Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This book constitutes the refereed proceedings of the 7th International Conference on Grid and Pervasive Computing, GPC 2012, held in Hong Kong, China, in May 2012. The 9 revised full papers and 19 short papers were carefully revised and selected from 55 submissions. They are organized in topical sections on cloud computing, grid and service computing, green computing, mobile and pervasive computing, scheduling and performance, and trust and security. Also included are 4 papers presented at the 2012 International Workshop on Mobile Cloud and Ubiquitous Computing (Mobi-Cloud 2012) held in conjunction with GPC 2012.

Engineering & Applied Sciences --- Computer Science --- Computational grids (Computer systems) --- Ubiquitous computing --- Computer science. --- Computer communication systems. --- Computer programming. --- Software engineering. --- Operating systems (Computers). --- Algorithms. --- Computer Science. --- Algorithm Analysis and Problem Complexity. --- Computer Communication Networks. --- Information Systems Applications (incl. Internet). --- Software Engineering. --- Operating Systems. --- Programming Techniques. --- Algorism --- Algebra --- Arithmetic --- Computer operating systems --- Computers --- Disk operating systems --- Systems software --- Computer software engineering --- Engineering --- Electronic computer programming --- Electronic data processing --- Electronic digital computers --- Programming (Electronic computers) --- Coding theory --- Communication systems, Computer --- Computer communication systems --- Data networks, Computer --- ECNs (Electronic communication networks) --- Electronic communication networks --- Networks, Computer --- Teleprocessing networks --- Data transmission systems --- Digital communications --- Electronic systems --- Information networks --- Telecommunication --- Cyberinfrastructure --- Network computers --- Informatics --- Science --- Foundations --- Operating systems --- Programming --- Distributed processing --- Computer software. --- Software, Computer --- Computer systems --- Application software. --- Application computer programs --- Application computer software --- Applications software --- Apps (Computer software) --- Computer software --- Computer networks. --- Computer and Information Systems Applications.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This book constitutes the refereed proceedings of the 15th International Conference on Web Information Systems and Applications, WISA 2018, held in Taiyuan, China, in September 2018. The 29 full papers presented together with 16 short papers were carefully reviewed and selected from 103 submissions. The papers cover topics such as machine learning and data mining; cloud computing and big data; information retrieval; natural language processing; data privacy and security; knowledge graphs and social networks; query processing; and recommendations.

Data mining. --- Data Mining and Knowledge Discovery. --- Algorithmic knowledge discovery --- Factual data analysis --- KDD (Information retrieval) --- Knowledge discovery in data --- Knowledge discovery in databases --- Mining, Data --- Database searching --- Data mining

