Hybrid Event

Event program

Wednesday, 9/30/2020 3:00 PM - 5:00 PM, Camelia 2, Grand hotel Adriatic, Opatija DS - Data Science  3:00 PM - 3:30 PM Invited Lecture  • Z. Šojat (Ruđer Bošković Institute, Zagreb, Croatia) Information Communication Society - a Possible Ecology of the Future  As Henry A. Giroux recently stated, the situation caused by the present pandemic of the Novel Corona Virus (SARS CoV-2) is not only a medical crisis, but even much more a "Socio-Political and Ideological crisis", "crisis of Greed" "and Pedagogical crisis". Consequently it is now absolutely essential to develop our Collective Imagination to produce possible Visions of the Future in which a novel global system of ecological production, distribution and usage of long-life high-quality products will only by necessity support the basic economic local production, distribution and usage of human survival essentials. In this sense an Information Communication Society, heavily leaning on organisational "skills" of computer and networking applications, will have to be envisioned, developed, organised and implemented on the global scale. The Information Communication Technologies and Computer Science developments will have to heavily concentrate primarily on enabling such an Information Communication Society, a future global society of local societies based on Kairology (the science of the right moment) and Synergy (co-operation) and rooted in proper present and future developments of the knowledge of Cybernetics (the science of dynamic systems) and Ecology (the science of the whole). 3:30 PM - 4:50 PM Papers  1. M. Marjanović, T. Kramberger, R. Kramberger, I. Cesar (TVZ, Zagreb, Croatia) Evaluating Microsoft Face API in the Context of Student Classroom Attendance  Face recognition technology is having a tremendous impact on the vast majority of today's human activities – mostly for leisure and security purposes. In this paper, we evaluate Microsoft Face API technology in the context of calculating classroom student attendance. Also, a short overview of face recognition technology is presented, alongside common algorithms used in today's state-of-the-art face recognition systems. Finally, Microsoft Face API is evaluated with respect to image quality, blur, occlusion and in the experiment with an attempt to evaluate the efficiency of the recognition system in the context of student classroom attendance. 2. I. Prazina, I. Ivkovic-Kihic, T. Chahin, A. Jajcanin, S. Rizvic, V. Okanovic (Faculty of Electrical Engineering, Sarajevo, Bosnia and Herzegovina) Sarajevo War Tunnel - Mobile Virtual Cultural Heritage Application  This paper explores the new way of presenting one existing VR application, which was described in our previous work - Virtual Reality Experience of Sarajevo War Heritage. The goal of the application was to introduce more people with the Sarajevo siege and allow them to experience the Tunnel crossing at that time. Before this application, we made two versions,the first one for VR setup and the second for the web. In this paper, we introduce a mobile version with the same content. The challenge was to optimize the content for the mobile experience. The assets were optimized so a wider number of mobile phones with different hardware capabilities can run the application. The advantages and disadvantages of this approach are pointed out, and the limitations of the mobile application are emphasized. The memory usage and frame rate are measured for different Android devices with different operating system versions and hardware capabilities. The results show the optimized application can be run on different Android mobile devices. Nevertheless, for better user experience a higher number of frames per second is needed, which may include reducing the quality of the assets. 3. D. Tomić, D. Davidović (INSTITUT RUĐER BOŠKOVIĆ, ZAGREB, Croatia), T. Šubić (Arctur d.o.o, Nova Gorica, Slovenia), J. Mesarić (Fakultet Elektrotehnike i Računarstva, Zagreb, Croatia), K. Skala (Institut Ruđer Bošković, Zagreb, Croatia) Improving 3D Printing of Garments by Using HPC Cloud  3D garment printing is a technology that is experiencing rapid development, and more and more garments are being printed on 3D printers. Advanced processes such as Selective Laser Sintering and PolyJet 3D printing are used. However, 3D printing of clothes still experiences some disadvantages, of which the biggest is the slow 3D printing speed and usually a small printing area of 3D printers. The novel, large multi-head printers can significantly increase the printing speed, but such printers require an additional step to prepare given 3D models that increase. However, pre-processing time grows exponentially with the number of printer heads, therefore computer resources that exceed the capabilities of a single workstation are required to prepare a printout. This implies the use of HPC resources, and because of its flexibility, lower cost and ease of use, HPC Cloud is optimal platform for such jobs. We have shown that the structure of these jobs fits perfectly in the HPC Cloud environment. 4. M. Babič (Faculty of Information studies, Novo mesto, Slovenia), B. Soldo (University of North, Varaždin, Croatia), J. Povh (Faculty of Mechanical Engineering, University of Ljubljana, Ljubljana, Slovenia) Public Transport Analyzing Using Bee Colony  Urban public transport is crucial to the functionality of every urban city. Good urban transport affects the functionality of the city itself and of course reduces the number of cars on the roads. The wider Novo mesto region, as a key industrial center, is heavily dependent on urban transport. Unfortunately, urban transport in Novo mesto still has a relatively weak effect on the traffic connections of the wider area. In fact, its functionality has never been systematically scientifically researched. Thursday, 10/1/2020 9:00 AM - 1:30 PM, Camelia 2, Grand hotel Adriatic, Opatija BE - Biomedical Engineering  9:00 AM - 9:30 AM Invited Lecture  • Elaine Åstrand (Mälardalen University, Västerås, Sweden) Insights into Multivariate Decoding for Estimating the Temporal Dynamics of Representations Encoded in the Brain  9:30 AM - 11:10 AM Papers  1. T. Smolčić ( University of Zagreb, School of Medicine, Zagreb, Croatia), T. Petković (University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, Croatia), S. Škokić ( University of Zagreb, School of Medicine, Zagreb, Croatia) Automatic Brain Extraction in Nissl-Stained Histological Sections  A simple and robust algorithm for automatic segmentation of Nissl-stained histological brain images is presented. The algorithm preprocesses the input image with anisotropic diffusion filter prior to applying standard binary image manipulations to isolate the brain from the background. Shape smoothing filter is applied to mimic the common practice of human evaluators. The developed code is tested on histological sections of mouse brains and shows good agreement with manual segmentation results. 