Abstract: This paper presents a concise literature overview in the field of biofeedback applications and systems in sports and rehabilitation. We searched for papers in three research databases: Scopus, Web of Science, and PubMed. The initial search yielded 1728 papers. After a rigorous and exacting elimination process, 112 papers were finally included into this literature overview. We focused on full papers describing applications and systems with a complete real-time feedback loop, which includes the use of sensors (inertial, motion, force), real-time processing, and feedback to the user (visual, auditory, haptic). A number of research questions were raised and papers were studied and accordingly. We present the first results of this literature overview.

Abstract: Anomaly detection is a very important class of data processing techniques and as such has found wide applications in many different fields. One of the problems often faced when developing systems for anomaly detection is the absence of labeled datasets, which means that unsupervised learning techniques need to be developed. Anomaly detection is most often used on data that exhibits some type of locality. This, in effect, means that anomalies are context-dependent. This paper analyzes existing approaches for extracting context information and defines a novel approach based on a bidirectional recurrent neural network. This paper defines an end-to-end system that is capable of autonomously flagging anomalies with no human interaction needed.

Abstract: In the traditional manufacturing industry, technical documentation comes in various forms and plays an important role at all stages of a product development and production. In recent years, the idea of Industry 4.0 has been popularized. One of the core elements of Industry 4.0 is a digital twin representing a virtual model of physical objects and processes in a factory. An initial version of the digital twin is based on the documentation, and all changes that later occur in the lifetime of the digital twin must also be reflected in the corresponding documents. Current forms of the documentation are unsuitable for both the efficient creation of the initial digital twin version, and for consequent updates. To bridge this gap, a novel approach is proposed in this paper. The core element of this approach is a domain-specific modeling language – MultiProLan. Its models serve as centralized representation of knowledge about the production processes. Different types of technical documentation can be automatically generated from the models, resulting in increased document consistency, quality and minimized number of errors introduced by humans. We also explore the possibility of automatic generation of different types of technical documentation from such models, as well as different production process modeling aspects that must be covered.

Abstract: Being an engaging physical activity, creative dance requires high levels of particular skills and body motion control. Automatically detecting dancing tempo has the potential of supporting many practical applications, from assessing dance move timing and overall performance to monitoring progress in the learning process. In this article, we present and evaluate a dance tempo estimation methodology that resides on a single, orientation independent, leg-attached 3D MEMS accelerometer. As the solution avoids using video cameras or IR imaging sensors, it is computational efficient, user-friendly and applicable for a variety of dancing situations - when the dancer is dancing alone, in the crowd, or in front of an audience. We focus our attention on Solo jazz dancing. Initial validation performed for a professional dancer dancing at seven different tempos ranging from 120 to 240 bpm with a 20 bpm tempo increment showed that the methodology is accurate up to 0.5 bpm. Further investigation, including different professional and amateur dancers, dancing following actual jazz music and not only metronome tempo is necessary in order to fully evaluate the presented methodology.

Abstract: Knowledge of legislation and regulation is of great importance, especially to legal workers. However, finding a document that contains the relevant information is not a trivial task. Existing services that enable storage, retrieval, and browsing of legislation and regulation usually offer simple search features. They do not utilize metadata to the fullest possible extent and do not link metadata with other data published on the (semantic) web. In this paper, we present a software solution based on linked data principles and technologies that offers key features that the existing legislation and regulation retrieval and browsing services have but also introduces advanced features such as advanced search using document metadata and better handling of document references and links. The content is semantically enriched by using data from DBpedia, a project aiming to extract structured knowledge from the content created in various Wikimedia projects, which makes it more usable by other systems. However, the solution has its constraints because it requires documents to follow a specific legal document format.

Abstract: The human ilium bone, as an entity of the hip bone, represents a very complex morphological structure of irregular shape. Building accurate subject specific 3D model requires complete geometric morphometry of given bone. Therefore, it is necessary to define reliable anatomical landmarks as points which have unique and consistent positions at the each bone, defined by coordinate values. Due to the fact that the number of these point depends on the complexity of the shape, in the case of the ilium bone 15 bilateral landmarks are separated. Based on these landmarks 26 parameters are determined as linear distances. Using statistical approach it is possible to predict coordinate values for 11 landmarks, based on the values for 4 points whose positions are easy to determine. Input data in the form of coordinate values are taken from anatomical points, localized at the sample of 32 polygonal models of the ilium bone. The tools of descriptive statistic and regression analysis are used for establishing proper dependencies between coordinate values, which results in 33 mathematical equations (6 linear, 18 squared and 9 logarithmic). These results are statistically significant, due to the the value of variance R2 (up to 0.83511) and the p-value which is less than 0.01 for regression coefficients. Based on measured and predicted coordinate values it is possible to calculate values for all parameters, using an expression for distance between two points in 3D, in proper Graphical User Interface (GUI) developed for the purpose of this study. The results of study, tested on randomly chosen male hip bone, proved proper accuracy. Landmark-driven approach presented here allows simple and fast prediction of the subject specific morphometry as the first step in building 3D bone surface model.

