The paper proposes the solutions for applying the adaptive real-time control algorithms, embedded devices and unmanned aerial vehicles to minimize the risk of collisions in different transport systems. The main goal of the research is to develop the adaptive algorithms for transport control and optimization. The proposed anti-collision system (TACS) is based on artificial immune system concept on the neural network basis with the real-time ability of self-training to detect potentially dangerous states of the system and perform the actions to avoid and prevent crashes caused by collisions of vehicles.
The IoT is innovative and important phenomenon prone to several services and applications such as the blockchain, but it should consider the legal issues related to the data protection law. We should be taken into account the legal issues related to the data protection and privacy law. Technological solutions are welcome, but it is necessary, before developing applications, to consider the risks which we cannot dismiss. Personal data is a value. It is important to evaluate the European Regulation n. 679/2016, European General Data Protection Regulation (GDPR) that will enter into force on 25 May 2018. The GDPR introduces Data Protection by Design and by Default, Data Protection Impact Assessment (DPIA), data breach notification and significant administrative fines in respect of infringements of the Regulation. It is fundamental to evaluate the legal issues and prevent them, adopting in each project the Data Protection by Design approach. Regarding the data protection and security risks, there are some issues with potential consequences for data and liability. A correct law analysis allows evaluating risks preventing the wrong use of personal data. The contribution describes the main legal issues related to privacy and data protection focusing on the Privacy by Design approach, according to the GDPR.
Health-related message boards are common digital health platforms with great practical and sociological relevance. In this work, we present an investigation of the topological structure of interactions in such forums. We visualise the interactions among users and between users and forum administrators using a simple yet effective network visualisation tool called the Bitmap of Sorted Adjacency Matrix (BOSAM) algorithm. We apply the BOSAM algorithm to six health-related message boards of various scopes and sizes. The outcome reveals major differences between the user interaction networks of these forums. The results of our analysis correlate closely with the characteristics of the respective message boards, including their topic coverage, presence of user communities, commercial or community nature and management style.
Alzheimer's disease (AD) and Parkinson's disease (PD) are two common, progressive neurodegenerative brain disorders. Their diagnosis is very challenging at an early disease stage, if based on clinical symptoms only. Brain imaging techniques such as [18F]-fluoro-deoxyglucose positron emission tomography (FDG-PET) can provide important additional information with respect to changes in the cerebral glucose metabolism. In this study, we use machine learning techniques to perform an automated classification of FDG-PET data. The approach is based on the extraction of features by applying the scaled subprofile model with principal component analysis (SSM/PCA) in order to extract characteristics patterns of glucose metabolism. These features are then used for discriminating healthy controls, PD and AD patients by means of two machine learning frameworks: Generalized Matrix Learning Vector Quantization (GMLVQ) with local and global relevance matrices, and Support Vector Machines (SVMs) with a linear kernel. Datasets from different neuroimaging centers are considered. Results obtained for the individual centers, show that reliable classification is possible. We demonstrate, however, that cross-center classification can be problematic due to potential center-specific characteristics of the available FDG-PET data.
We propose a novel algorithm for reconstruction of optical flow group transformation parameters from sequences of multi-channel images. Our method is based on “early integration” paradigm using all of the available spectral components and the vector field representation of the transformation group generators. This way the reconstructed flow carries the total information contained in the images and the singularity of the inverse problem is potentially reduced. The method avoids the more complex and ambiguous task of first reconstructing the local vector field and subsequently fitting the group transformation templates. Remaining singularity of the structural tensor is removed by a modified Tikhonov-type of regularization. The algorithm is quantitatively validated with recorded images transformed with generated vector fields that are then compared with the reconstructed optical flow. The dependence of reconstruction accuracy on both the parameters of the images and the magnitudes of the vector deformation fields is presented. We also show the application of the method to the real-world task of video-based detection of convulsive epileptic seizures and compare the output to the previously published results using standard optical flow algorithm.