2. V. Miletić, M. Ašenbrener Katić (University of Rijeka Department of Informatics, Rijeka, Croatia), Ž. Svedružić (University of Rijeka Department of Biotechnology, Rijeka, Croatia) High-throughput Virtual Screening Web Service Development and Its Applications to SARS-CoV-2 Target Protein Inhibitors Drug Design  The available structures of viral proteins and RNA molecules related to SARS-CoV-2 are used to screen and design a new set of drugs using the commercial databases and molecular docking protocols. The selected results are then described to finer details using molecular dynamics. Based on our earlier experiences we can target proteases, enzymes in DNA and RNA metabolism, and protein—protein interactions. Screening for and design of new drugs are the major aims of our work. Prior to the screening campaign, we will develop new open-source computational infrastructure, with two major outcomes. A new database containing all commercially available small-molecule ligands will be developed and deployed on the University of Rijeka. A docking server with a web-based user interface will be developed and interfaced with the compound database. The docking server will use the database for sourcing of the molecules for the high-throughput virtual screening. Our approach offers major advantages that can bypass the problems that have traditionally plagued the pharmaceutical industry: our protocols are faster, cheaper, versatile, and offer minimal risks. We are developing new drugs using commercial databases, which allows us to buy the lead compounds for affordable prices that can bypass expensive and slow organic synthesis protocols. 3. H. Petterson, J. Rehnholm, S. Vikström, M. Åslund, E. Åstrand, I. Tomasic (Mälardalen Univeristy, Västerås, Sweden) Iris Identification Using Wavelet Decomposition and Gabor Filter  Biometric authentication has seen a widespread increase in popularity as supporting technology has become common in mass produced consumer electronics. Like fingerprints, each individual has unique patterns in the iris, which makes it a common approach for implementing visual biometric authentication. The paper describes a novel system for extracting the iris pattern and using it for identification of people. The system uses Haar wavelet decomposition and 2D Gabor filtering to extract the pattern data. The pattern data is then used with bitwise XOR comparison for final identification matching. Instead of manually selecting parameters for the Gabor filter, a machine learning method called Particle Swarm Optimization was used. The parameters that gave the best matching result were then implemented in the filter design. The implemented system was evaluated on images obtained from 6 individuals in different settings. The evaluation showed that matching identification could be achieved for the database used. The prepossessing of images with Independent Component Analysis was also used to remove the reflections on the images but that did not improve the classification significantly. Still we were able to perfectly distinguish between the individuals. Further preprocessing and a larger training database would be required to get more general and robust results. 4. R. Trobec (INSTITUT "JOŽEF STEFAN", LJUBLJANA, Slovenia), M. Lindén (Mälardalen Univeristy, Västerås, Sweden), M. Šinkovec (University Medical Centre, Ljubljana, Slovenia) Impact of Subthreshold Transcutaneous Auricular Vagus Nerve Stimulation on the Heart Rate Variability and Atrial Arrhythmias  Dysfunction of cardiac intrinsic autonomic nervous system could be one of possible triggers for atrial fibrillation. Different studies show that it is possible to modulate the cardiac intrinsic autonomic activity with electrical stimulation through the skin in the ear, innervated by auricular branch of the vagus nerve (tVNS). The aim of our study was to determine, weather tVNS reduces the burden of atrial arrhythmias in patients with paroxysmal atrial fibrillation. A randomised double-blind study, with one-hour daily subthreshold tVNS, over three-day blocks of sham or true stimulation, was performed. A continuous single-channel ambulatory ECG was recorded. Significant technical assistance to computer-based ECG interpretation was necessary for highest possible interpretation quality. The obtained results were compared between periods with true and sham stimulation to discover eventual differences. We have successfully validated the designed methodology. The subthreshold tVNS has a detectable impact on heart rate variability. However, there was no significant impact of tVNS on the burden of atrial arrhythmias. A 3-day tVNS is probably too short for such evaluation. 5. A. Jović, I. Stančin, K. Friganović, M. Cifrek (University of Zagreb Faculty of Electrical Engineering and Computing, ZAGREB, Croatia) Clinical Decision Support Systems in Practice: Current Status and Challenges  Decision support systems (DSS) are computer programs based on artificial intelligence methods that contribute to reaching a correct decision in an often narrow domain of interest. Clinical decision support systems (CDSS) are such DSS that may be used by medical professionals in clinics and hospitals. They are used for diagnosis, treatment protocol recommendations, treatment outcome predictions and other tasks. CDSS are constructed based on symbolic, connectionist or hybrid approaches to represent medical knowledge. The aim of this work is to provide an overview of past and current methods in designing a successful CDSS. The study only considers the systems that were claimed to be implemented in clinical or primary care practice. Currently, the development of CDSS is mostly pursued in two directions: 1) a more traditional approach based on rules, ontologies, probabilistic models, and the use of standards; 2) deep learning based approach. Both approaches may be used complementary within a healthcare information system. This work seeks to provide an objective view on the advantages and limitations of the approaches as well to discuss future research avenues that could lead to more accurate and trustworthy CDSS and improved healthcare. 