A Phase Space Box-counting based Method for Arrhythmia Prediction from Electrocardiogram Time Series

Arrhythmia is one kind of cardiovascular diseases that give rise to the number of deaths and potentially yields immedicable danger. Arrhythmia is a life threatening condition originating from disorganized propagation of electrical signals in heart resulting in desynchronization among different chambers of the heart. Fundamentally, the synchronization process means that the phase relationship of electrical activities between the chambers remains coherent, maintaining a constant phase difference over time. If desynchronization occurs due to arrhythmia, the coherent phase relationship breaks down resulting in chaotic rhythm affecting the regular pumping mechanism of heart. This phenomenon was explored by using the phase space reconstruction technique which is a standard analysis technique of time series data generated from nonlinear dynamical system. In this project a novel index is presented for predicting the onset of ventricular arrhythmias. Analysis of continuously captured long-term ECG data recordings was conducted up to the onset of arrhythmia by the phase space reconstruction method, obtaining 2-dimensional images, analysed by the box counting method. The method was tested using the ECG data set of three different kinds including normal (NR), Ventricular Tachycardia (VT), Ventricular Fibrillation (VF), extracted from the Physionet ECG database. Statistical measures like mean (μ), standard deviation (σ) and coefficient of variation (σ/μ) for the box-counting in phase space diagrams are derived for a sliding window of 10 beats of ECG signal. From the results of these statistical analyses, a threshold was derived as an upper bound of Coefficient of Variation (CV) for box-counting of ECG phase portraits which is capable of reliably predicting the impeding arrhythmia long before its actual occurrence. As future work of research, it was planned to validate this prediction tool over a wider population of patients affected by different kind of arrhythmia, like atrial fibrillation, bundle and brunch block, and set different thresholds for them, in order to confirm its clinical applicability.

Embedding Video Object Detection Capabilities on Low-Power Nano-Drones

Our objective for this thesis work was the deployment of a Neural Network based approach for video object detection on board a nano-drone. Furthermore, we have studied some possible extensions to exploit the temporal nature of videos to improve the detection capabilities of our algorithm. For our project, we have utilized the Mobilenetv2/v3SSDLite due to their limited computational and memory requirements. We have trained our networks on the IMAGENET VID 2015 dataset and to deploy it onto the nano-drone we have used the NNtool and Autotiler tools by GreenWaves. To exploit the temporal nature of video data we have tried different approaches: the introduction of an LSTM based convolutional layer in our architecture, the introduction of a Kalman filter based tracker as a postprocessing step to augment the results of our base architecture. We have obtain a total improvement in our performances of about 2.5 mAP with the Kalman filter based method(BYTE). Our detector run on a microcontroller class processor on board the nano-drone at 1.63 fps.

A study of two Languages for Active Objects with Futures

In computer systems, specifically in multithread, parallel and distributed systems, a deadlock is both a very subtle problem - because difficult to pre- vent during the system coding - and a very dangerous one: a deadlocked system is easily completely stuck, with consequences ranging from simple annoyances to life-threatening circumstances, being also in between the not negligible scenario of economical losses. Then, how to avoid this problem? A lot of possible solutions has been studied, proposed and implemented. In this thesis we focus on detection of deadlocks with a static program analysis technique, i.e. an analysis per- formed without actually executing the program. To begin, we briefly present the static Deadlock Analysis Model devel- oped for coreABS−− in chapter 1, then we proceed by detailing the Class- based coreABS−− language in chapter 2. Then, in Chapter 3 we lay the foundation for further discussions by ana- lyzing the differences between coreABS−− and ASP, an untyped Object-based calculi, so as to show how it can be possible to extend the Deadlock Analysis to Object-based languages in general. In this regard, we explicit some hypotheses in chapter 4 first by present- ing a possible, unproven type system for ASP, modeled after the Deadlock Analysis Model developed for coreABS−−. Then, we conclude our discussion by presenting a simpler hypothesis, which may allow to circumvent the difficulties that arises from the definition of the ”ad-hoc” type system discussed in the aforegoing chapter.

Tecniche di utilizzo e classificazione di immagini satellitari multispettrali in un'ottica di pianificazione e gestione delle emergenze umanitarie

Il presente lavoro è inserito nel contesto di applicazioni che riguardano la pianificazione e gestione delle emergenze umanitarie. Gli aspetti che si sono voluti mettere in evidenza sono due. Da un lato l'importanza di conoscere le potenzialità dei dati che si hanno di fronte per poterli sfruttare al meglio. Dall'altro l'esigenza di creare prodotti che siano facilmente consultabili da parte dell'utente utilizzando due diverse tecniche per comprenderne le peculiarità. Gli strumenti che hanno permesso il presente studio sono stati tre: i principi del telerilevamento, il GIS e l'analisi di Change Detection.