We propose a novel algorithm for optical flow reconstruction, the Spectral Optical Flow Iterative Algorithm (SOFIA). It uses local structural information in color images. The reconstructed flow carries the total information contained in the images and the singularity of the inverse problem is potentially reduced when the spectral components are not mutually redundant. In addition, the method is extended to provide smoothening functionality by averaging the structural information represented by the structural tensor in local neighborhoods, avoiding thus gradient cancellation effects present when the image is directly smoothened. An iterative multi-scale scheme is proposed where the optical flow vector reconstructed at coarser scales is used to generate the source image for the reconstruction at finer scales. The algorithm is quantitatively validated with recorded images transformed with generated synthetic displacement vector fields. The dependence of reconstruction accuracy on the parameters of both the images and the vector deformation fields is presented.
In order to improve the performance of Convolutional Neural Networks (CNN) in the classification of mammographic images, many researchers choose to apply a normalization method during the pre-processing stage. In this work, we aimed to assess the impact of 6 different normalization methods in the classification performance of 2 CNNs. We have also explored 5 classifiers, being the first one the CNN itself. The other 4 correspond to Support Vector Machine (SVM), Random Forest (RF), Simple Logistic (SL) and Voted Perceptron (VP) classifiers, all of them fed with features extracted from one of the layers – comprised between the sixteenth and the nineteenth – of the CNN. The last 3 classifiers were tested with different options for data testing presentation, according to the Weka software: Supplied Test Set (STS), 10-fold Cross Validation (10-FCV) and Percentage Split (PS). Results indicate that the effect of image normalization in the performance of the CNNs depends on which network is chosen to make the classification; besides, the normalization method that seems to have the most positive impact is the one that subtracts to each image the corresponding image mean and divide it by the standard deviation (best AUC mean values were 0.786 for CNN-F and 0.790 for Caffe; the best run AUC values were, respectively, 0.793 and 0.791. Layer 1 freezing decreased the running time and did not harm the classification performance. Regarding the different classifiers, CNNs used alone with softmax yielded the best results, with the exception of the RF and SL classifiers, both using the 10-FCV and PS options; however, with these options, we cannot guarantee that the test set images are presented for the first time to the network.
A stimulation-based measurement paradigm is proposed that can provide a biomarker, which we call “separatrix proximity marker”, able to assess how close is a neural system from an epileptic transition. We use a distributed network of potentially multi-stable local systems that can have different levels of susceptibility for transitions to oscillatory states, considered surrogates of epileptic seizures. The simplest computational model of the bistable system is the complex-valued polynomial model, also called z6 model that has been developed by our group. A two-dimensional grid of (9 x 9) z6 units connected with non-linear aperture based interactions is generated for 1000 random sets of parameters, including local parameters and connectivity weights resulting in various degrees of epileptic condition. We investigated an “active” closed-loop protocol where stimulation sequences of increasing amplitudes were delivered to the model system. We found that the separatrix proximity can be used to estimate the relative closeness of the system to the state transition. In this way, our model prescribes a universal biomarker based on the single assumption that the epileptic state is caused by bistable dynamics and does not rely on the specific level of model detail. We also show that the only spatial information needed to do the analysis is the location of the stimulated area and therefore no locally acquired signals are necessary. The findings in this work can be exploited to increase the efficiency and accuracy of pre-surgical epileptogenic zone localisation in cases of focal epileptic seizure onsets or to determine the effective dose and discriminate responsiveness to anti-epileptic drugs.
In this paper an algorithm is proposed for detection of severity of Diabetic Macular Edema (DME) from Retinal Images, which are invariant to image rotation. The proposed work includes detection of optic disc, macula, exudates, region of interest localization and a level indicator which indicates the severity of disease as severe, moderate or normal DME. Rotation Invariant detection of macula makes this method efficient. A set of NPDR data of fundus image of an eye is used to test the proposed algorithm. The results show a better comprehensive performance of disease severity checker and computationally efficient.
The Macula is one of the important parts of the retina of the human eye, responsible for sharp central vision. Detection of the macula is an important task for an automated screening tool for different ocular pathologies. The proposed method presents an imaging method detect macula from fundus images automatically and which is invariant to image rotation. The proposed method includes a strategic region based method based on an adaptive arc which is used to detect macula efficiently. For experimentation purposes, normal and affected fundus images were collected from a local eye hospital to test the performance of the proposed method and achieved encouraging results. The method proposed for the detection of the macula is accurate, efficient and hence can be used in real time automated screening of various eye diseases.