11:10 AM - 11:20 AM Break  11:20 AM - 1:20 PM Papers  6. B. Širaiy, U. Stanič, Z. Katkič, N. Grishin (School of Health Society, Domžale, Slovenia) Self-help for Elderly Individuals with Morning Gymnastics “1000 Movements” and “Method 23+5”  Medical doctors have many patients with chronic health problems such as headaches, balance issues, high blood pressure, leg, shoulder and back pain... In such cases doctors are often in a quandary, should they start with medicaments therapy, send the patient to a specialist or wait for the pain to pass by itself. The duration of diagnosis should not exceed 1 month, but if a doctor decides to send a patient to the specialist it can take quite a long time to have a final diagnosis, and that could be very frustrating for the patient. To make it easier for doctors and patients alike original selfhelp methods gymnastics “1000 movements”, and “method 23+5”, are presented in this paper. These methods have been applied to many patients with chronic health problems. The methods are based on the exercises which are understandable, simple and reasonable for everyone. The effect of exercises is multifaceted: improved blood flow to all parts of the body, relaxed musculoskeletal system, improved functioning of the organs, decreased chronic pain, and prevention of many diseases. In utilising one of the methods presented in this paper the patients becomes an active participant in self-healing, starts to believe in themselves, and last but not least they save a lot of time and money which would be spent on medical treatments. 7. M. Gusev (University Sts Cyril and Methodius, Skopje, Macedonia), E. Guseva (Innovation Dooel, Skopje, Macedonia), L. Poposka (University Sts Cyril and Methodius, Skopje, Macedonia) Design of a Non-invasive ECG-Based Glucose Measurement  Diabetic patients have to pay for each glucose reading with a blood drop and a small fortune. In addition, routine finger pricking is troublesome for diabetic patients because it can lead to scarring. It is no surprise then that the idea that glucose measurement can be done cheaply and in a non-invasive way surpasses the wildest dreams of diabetic patients. The goal of this paper is to present the design of a new technology solution for non-invasive glucose measurement based on processing the lectrocardiogram obtained via a light easyto- wear ECG monitor. We present details on how to develop a service that tracks glucose levels based on real-time ECG monitoring, and using sophisticated machine learning and related technologies. Our initial analysis shows that no similar solution is present on the market today, although several research initiatives are ongoing. 8. L. Beqiri, A. Velinov (University GD Shtip, Shtip, Macedonia), B. Fetaji, L. Loku (Mother Teresa University, Skopje, Macedonia), A. Bucuku , Z. Zdravev (University GD Shtip, Shtip, Macedonia) Analysis of Diabetes Dataset  The focus of the research study was analysis of diabetes dataset and how it will perform if we try to do a prediction of diabetes with different machine learning algorithms. We used the original dataset from the National Institute of Diabetes, and Digestive and Kidney Diseases. The dataset can be used to predict whether or not a patient has diabetes, based on certain diagnostics. For analysis we used Amazon Web Services. We used AWS S3 service to store our dataset, and Amazon Sagemaker to perform an analysis. For the given dataset we applied three classification models: Logistic Regression Model, K-nearest Neighbours and Support Vector Machines. For each of the models we also performed a performance measurement. We also compared all the results we got and according to the results, Support Vector Machines has the best performance. Insights and recommendations are provided. 9. M. Gusev, L. Poposka, E. Guseva, M. Kostoska, B. Koteska, M. Simjanoska, N. Ackovska, A. Stojmenski (University Sts Cyril and Methodius, Skopje, Macedonia) Trends from Minimally Invasive to Non-invasive Glucose Measurements  Approximately 7% of elderly over 45 are diagnosed and it is believed that 3% of the population is undiagnosed and can be classified in several diabetes forms. The usual way of glucose level measuring is by applying a blood drop to a chemically treated, disposable “test-strip” inserted into an electronic device. In this study, we give a short presentation of available products and initiatives for continuous glucose measurement based on minimally invasive and non-invasive glucose measurement methods. 10. E. Merdjanovska (Department of Communication Systems, Jožef Stefan Institute, Ljubljana, Slovenia), A. Rashkovska (Jozef Stefan International Postgraduate School, Ljubljana, Slovenia) Respiration Extraction from Single-Channel ECG using Signal-Processing Methods and Deep Learning  The measured bio-potential on the surface of the body can give most of the information on the health status of an individual. Sensors that measure the ECG potential difference on the body surface could also provide information on other vital functions indirectly, like respiration, by a customized analysis of the ECG signal. Respiration is one of the most characteristic vital signs and can reflect the status of a patient or the progression of an illness. In this paper, we utilize signal-processing and deep learning methods for the extraction of the respiratory signal from the differential surface potential of a single-channel ECG. From signal processing, we investigate feature-based and filter-based methods, while from deep learning, an encoder-decoder architecture. Simultaneous measurements of a single-channel ECG and respiration have been obtained from 61 subjects before and after cardiac intervention in several positions of the body. We also investigate the power of the methods for respiration extraction depending on the period (pre-/post-operation) and the position of the body when the signal is obtained. The results show that the deep learning approach performs better than the filter-based methods but worse than the feature-based. Moreover, we conclude that different body positions do not influence respiration extraction significantly before and after the operation. 