Surface cloud radiative forcing from broadband radiation measurements on the Antarctic plateau

Surface based measurements systems play a key role in defining the ground truth for climate modeling and satellite product validation. The Italian-French station of Concordia is operative year round since 2005 at Dome C (75°S, 123°E, 3230 m) on the East Antarctic Plateau. A Baseline Surface Radiation Network (BSRN) site was deployed and became operational since January 2006 to measure downwelling components of the radiation budget, and successively was expanded in April 2007 to measure upwelling radiation. Hence, almost a decade of measurement is now available and suitable to define a statistically significant climatology for the radiation budget of Concordia including eventual trends, by specifically assessing the effects of clouds and water vapor on SW and LW net radiation. A well known and robust clear sky-id algorithm (Long and Ackerman, 2000) has been operationally applied on downwelling SW components to identify cloud free events and to fit a parametric equation to determine clear-sky reference along the Antarctic daylight periods (September to April). A new model for surface broadband albedo has been developed in order to better describe the features the area. Then, a novel clear-sky LW parametrization, based on a-priori assumption about inversion layer structure, combined with daily and annual oscillations of the surface temperature, have been adopted and validated. The longwave based method is successively exploited to extend cloud radiative forcing studies to nighttime period (winter). Results indicated inter-annual and intra-annual warming behaviour, i.e. 13.70 W/m2 on the average, specifically approaching neutral effect in summer, when SW CRF compensates LW CRF, and warming along the rest of the year due prevalentely to CRF induced on the LW component.

Laser shock peening treatment to control and moderate fatigue crack growth in aircraft structure based on residual stress engineering approach

Laser Shock Peening (LSP) is a surface enhancement treatment which induces a significant layer of beneficial compressive residual stresses of up to several mm underneath the surface of metal components in order to improve the detrimental effects of the crack growth behavior rate in it. The aim of this thesis is to predict the crack growth behavior in metallic specimens with one or more stripes which define the compressive residual stress area induced by the Laser Shock Peening treatment. The process was applied as crack retardation stripes perpendicular to the crack propagation direction with the object of slowing down the crack when approaching the peened stripes. The finite element method has been applied to simulate the redistribution of stresses in a cracked model when it is subjected to a tension load and to a compressive residual stress field, and to evaluate the Stress Intensity Factor (SIF) in this condition. Finally, the Afgrow software is used to predict the crack growth behavior of the component following the Laser Shock Peening treatment and to detect the improvement in the fatigue life comparing it to the baseline specimen. An educational internship at the “Research & Technologies Germany – Hamburg” department of AIRBUS helped to achieve knowledge and experience to write this thesis. The main tasks of the thesis are the following: •To up to date Literature Survey related to “Laser Shock Peening in Metallic Structures” •To validate the FE model developed against experimental measurements at coupon level •To develop design of crack growth slowdown in Centered Cracked Tension specimens based on residual stress engineering approach using laser peened strip transversal to the crack path •To evaluate the Stress Intensity Factor values for Centered Cracked Tension specimens after the Laser Shock Peening treatment via Finite Element Analysis •To predict the crack growth behavior in Centered Cracked Tension specimens using as input the SIF values evaluated with the FE simulations •To validate the results by means of experimental tests

Deep learning for computer vision: a comparison between convolutional neural networks and hierarchical temporal memories on object recognition tasks

In recent years, Deep Learning techniques have shown to perform well on a large variety of problems both in Computer Vision and Natural Language Processing, reaching and often surpassing the state of the art on many tasks. The rise of deep learning is also revolutionizing the entire field of Machine Learning and Pattern Recognition pushing forward the concepts of automatic feature extraction and unsupervised learning in general. However, despite the strong success both in science and business, deep learning has its own limitations. It is often questioned if such techniques are only some kind of brute-force statistical approaches and if they can only work in the context of High Performance Computing with tons of data. Another important question is whether they are really biologically inspired, as claimed in certain cases, and if they can scale well in terms of "intelligence". The dissertation is focused on trying to answer these key questions in the context of Computer Vision and, in particular, Object Recognition, a task that has been heavily revolutionized by recent advances in the field. Practically speaking, these answers are based on an exhaustive comparison between two, very different, deep learning techniques on the aforementioned task: Convolutional Neural Network (CNN) and Hierarchical Temporal memory (HTM). They stand for two different approaches and points of view within the big hat of deep learning and are the best choices to understand and point out strengths and weaknesses of each of them. CNN is considered one of the most classic and powerful supervised methods used today in machine learning and pattern recognition, especially in object recognition. CNNs are well received and accepted by the scientific community and are already deployed in large corporation like Google and Facebook for solving face recognition and image auto-tagging problems. HTM, on the other hand, is known as a new emerging paradigm and a new meanly-unsupervised method, that is more biologically inspired. It tries to gain more insights from the computational neuroscience community in order to incorporate concepts like time, context and attention during the learning process which are typical of the human brain. In the end, the thesis is supposed to prove that in certain cases, with a lower quantity of data, HTM can outperform CNN.