Abstract: This paper provides an analysis and comparison of the most important characteristics of entropy-based techniques and different categories of machine learning approaches. The main goal is to better understand the existing techniques and results, which can be found in a wide range of scientific studies. Although both classes of approaches rely on the network traffic structure analysis, the entropy-based techniques are inherently much simpler for the application, and in some cases can easily deal with the detection of the zero-day attacks. On the other side, due to the lack or presence of the labels, the machine learning algorithms can be applied for specific attack cases, and under such conditions providing more accuracy when compared with some entropy-based methods.

Abstract: Data in the energy domain grows at unprecedented rates. Despite the great potential that IoT platforms and other big data-driven technologies have brought in the energy sector, data exchange and data integration are still not wholly achieved. As a result, fragmented applications are developed against energy data silos, and data exchange is limited to few applications. Therefore, this paper identifies semantic models that can be reused for building interoperable energy management services and applications. The ambition is to innovate the Institute Mihajlo Pupin proprietary SCADA system and to enable integration of PUPIN services/applications in the European Union (EU) Energy Data Space. The selection of reusable models has been done based on a set of scenarios related to electricity balancing services, predictive maintenance services, and services for residential, commercial and industrial sector.

Abstract: Observation of air pollution levels at certain points in space and time is done by using mobile and static sensor networks. The values of air pollution levels at points where no measurements were made are mostly assumed by numerous types of interpolation between known values at measured points. The authors of this paper propose techniques for predicting air pollution levels in points in space where there are no measurements. The proposed techniques are based on the analysis of measurements from the sensor network that are affected by the same sources of pollution. Three approaches for identifying unknown air pollution sources by collecting measures from sensors mounted on public service vehicles are defined, implemented, and evaluated. The first approach can be treated as the optimization problem, the second approach is based on clustering in a multidimensional space and the third one is a fast and light method for a specific simplified case of the problem. The system is also implemented for a distributed computer cluster that applies machine learning algorithms over data streams for efficient estimation of dominant pollution sources in real-time.

Abstract: Total hip arthroplasty is considered as one of the most successful surgeries. Cementless hip implant surface structure has been identified as a significant factor when choosing the implant that will be inserted in a body. Our aim was to numerically analyze how three different Ti-6Al-4V hip implant surface topographies affect the shear stress distribution under the static load corresponding to single leg stance. Finite Element Analysis was performed for three developed hip implant and bone models, using material properties and boundary conditions adapted from literature. Based on the criteria that shear stress at the implant-bone interface should be minimized to promote bone ingrowth, we were able to conclude which model would be the best choice for the hip implant surface topography as well as to determine if location of surface topography influences the shear stress results.

Abstract: This paper presents development of an automatic diagnostic tool based on machine learning that analyses cardiac ultrasound images of patients with cardiomyopathy in several views (4 chamber apical, 2 chamber apical and M mode view). The main aim of the developed tool is to perform automatic left ventricle (LV) segmentation and to extract relevant parameters in order to estimate the severeness of cardiomyopathy in patients. Dataset included 1809 images with apical view and 53 images with M view from real patients collected at three Clinical Centers in UK and Serbia. Separate methodologies have been implemented for analyzing apical and M mode view, including U-net for segmentation, after which parameters such as left ventricular length (LVL), internal dimension (LVID), posterior wall thickness (LVPW) and interventricular septum thickness (IVS) are calculated, both in systole and diastole. The tool has also been implemented on the platform with a user-friendly interface, which allows these two modules to be used either separately or combined. In order to validate the model and compare the results between gold standard and developed methodology, two cardiology specialists have independently manually annotated LV and measured relevant parameters. The results show that the model achieves dice coefficient of 92.091% for segmentation and average root mean square error (RMSE) of 0.3052cm for parameter extraction in apical view images and average RMSE of 1.3548cm for parameter extraction in M mode view. Fully automatic detection of cardiomyopathy in cardiac LV ultrasound images can help clinicians in supporting diagnostic decision making and prescribing adequate therapy.

Abstract: In the era of personalized medicine and improved prediction power, the cardiovascular diseases such as carotid atherosclerosis have to be analyzed by using advanced machine learning techniques in order to better estimate the patient’s condition and decrease the risk from severe catastrophic events over time, such as stroke and Transitional Ischemic Attack (TIA). Among various diagnostic imaging techniques, the US images were used in this study for the detection of segments of the carotid artery as this technique is low-cost and widespread in examination of carotid atherosclerosis. The clinical US images were analyzed by using deep learning techniques. The automatic segmentation of carotid artery’s lumen has been done by using the U-Net based deep convolutional networks. The US images of carotid arteries underwent the preprocessing, such as resizing and classification of US images, as well as the annotation of lumen area. The whole dataset was randomly divided into, training, validation and testing sub-sets, while the model robustness was tested on unseen images (test dataset). The obtained results show high accuracy in region detection and segmentation (Precision - 0.90, Recall - 0.92, Dice coefficient (F1-score) - 0.91). The automatic extraction of US features such as carotid lumen gives the specific segmentation of individual patient-specific anatomy and can be used for further analysis of the patient, as well as for building up the patient-specific models for computational CFD simulations.