11. V. Kadunc Kos (Clinic for Reproduction and Large Animals, Veterinary Faculty, University of Ljubljana, Ljubljana, Slovenia), M. Brložnik, A. Domanjko Petrič (Small Animal Clinic, Veterinary Faculty, University of Ljubljana, Ljubljana, Slovenia), V. Avbelj (Department of Communication Systems, Jožef Stefan Institute, Ljubljana, Slovenia) Simultaneous Phonocardiography and Electrocardiography Using Smartphone in Dogs, Cats and Horses  A phonocardiographic device was constructed using a microphone connected to a stethoscope tube and a smartphone and synchronized with a wireless electrocardiographic sensor attached precordial to the animal chest and connected to the phone via Bluetooth technology. Cardiac sounds and murmurs were recorded simultaneously with the animal’s electrocardiogram on the smartphone in 17 animals (eight dogs, six horses and three cats). Ten of these animals (six dogs, three horses and one cat) were cardiac patients with various murmurs detected on the phonocardiogram and 7 were healthy animals. The quality and synchronization of recorded heart sounds and murmurs relative to the recorded electrocardiogram was assessed in order to analyze this new prototype device in practice. Thursday, 10/1/2020 3:00 PM - 7:15 PM, Camelia 2, Grand hotel Adriatic, Opatija BE - Biomedical Engineering  3:00 PM - 5:40 PM Papers  1. V. Avbelj (Department of Communication Systems, Jožef Stefan Institute, Ljubljana, Slovenia), M. Brložnik (Small Animal Clinic, Veterinary Faculty, University of Ljubljana, Ljubljana, Slovenia) Phonocardiography and Electrocardiography with a Smartphone  Phonocardiogram, a graphic recording of auscultatory findings of the cardiac sound vibrations, was acquired through an add-on to a stethoscope: a microphone was inserted into the stethoscope tube and connected to a smart device. Phonocardiographic data were synchronised to electrocardiogram, a recording of cardiac electrical activity, that was obtained with a wireless body sensor. Simultaneous phonocardiographic and electrocardiographic data enable a much more detailed understanding of the cardiac function than auscultation of the heart sounds and murmurs with a stethoscope only. Furthermore, the possibility to study phonocardiogram synchronised to electrocardiogram seems a valuable diagnostic tool for the clinician and provides a great teaching material for students. We demonstrate a simple way to construct a device that records simultaneous phonocardiographic and electrocardiographic data, and hence enables a thorough insight into the cardiac electromechanical interactions. Cardiac activity was analysed in 9 volunteers, where timing relations of various electromechanical events were measured. 2. E. Ajdaraga Krluku, M. Gushev (Faculty of Computer Science and Engineering, Skopje, Macedonia) Detection of Uninterpretable ECG Signal Segments  Remote diagnosis represents one of the fundamental reasons for the introduction of telemedicine services. Specialized wearable health monitoring devices collect large amounts of data, which are transmitted to cloud collection centers for further monitoring and interpretation. However, the presence of noise corrupts the ECG signals, especially in wearable sensors, due to physical activities and movements. This significantly decreases the diagnosis accuracy and performance. Therefore, timely noise detection and identification of uninterpretable ECG segments are crucial for wearable devices. In this paper, we present results from our research to detect noisy segments in ECG signals without a goal to eliminate them and improve the QRS detection, but to detect where QRS detection would be impossible and avoid detection and interpretation mistakes. Our work includes two algorithms and multiple related variables that add to the success of the proposed algorithms. Finally, we achieved high performance for detecting signals where the signal to noise ratio is lower than 6 dB, with sensitivity and a positive predictive rate of over 90%. 3. E. Shaqiri (Innovation DOOEL, Skopje, Macedonia), M. Gusev (University Ss Cyril and Methodius, Faculty of Computer Science and Engineering, Skopje, Macedonia) A Programmatic Approach for Development of the ViewHRV Service Platform with Accurate and Reliable Results  There are dozens of available packages and libraries that claim to calculate HRV. This paper aims at comparing the results from the calculation of a single array of intervals between normal beats, including the most popular open source Python HRV measurement packages available on GitHub. Furthermore, the same array was ran through the Kubios software and compared to the previous results. In order to compare the accuracy of the results, as a benchmark we used the C programs provided by Physionet. The results showed a huge difference in the results with almost all the indices, in fact the simplest measurement that of Standard Deviation of NN Intervals showed to be incorrect in Kubios and in most of the Python packages. Results like these are the reason we decided to develop our own package to calculate HRV. Finally, the goal of this paper is to present details on developing a publicly available web service platform ViewHRV with guaranteed precision obtaining accurate and reliable results. 4. F. Salgueiro, P. Amado Assuncao (Instituto de Telecomunicações, Leiria, Portugal), R. Fonseca-Pinto (ciTechCare - Center for Innovative Care and Health Technology, Leiria, Portugal) Laser Speckle Stereo System in Biomedical Imaging  Speckle is an optical interference phenomenon generated by the reflection of a highly coherent beam of light (e.g., laser) on a rough surface. This phenomenon has been applied in several research areas aiming the characterization of static surfaces or the extraction of properties in dynamic processes. In particular, laser speckle has shown huge potential in the identification of high varying perfusion regions in medical images of skin and brain. The acquisition setup in laser speckle is broadly composed by a laser source, a magnifying lens and one CCD camera. From the obtained sequence of video frames several image processing based descriptors are calculated to characterize the rough surface and its varying patterns. In recent years a new acquisition perspective in laser speckle was proposed by assemble an acquisition setup using two CCD cameras, thus producing the so called stereo laser speckle. This new approach is challenging regarding acquisition and imaging processing, introducing several variables to be controlled (distance from the camera lens to the surface, direction and width of the reflected beam of light). In this paper, a revision of recent results in stereo laser speckle will be presented, highlighting the challenges in acquisition and image processing. Moreover, a new setup will be proposed to test the influence of the identified critical variables in stereo laser speckle to be applied in medical imaging contexts. 