all the parties involved in their diagnosis and treatment. The timely diagnosing and early treatment have a great social and economic importance. The neuropsychological and neuropsychiatric assessments provide a solid base for the diagnosis and prediction of cognitive disorders. The present research work aims to find dependencies between factors and conditions for disease’s recognition. It is based on data collected for three very commonly applied examinations for functional, mental and neuropsychological assessment. Clustering analysis is applied as appropriate datamining technique. The results show that the functional assessment data present more clear structuring than the separation ability of the other examinations. The other part of the research conclude that the diagnosis’s prediction based on groups found by the clustering analysis is improved by analysis of the merged datasets instead the individual ones.

Abstract: After briefly reviewing most recent definitions of digital twin concept in the context of Industry 4.0, this paper focuses on the concept applied in the AEC (Architecture, Engineering, and Construction) industry, and more precisely, in the domain of built heritage. Analyzing the various building reconstruction and restoration reports that consider the digital twin usage, we looked for its application aspects, both the ongoing and planned ones. To do so, we dominantly focused on the reconstruction-case of the Notre Dame Cathedral in Paris, because it is an actual (in progress) and well-known one and just supported by the usage of digital twin technology. Investigating that case, we highlighted various utilization aspects of the employed digital twin. Due to complexity, this case study permitted to scientifically draw conclusions and generalizations that could be applied broadly in the field of interest. This study underlines that the digital twin concept becomes a sustainable and inevitable technology which allows to integrate ICTs into the interactive processes of numerous building operations and, thus, very promising one in the field of built heritage preservation praxis.

Abstract: The Digital Twin (DT) as a virtual representation of physical assets aims to support transition to smart manufacturing and Industry 4.0. Today, DT remains challenging to implement on production systems because of its original technical complexity, various functions and purposes. Recent researches are focusing on Asset Administration Shell, a promising concept allowing a systemically defined and semantically described components to form distributed interoperable systems according to Industry 4.0 principles. The paper explores benefits of Asset Administration Shell and Model-Based Systems Engineering approaches to provide a conceptual framework of DT implementation in production systems.

Abstract: Everyday activities can lead to damage in our cartilage. Focal cartilage lesions have been associated with the progressive degeneration of the surrounding cartilage tissue which could lead to additional health problems such as early-onset osteoarthritis. Our aim was to compare the mechanical response of the knee joint with damaged femoral cartilage for normal and high body mass index (BMI) values during the stance phase of the gait cycle, using the finite element method. The location of the hypothesized lesion was above the anterior section of the lateral meniscus. Comparison of the obtained results has shown a significant influence of BMI on the Von Mises stress values, even in the case of simplified material properties. Keywords knee joint, femoral cartilage, focal lesion, finite element methods, stress distribution

Abstract: Melanoma is one of the deadliest forms of skin cancer and a global health problem. Early and confident detection of melanoma increases chances of survival. Currently, there are various methods applied in medical practice for computer-assisted diagnosis of melanoma. This paper considers the need for mobile-based decision support system for melanoma detection. It includes the description of a prototype based on state-of-the-art machine learning methods and it proves there is a possibility of developing a system that serves as an early detection tool. The prime goal of the system is to classify skin lesions as benign or malignant, helping practitioners increase confidence of early examinations. The system consists of an Android application for clients, the database, skin lesion classifier and data exchange service on the server side. Three public available melanoma image databases have been used to develop this classifier, which is now able to achieve accuracy of more than 91%. The system was tested in practice and its accuracy confirmed comparing with the experts on another dataset and in a prospective study at the clinic. To the best of the author's knowledge, this is the first mobile automated supporting system for melanoma detection based on deep neural networks.

Abstract: Atherosclerosis is a local inflammatory disease characterized by maladaptive build-up of lipids, leukocytes, cholesterol, cellular waste products and extracellular matrix inside the artery wall. The described components contribute to plaque development and progression, each of them with proper features and rule-based behavior. Plaque progression is determined by the interaction between these components and the environment in which these components evolve. Agent-based modelling (ABM) is selected as an adequate approach to reproduce the evolution of plaque progression by simulating the behavior of autonomous cellular agents. The agent-based model proved to be a reliable model that is able to predict the history of atherosclerosis development and progression accurately. Our proposed method is based on 2D modelling of cross sections of artery wall and plaque development. Also, our atherosclerosis agent-based model included different characteristics such as behavior of cells and lipid dynamics in a variety of vessel cross-sections. Cell behavior which had a substantial impact on the lumen area, have the highest influence on our model. INTRODUCTION Atherosclerosis is a local inflammatory disease characterized by the recruitment of atheroma (plaques) in the arterial wall. Plaques can be comprised of derived foal cells, lipids, fatty substances, cholesterol, cellular waste products, elastin, collagen, fibrin, calcium and other constituents. The rate of production of these constituents are not identical in different stages of the disease progression. In the beginning, the plaque progression induces artery outwards remodeling to accommodate the volume of the plaque without lumen area reduction. After the initial progression, the remodeling happens towards inside the lumen causing the lumen narrowing which is called stenosis (Figure 1). Figure 1. Different stages of the plaque formation and progression. Depending on the type of the plaque, some can be vulnerable to rupture and cause secondary vascular, sometimes even lethal, outcome for the patient. Initial damages to the endothelium, which is a selective layer, cause an increase in permeability of the layer. Fats integrated into lipoprotein LDL particles are absorbed into the intima by passing the endothelium lining. Some of these LDL particles become oxidized and this attracts monocytes, which differentiate into macrophages that uptake the particles. Due to these activates, cytokines are released, which in turn attract more monocytes. Artery wall is at fatty streak stage now. In some regions of increased macrophage activity, macrophage-induced-enzymes erode the fibrous membrane beneath the endothelium so that the cover separating the plaque from blood flow in the lumen becomes thin and fragile, vulnerable to rupture. The problem stated above describes numerous components contributing to plaque creation and progression, each one with proper characteristics, behavior, and rulesets. The interaction between these components and the environment in which they evolve determines the plaque progression. For this reason, we selected ABM as the proper approach to mimic the evolution of plaque progression and artery reshaping (dynamical system) by simulating the behavior of autonomous cellular components (agents). We hypothesize that ABM model represents a reliable model that can describe properly history of atherosclerosis development. When the response of complex biological systems strictly depends on cellular behaviors, which in turn are influenced by the changes in the micro-environmental factors, a multi-scale modelling approach is preferable. Pappalardo et al. [1] developed agent-based model to describe the very early stage of the atherosclerosis, the stage before formation of a calcified plaque. Deo et al. [2] also used an ABM to simulate inflammatory processes in atherosclerosis, particularly focusing on plaque formation and rupture. Curtin et al. [3] instead developed a two-dimensional grid space ABM to simulate the restenosis development in a blood vessel following an angioplasty and bare-metal stent implantation. They showed the body’s response to the intervention and explored how different vessel geometries or stent arrangements may affect restenosis development. Starting from the post-procedural vessel lumen diameter and stent information, they generated the final lumen diameter, the percent change in lumen cross-sectional area, the time to lumen diameter stabilization, and the local concentrations of inflammatory cytokines. Olivares et al. [4] also developed a 3D ABM to simulate relevant cellular and molecular phenomena involved in the early formation