Abstract: The world stent market has an estimated value of €6.4 billion, of which 37% is generated in the US and 10% in the EU. Coronary stents are now the most commonly implanted medical devices, with more than 1 million implanted annually. Coronary stents are currently the most widely used for treating symptomatic coronary disease. This paper presents the innovative platform and its separate modules that can be used as a standalone tools developed within EU funded project InSilc, for designing, developing and assessing coronary stents. This integrated solution was developed by international consortium in H2020 project www.insilc.eu. After three years of intensive work on different modules, presented in this paper and their integration in the platform which will be offered as a service to interested stakeholders, consortium is planning the commercialization of the solution. In this paper the potential pathways for further exploitation and commercialization are presented together with market analysis, business plan and business model. The technological level of each module is also elaborated, as well as different business models for its market introduction. Future work will include testing and validation on larger databases, and improvement of the tool based on feedback from the users from different categories (stent industry, clinicians, researchers etc.). The main goal is commercialization of the InSilc solution and its fast market uptake.

Abstract: Since the Coronavirus outbreak in December of 2019, people have put a lot of effort into its early diagnosis by assessing signs and symptoms obtained from various studies. To this goal, deep learning models are trained to detect the presence of COVID-19 from three most commonly used lung-imaging modes: X-Ray, Ultrasound, and Computed Tomography (CT) scan. This survey compares current Deep Convolutional Neural Networks approaches for the classification of CT scans considering datasets, neural network architectures, and evaluation metrics. It gives new researchers in this field an excellent starting point providing currently most popular and prosperous CNN architectures and large available diverse datasets.

Abstract: Internet of Things and Blockchain are considered two major technologies. Internet of Things is facing many challenges such as poor interoperability, security vulnerabilities, privacy, and lack of industry standards. Most Internet of Things Devices needs a constant connection to the internet which brings many challenges in their protection and security the data. This is where Blockchain technology comes into use. The Blockchain provides decentralization and authentication that makes is impossible for third parties to gain access to the network. It provides much needed privacy and flexibility which is currently missing from IoT infrastructure. Our work provides a solution with overview what can be done to implement Blockchain in existing IoT infrastructure and furthermore improve security and privacy of the system.

Abstract: The goal of this study is to develop an expeditious methodology for designing an orthosis which can be used both for wrist fracture immobilisation and later rehabilitation. This requires an orthosis that can (depending on the patient’s needs) be flexible or fully stiff. It is possible to achieve this by making the orthosis out of a flexible material and by adding external plates for stiffness (which can be removed when the orthosis is used for rehabilitation purposes). Another reason why the orthosis should be made of a flexible material is to ensure that it can adapt to changes in swelling that may appear during the healing process. The design of the orthosis is expeditious in order to facilitate possible practical application of the methodology in clinics.

Abstract: The basic setting of the paper is related to the improvement of the current way of decision-making by using the business intelligence system, important for decision-making on a specific problem, and in order to achieve benefits in decision-making at the highest level. The improvement must be appropriate to the development of the BI system in conditions when there is already previously accumulated knowledge about the problem. The obtained results should show that the business intelligence system can identify dangerous places on the roads, as the first and very important stage in the process of managing "black points", and is a procedure for detecting specific locations on the road network where appropriate concrete measures need to be taken.

Abstract: This paper represents a teaching approach to the transformation of classical lectures teaching. Paper elaborates use of the online platform that enables students with and without disabilities to follow classes without hindrance during the lecture period. The main advantage of this kind of teaching is the possibility of attending classes from any location and from any device, it is only important to be connected to the Internet connection. After the lecture, students also have the ability to view video and presentation materials. This paper describes a new approach to teaching and illustrates the expected benefits of online teaching. Full integration with the already existing Faculty Information System has been performed. The platforms used in this integration are Microsoft Azure, Microsoft Office 365 Admin, Microsoft Teams, Microsoft Stream and Microsoft SharePoint.

Abstract: Smart homes consist of electronic devices that consume electricity from the electricity grid (EG) or renewable energy sources. The system proposed in this paper aims to lower the cost of consumed energy in smart homes. Cost reduction can be achieved by training a smart HEMS (House Energy Management System) to orchestrate the schedule of loads energy consumption according to the time-varying energy price and the residents' preferences. HEMS can be trained by Reinforcement Learning (RL) if provided a realistic environment. In this paper, we propose a simulated household environment with the help of the Typhoon HIL application. To make our simulated environment realistic, we need realistic measurements of external conditions, such as external temperature and solar irradiation. Thus, we use the Mainflux platform that supplies the simulated environment with real-world data. This paper focuses on integrating the Mainflux IoT platform with Typhoon HIL simulation of smart home devices. In this paper, Mainflux provides two real-world parameters: solar irradiation and outdoor temperature, vital inputs for the realistic simulation of smart home devices such as PV panels and Air Conditioners.

Abstract: For each trading day, there is a huge number of trade events registered at the stock market, and thus a large volume of data is recorded. In this research, machine learning approaches are particularly employed in order to learn from stock market data. More precisely, in order to capture the time dependency between rows in the market data, the model based on the Long short-term memory (LSTM) network is employed.