5. J. Nuhić, J. Kevrić (International Burch University, Sarajevo, Bosnia and Herzegovina) Lung Cancer Typology Classification Based on Biochemical Markers Using Machine Learning Techniques  Clinical mistreatment and mismanagement are big issue that are caused by detection of too many false negative patients. Therefore, lung cancer diagnostic inaccuracy and methods to surpass it in minimally invasive way is often the subject of research, as it is case of this study. This study focuses on the use of machine learning algorithms as a noninvasive tool for differentiation of malignant pleural effusions from benign effusions. It provides performance comparison between Adaptive neuro-fuzzy inference system (ANFIS), Support vector machine (SVM), RUS Boosted Tree (RUSBoost) and K-Nearest-Neighbor (K-NN) techniques for lung cancer detection. The proposed algorithms were chosen based on current state of the art in the field of pulmonary diagnostics. The novelty of this work is development of machine learning models for classification of lung cancer based on expression of tumor markers obtained from serum and pleural fluid. The performance of all models is compared and validated on data samples of 168 patients and linear SVM model outperformed others. 6. R. Fonseca-Pinto, E. Silva (CiTechCare, Leiria, Portugal), R. Martinho, . Rijo ( Polytechnic Institute of Leiria, Leiria, Portugal), . Januário, A. Antunes (CHL – Centro Hospitalar de Leiria, Leiria, Portugal) MOVIDA.eros : an eHealth Solution for Cardiac Rehabilitation Programs  MOVIDA.eros platform is a rehabilitation program emerging from a partnership between academia and a central hospital in Portugal. The platform combines a mobile application and a web backoffice platform to presceibe and monitoring physical activity. The system also allows changes to be made to the prescription, depending on physical performance and physiological response. These changes are automatically updated in the mobile app. The results regarding the adhesion and usability of the platform are encouraging. So far, the integration of the MOVIDA.eros platform into a classic cardiac rehabilitation program has allowed an increase in the number of patients under supervision, and has contributed to increasing the feeling of security during physical training. This program, in 1.5 years, has allowed the number of patients being monitored and treated to currently be 150, in a continuous monitoring regime (the classic program was designed for 5 patients every 3 months). In the near future, the platform will be able to be easily integrated into Cardiac Rehabilitation Programs in other hospital centers and contribute to the increase in the number of patients in rehabilitation, which is a central strategy in the management of heart diseases, with special emphasis on myocardial infarction. 7. A. Abugabah (Zayed University , Abu Dhabi, United Arab Emirates), N. Nizamuddin (Zayed University , Abu Dhabi , United Arab Emirates) Smart Healthcare Ecosystem for Elderly Patient Care  The healthcare sector is one of the rapidly growing service based sectors in the world. Constant technological adoptions have been implemented in order to provide enhanced patient care and other healthcare related services. Recently, the rapidly growing Internet and sensing technologies, cloud platforms, and remote healthcare monitoring have paved a way to build smart healthcare ecosystems. However, special attention needs to be given for delivering high quality clinical services for the growing elderly population and critically ill patients who are finding it difficult to reach out for professional medical help either due to terminal illness or because of their remote geographical location. In this paper, we propose a framework for a smart healthcare ecosystem where the patient data is collected via medical IoT sensors connected to the patient, sensing data is stored in cloud infrastructure and is analyzed by an expert from a remote telemedicine center. Moreover, an authorized telemedicine infrastructure’s person can regularly monitor the activities of the caregiver and interact with the patient without having the patient to visit the hospital. This framework aims to improve trust, reliability, and cost effectiveness of the overall healthcare service delivery. 8. D. Tomić, D. Davidović (INSTITUT RUĐER BOŠKOVIĆ, ZAGREB, Croatia), V. Janđel (Clinical Hospital Centre Zagreb, Zagreb, Croatia), J. Mesarić (Mrežne tehnologije Verso d.o.o, Zagreb, Croatia), K. Skala, T. Lipić (Institut Ruđer Bošković, Zagreb, Croatia) Drug Screening with the Autodock Vina on a Set of Kinases without Experimentally Established Structures  Virtual drug screening is one of the most widely used approaches for finding new drugs candidates. The process consists in selecting one or more chemical compounds with the highest binding free energy to target proteins. Given that the empirical space of chemical compounds is extremely large and estimated to has over 50 millions of them, finding the most effective drug becomes computationally challenging. Furthermore, the vast majority of proteins still lack the experimentally obtained 3D structures required for most of the molecular computer tools available, making it impossible to calculate their binding free energies with chemical compounds. In view of this, the aim of our study is to asses the effectiveness of the Autodock Vina tool in a large environments with unstructured proteins, those without defined 3D structure. The ultimate goal is to enable a fast and efficient virtual drug screening in such an environments, and to apply it for discovery of a new drug candidates. 5:40 PM - 5:50 PM Break  Data Science  5:50 PM - 7:10 PM Papers  9. I. Kinder, K. Friganović (Fakultet elektrotehnike i računarstva, Zagreb, Croatia), J. Vukojević, D. Mulc, T. Slukan, D. Vidović, P. Brečić (Klinika za psihijatriju Vrapče, Zagreb, Croatia), M. Cifrek (Fakultet elektrotehnike i računarstva, Zagreb, Croatia) Comparison of Machine Learning Methods in Classification of Affective Disorders  Depression belongs to a group of psychiatric disorders called affective disorders. In medical practice, patients are diagnosed according to the criteria in standardized diagnostic manuals. The criteria for diagnosing such disorders focus on the symptoms presented by the patient as well as on disqualifying other potential causes of the symptoms. Electroencephalography (EEG) is a non-invasive brain imaging technique that measures the electrical activity of the brain across different sites on the surface of the scalp. In this paper, 15 EEGs of depression patients and 15 EEGs of healthy control subjects are observed. The depressed and healthy subjects are paired according to age and gender to achieve a dataset that is balanced across classes, gender, and age of subjects. 475 different features are extracted from each EEG and used in the evaluation of different binary classification methods. The best F1-score of 0.7586 is achieved with the K-nearest neighbor algorithm. Sequential feature selection is performed, and sequentially selected features are used to evaluate the former binary classification methods. The best F1-score of 0.8750 is achieved with the K-nearest neighbor algorithm. Classification results are compared across different methods, as well as before and after excluding features that were not deemed significant by the sequential selection algorithm. 10. A. Sagdic, C. Tekinbas, E. Arslan, T. Kucukyilmaz (TED Univeristy, ANKARA, Turkey) A Scalable K-Nearest Neighbor Algorithm for Recommendation System Problems   Memory-based classification techniques are commonly used for modeling recommendation problems. They rely on the intuition that similar users and/or items behave similarly, facilitating user-to-item, item-to-item, or user-to-user proximities. A significant drawback of memory-based classification techniques is that they perform poorly with large scale data. Thus, using the off-the-shelf classification techniques for recommendation problems generally lead to impractical computational costs. In this study, we propose a recommendation problem-specific enhancement for a widely known memory-based classification algorithm, K-Nearest Neighbor. For this purpose, the movie recommendation problem is selected, and the scalability of the proposed enhancement is evaluated on two publicly available datasets. In the proposed enhancement, user- and item-proximities are pre-calculated during the first offline recommendation, while an auxiliary data structure is constructed for keeping user-to-user proximities. The stored neighborhood information is then facilitated in order to speed up later recommendations. The experiments show that the proposed algorithm has performed superior to both classical classification model and the state-of-the-art off-the-shelf toolkits. 11. M. Melinščak (Karlovac University of Applied Sciences, Karlovac, Croatia), S. Lončarić (Faculty of Electrical Engineering and Computing, Zagreb, Croatia) Retinal OCT Image Segmentation: How Well do Algorithms Generalize or How Transferable are the Data?  The success of deep learning depends, among other things, on a large amount of labeled data. However, in medical applications, large labeled datasets are the exception, rather than the rule. Manual image labeling is time-consuming and is generally performed only with the purpose of developing algorithms, and not as a part of standard clinical practice. The goal of this study is twofold. Since there is always a trade-off between the ability to collect data and achieve the best possible performance, we wanted to explore how performance depends on the amount of data. For this purpose, a database of manually annotated OCT images was collected. Also, we wanted to see how much transfer learning can help. Retinal OCT images vary depending on the type of device, therefore developed methods should be as robust as possible. Transfer learning was performed so that the model was trained with similar OCT images and then fine-tuned with images from the collected database. It has been shown that transfer learning helps in terms of generalization and better prediction in case the source database is similar to the target database. We can also assume that further improvement can probably be achieved by adding images from another distribution (medical or nonmedical). 12. M. Njirjak, E. Otovic (Tehnicki fakultet, Rijeka, Croatia), M. Budimir (Medicinski fakultet, Rijeka, Croatia), H. Vlahovic (Fakultet zdravstvenih studija, Katedra za fizioterapiju, Rijeka, Croatia), M. Tomic (Tehnicki fakultet, Rijeka, Croatia), V. Marijančić (Fakultet zdravstvenih studija, Katedra za fizioterapiju, Rijeka, Croatia) Smartphone Based Range of Motion Measurement in Physiotherapy  Range of motion measurement is often performed in physiotherapy. It is important for both initial patient's joint health status evaluation, and later on for tracking progress of a rehabilitation therapy. Standard goniometers are often cumbersome to use and many therapists resort to approximative by eye angle assessment. In this paper, we investigate a modern smartphone application goniometer, based on an inertial measurement unit (IMU). We analyze IMU properties and their influence on angle measurement accuracy. The application is tested in a controlled environment and later on in real-life situations. The same set of measurements is performed by experienced physiotherapist using both the smartphone application and traditional goniometers. Shoulder abduction and elbow pronation and supination were examined. The results obtained clearly show that the smartphone application is comparable in accuracy to traditional ways of measurement. There was no statistically significant difference between the measurements made with traditional instruments and the mobile application. It was shown that for some ROM measurements it was efficient and easy to use and can be considered to be a viable and a better alternative to traditional goniometers. Friday, 10/2/2020 9:00 AM - 1:00 PM, Camelia 2, Grand hotel Adriatic, Opatija DS - Data Science  9:00 AM - 10:40 AM Papers  1. I. Ivanoska, E. Petreska, B. Risteska Stojkoska, S. Kalajdziski, K. Trivodaliev (Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University, Skopje, Skopje, Macedonia) Comparative Analysis of Network Embeddings for Functional Annotation in Protein Interaction Networks  One of the major problems in bioinformatics is the computational prediction of functions for the large number of sequenced proteins which will facilitate the expensive and long process of wet lab verification. Protein-protein interaction networks (PINs) are considered as one of the richest sources of information for solving this problem. PINs can be represented as graphs, where the nodes are the proteins with their functions as node labels and the edges are their physical interactions. In this paper embedding vectors are created to represent the nodes of the graph which are later used as the input data for a classification model. This is a graph node classification problem and because of the property of proteins to have multiple functions, it is also a multi-label problem. The classification model used is linear SVM, while the embeddings are built with 4 algorithms, HOPE, SDNE, GF and node2vec and then a comparative analysis is done on the results. Hamming loss is used as an evaluation metrics, because of the multi-label problem. Based on the comparative evaluation recommendation for using a specific network embedding in specific scenarios is given. 2. I. Vasileska, P. Tomšič, L. Kos (Faculty of Mechanical Engineering, University of Ljubljana, Ljubljana, Slovenia, Ljubljana, Slovenia) Modernization of the PIC Codes for Exascale Plasma Simulation  Particle simulation in the field of nuclear fusion is a well-established technique which has spawned dozens of codes around the world through years (e.g. BIT1, VPIC, VSIM, OSIRIS, REMP, EPOCH, SMILEI, FBPIC, GENE, WARP, PEPC) with varying degrees of specialization for different physic areas and accessibility. By leveraging advanced algorithmic features we propose to implement key for genuine pre-exascale capability. While more expensive than conventional mesh-based particle simulation, proposed approach eliminates artificial grid-generated noise by design and offers a more direct method for treating complex simulation case. A combined library would make coupling of mesh-based and mesh-free methods a lot easier. Strategically this work is expected to benefit not only by future users, but will also provide a fully alternative approach for high-fidelity modelling. 3. M. Jančič, J. Slak, G. Kosec (Jozef Stefan Institute, Ljubljana, Slovenia) GPU Accelerated RBF-FD Solution of Poisson’s Equation  The Radial Basis Function-generated finite differences became a popular variant of local meshless strong form methods due to its robustness regarding the position of nodes and its controllable order of accuracy. However, this comes at the cost of larger neighbourhood sizes. In this paper, we present a GPU accelerated numerical solution of Poisson's equation on scattered nodes in 3D for orders from 2 up to 8. We specifically study the effect of using different orders on GPU acceleration efficiency. 4. U. Duh (Jozef Stefan Institute and University of Ljubljana, Ljubljana, Slovenia), M. Depolli (Jozef Stefan Institute, Ljubljana, Slovenia), J. Slak (Jozef Stefan Institute and University of Ljubljana, Ljubljana, Slovenia), G. Kosec (Jozef Stefan Institute, Ljubljana, Slovenia) Parallel Point Sampling for 3D Bodies  Point sampling in interiors and on surfaces of 3D bodies has important applications in computer graphics, visualization and physical simulations. Some of those application require certain statistical properties of the generated sample, but many don't and it is often only desired that the final points are distributed sufficiently regularly, with approximately equal distances to the neighboring points. We present a parallel version of an algorithm for point sampling that supports generation of locally regular point sets, but with globally variable point spacing. The behaviour of the algorithm is show-cased on interiors and surfaces of 2D and 3D models. The speedups of the parallel execution are also reported. 5. T. Kovačević, L. Mrčela, A. Merćep, Z. Kostanjčar (Fakultet elektrotehnike i računarstva, Sveučilište u Zagrebu, Zagreb, Croatia) Modeling Agricultural Production Activities Using Weather and Soil Parameters  Activities concerning agricultural production processes are highly influenced by weather and soil parameters. Such parameters determine when certain activities in the agricultural process should be carried out. For example, it is known that fertilization should be applied when humidity of the soil is appropriate for fertilizer to get worked into it, as too dry soil causes fertilizer to be washed away. Furthermore, pest and disease control is mostly affected by wind direction and speed, which may cause pesticide to be overly spread or concentrated in some spot on the field. There is well known expert knowledge about connection between those parameters and scheduling production process activities. However, expert knowledge can be enriched and automatized with data-driven models, which are relatively unknown due to lack of research in this field. We propose a Bayesian logistic regression model in order to predict probability of conducting field activities, based on weather and soil parameters. We have trained and validated our model on historical data which contains activities during potato production in Croatia and weather station measurements gathered from several fields over a period of one season. 10:40 AM - 10:50 AM Break  10:50 AM - 12:50 PM Papers  6. I. Vlahović (Algebra University College, Zagreb, Croatia), M. Glunčić, K. Dekanić (Department of Physics, Faculty of Science, University of Zagreb, Zagreb, Croatia), L. Mršić, H. Jerković (Algebra University College, Zagreb, Croatia), I. Martinjak (Faculty of Science, University of Zagreb, Zagreb, Croatia), V. Paar (Croatian Academy of Sciences and Arts, Zagreb, Croatia) Global Repeat Map Algorithm (GRM) Reveals Differences in Alpha Satellite Number of Tandem and Higher Order Repeats (HORs) in Human, Neanderthal and Chimpanzee Genomes – Novel Tandem Repeat Database  Development of novel sequencing methods (NGS) enabled investigation of DNA sequence in centromere region that, as is shown in previous studies, are replete with alpha satellite sequences, 171bp long monomer units. Those units (monomers) are found in each of human chromosomes as tandem repeats or as higher order repeat (HOR) structures. Except for their role in chromosome segregation and microtubule attachment to kinetochores, recent studies show that alpha satellite copy number variations are associated with cancer outcomes, aneuploidy, aging, differences in gene expression and evolutionary development. Using Global Repeat Map algorithm (GRM) and its extended Alphasub tool for finding all alpha monomers in genomic sequence, we created tandem repeat database for human, Neanderthal and chimpanzee genomes. This database have almost 3000 records (only main data table is shown on-line, genom.hazu.hr) with information about build of inserted genomes, start position in chromosome, sequence length of array, monomer length and number of monomers for specific array. Analysis of large set of data based on alpha satellites will help us to understand their role in centromeric organization and function and enable us to make conclusions about evolution between closely related species even though technologic limitation of assembly of centromere region are still present. 7. I. Kesedžić, J. Božek (Faculty of Electrical Engineering and Computing, University of Zagreb, Zagreb, Croatia), M. Radoš (Croatian Institute for Brain Research, School of Medicine, University of Zagreb, Zagreb, Croatia), S. Popović, K. Ćosić (Faculty of Electrical Engineering and Computing, University of Zagreb, Zagreb, Croatia) Air Traffic Control Competencies Measurements Based on Functional Near-Infrared Spectroscopy  Air traffic control officers (ATCOs) are rigorously selected from the vast number of applicants and extensively trained, in order to have the required psychological and behavioural characteristics for this stressful occupation, as well as specialised performance competencies while controlling air traffic, such as situational awareness, workload management, problem-solving and decision-making. Recent studies have demonstrated the possible improvements of ATCO selection and training processes by using physiological, speech and oculometric features related to stress resilience, cognitive load, visual attention and fatigue. The functional near-infrared spectroscopy (fNIRS) is a brain imaging technique which provides a good insight into brain activation by measuring the haemoglobin concentration in the prefrontal cortex, which is one of the key brain areas responsible for complex cognitive behaviour, decision-making, working memory, and stress regulation. This paper focuses on the ATC performance competencies that could be measured with the fNIRS, demonstrates computational processing of signals obtained with this technique, and discusses the applicability of fNIRS measurements in the current ATC selection and training procedures. The experimental paradigm and the results obtained from the paradigm demonstrate that fNIRS shows differences in brain activation among individuals during cognitive tasks, thus allowing the more objective performance measurement. 8. M. Zanin, E. Menasalvas, A. Rodriguez González (Universidad Politécnica de Madrid, Pozuelo de Alarcón, Spain), P. Smrz (Brno University of Technology, Brno, Czech Republic) An Analytics Toolbox for Cyber-physical Systems Data Analysis: Requirements and Challenges  The fast improvement in telecommunication technologies that has characterised the last decade is enabling a revolution centred on Cyber-Physical Systems (CPSs). Elements inside cities, from vehicles to cars, can now be connected and share data, describing both our environment and our behaviour. These data can also be used in an active way, by becoming the tenet of innovative services and products. Still, having data is not tantamount to having knowledge, and an important overlooked topic is how should them be analysed. In this contribution we tackle the issue of the development of an analytics toolbox for processing CPS data. Specifically, we review and quantify the main requirements that should be fulfilled, both functional (e.g. flexibility or dependability) and technical (e.g. scalability, response time, etc.). We further propose an initial set of analysis that should in it be included. We finally review some challenges and open issues, including how security and privacy could be tackled by emerging new technologies. 9. L. Tišljarić, Ž. Majstorović, T. Erdelić, T. Carić (Fakultet prometnih znanosti, Zagreb, Croatia) Measure for Traffic Anomaly Detection on the Urban Roads Using Speed Transition Matrices  Road traffic anomaly detection is an essential research topic within the Intelligent Transport System (ITS) context. Urban road anomaly detection systems are a crucial part of the ITS regarding the trip planning, road security, and congestion estimation applications. In this paper, the method for traffic anomaly detection using Speed Transition Matrices (STM) is presented. The paper's main goal is to present the novel method for measuring the distance between two STM, as standard distance measures are inapplicable for the anomaly detection and road traffic analysis interpretation The method is based on the Euclidean distance measure between STM’s Center of Mass (COM), and the average STM that represents normal traffic conditions. The Global Navigation Satellite System (GNSS) data on the road-network of City of Zagreb were used as a case study, as it is, the capital and the largest city in Croatia, suitable for the application of the proposed methodology. The anomaly detection method resulted in 73 anomalous points which are presented on the digital map. The proposed method is compared to the other distance metrics used in the literature, and advantages over each of the metrics are highlighted. 10. C. Attard (Universiity of Malta, Msida, Malta), A. Naudi (University College London , London, United Kingdom), S. Mallia, D. Gauci, R. Farrugia (University of Malta, Msida, Malta) Assisting Motorists Using Parking Prediction through a Car App  More persons depend on private cars, particularly when alternative transport such as public transport is not as efficient as required. The majority of motorists get caught in queues moving slowly through large cities. Parking becomes more of a challenge in areas where existing car parks provide limited parking spaces. The model for the study was created following an observational study. This required a drone taking top-down images for building a dataset, which in turn was used to flag available parking slots consulting historic patterns. The dataset is currently available for research purposes. The vehicle-detection tool developed for this study was used to evaluate the manual logs of the dataset and obtained generally satisfactory results, albeit presenting some limitations. Different regression algorithms were tested on the dataset and the best one overall was selected for making predictions. After considering various techniques, a car app using web technologies and a Node.js framework was built. Through this solution, predictions made using the dataset have been stored in a MongoDB database and passed on to a motorist through the app. A total of 18 motorists took part in a controlled experiment designed to enable the functional and usability testing of the app. 11. J. Kuzmanova, M. Gusev, V. Zdraveski (University Ss Cyril and Methodius, Skopje, Macedonia) Calculating the Fire Weather Index for US Wildfires using a GPU  The Forest Fire Weather Index allows the assessment of fire danger using weather variables in order to increase preparedness to prevent or halt the spread of wildfires. It often needs to be computed over large areas, taking weather data from hundreds of thousands of stations. CUDA parallel programming can be used to do this more efficiently. This paper presents a CPU and a GPU version as a solution to this problem, using historic datasets of wildfires and weather in the US to measure performance.