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The aging population has led to an increase in individuals over 60 years old in Belgium, resulting in the cost of social security projected to rise from 26\% to 32\% of GDP by 2050. Musculoskeletal diseases, such as low back pain, are becoming more prevalent due to the aging population, which is the second highest contributor to global disability. Accurate measurement of spine kinematics is essential for understanding the mechanisms behind spinal motion and facilitating the disclosure of the etiology of LBP. However, accurate spine kinematics measurement under dynamic conditions is a challenging task due to the complex anatomical structure of the spine. There is a need for non-radiative and non-invasive motion measurement tools that comply with key requirements, including accurate kinematic measurements of the lumbar spine with minimal error, and easy-to-use and cost-effective tools that enable clinical evaluation under dynamic conditions. Designing an ultrasound sensor module for spine fusion surgery involves several critical considerations and steps. Firstly, it is essential to research and comprehends the specific requirements of the surgery and the relevant guidelines to be followed. The set-out requirements for this sensor module are motion in 6 DOF, a universal design that can be used for most spine shapes, must have a good connection to the skin, and a good connection with the optical markers and the US transducer. The optimal distance between sensors for a spine motion tracker is determined by studying the anatomical features of the lumbar spine. It was discovered that the horizontal distance between the transverse process should be between 41-47 mm, and the lateral distance between L1 to L5 vertebrae is between 34-37 mm. In addition, it is necessary to know the maximum vertebrae rotation in pitch, yaw, and roll. For ease of prototyping, all the spine sensor models were 3D printed. During the study, 7 different versions were tested and improved. These versions differed in the material used in 3D printing, the first iterations used PLA, but it was later switched to TPU for the more ideal material properties when bending. Additionally, the vertebrae pair connection methods were tested to fit the requirement of extension and rotation, whom in the end flat spring design was chosen. To determine the precision of the ultrasound spine motion sensor proposed in this thesis, a control model will be used. This control model is an optical probe to touch the transverse process. The experimental setup comprises a partial vertebrae specimen and an A-mode ultrasound holder consisting of six ultrasound transducers with four optical markers each. The ultrasound holder is designed using SolidWorks and manufactured using PLA material in a 3D printer. A specially designed impulse generator is used to run the ultrasound, and a Picoscope is used to record the echo from the ultrasound signals. The experiment comprised of three setups- Static Water Experimental setup, Static Inclined Agar Experimental setup, and Static straight AGAN Experimental setup. The point-cloud registration algorithm was used to estimate the position and orientation of the bone model in 3D space. The results of this study showed that the spine sensor worked as planned and the vertebrae's transverse process locations could be detected with at least 5 mm error in all cases. All the set-out requirements were fulfilled with the final design.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Freehand three-dimensional (3D) ultrasound (US) is a non-invasive and radiation-free technique to acquire a 3D model of an anatomical feature – in this case a lumbar vertebra – by assembling a series of two-dimensional (2D) images. Probe calibration is necessary to find the transformation that converts the 2D coordinates of pixels in the ultrasound image into 3D coordinates in the frame of reference of an optical tracker. Three calibration phantoms – a single cross-wire phantom, a Z-fiducial phantom and a single-wall phantom – are designed, as well as three corresponding calibration algorithms, to find the necessary calibration parameters. This is done by the aid of an optical tracking device, that continuously measures the US transducer’s pose in a reference frame in real-time. A semi-automatic image processing algorithm detects the phantom features in the US image and extracts their coordinates. By linking these coordinates to the US transducer’s pose at that moment, and by doing this for many US images, the calibration parameters can be calculated. Further, the ideal motion profiles to scan the phantom are explored. A motion profile holds a direct relation with the obtained data set and has a vast impact on the calibration parameters. The ideal motion profile ensures that these parameters are accurate and consistent over multiple tests. We noticed the single cross-wire calibration is the most accurate and consistent method, with a average root mean square error (RMSE) of 0.92 mm. The Z-fiducial has the easiest motion profile since only one translating motion is required and the single-wall is the fastest with about 3 minutes. Our image processing algorithm works at a success rate of 100% for the Z-fiducial and single-wall phantom and 93.8% for the single cross-wire.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Pedicle screw placement (PSP) is een medische ingreep bij spinale fusie om ruggenwervels gefixeerd te houden t.o.v. elkaar. Een van de grootste risico bij PSP is het penetreren van de pedicle wand, hieronder genoemd als (pedicle breach), wat kan leiden tot complicaties. Om dit te voorkomen gebruikt een chirurg een ball tip feeler (BTF). Door palpatie van de wanden van het pedicle met een BTF, kunnen breaches gedetecteerd worden voordat de schroeven geplaatst worden. Deze techniek blijkt echter onbetrouwbaar, zelfs wanneer de chirurg ervaren is. Het doel van deze thesis is dus om een automatisch oplossing te vinden om deze breaches te herkennen. Verscheidene onderzoeken zijn uitgevoerd met echografie, versnellingsmeters, weegcellen en optische sensoren. De redenen waarom protoypes met deze technologiën nog niet zijn gebouwd en getest zou kunnen doordat ze oftewel te duur, te complex, te tijdsintensief of niet volledig automatisch zijn. De nood voor een intelligente manier om pedicle breaches te detecteren blijft, maar de oplossing zal naast goedkoop, snel en makkelijk ook accuraat moeten zijn. De inertial measurement unit en de weegcel zijn twee goedkope opties om de breaches te detecteren door de acceleraties, oriëntatieveranderingen en de krachten van de BTF te meten. Deze sensoren worden gefixeerd door zelfontworpen 3D geprinte onderdelen. Door de complexiteit van de sensor data wordt er gebruik gemaakt van deep learning. Dit om patronen en correlaties te identificeren tussen de data en het detecteren van breaches. Na het vergelijken en testen van meerdere neurale netwerken (NN) is een combinatie van zowel een 1D convolutional en een long short-term memory NN (CNN-LSTM) gekozen. Het CNN-LSTM model analyseert de data van de BTF en geeft live feedback aan de chirurg bij detectie van een pedicle breach met behulp van een piepgeluid. De prestaties van het bekomen model is eerst getest op een simpele houten plaat met gaten en daarna op 3D geprinte buisjes die pedicles representeren. Op de eerste setup is model behaald met een nauwkeurigheid van 95% op een dataset van 207 test samples. Dit model is echter niet te hergebruiken op de complexere 2de setup. Hier reageert de BTF te sensitief en geeft het meer dan 50% vals positieven. Na nieuwe train data te verzamelen is een model getraind met een nauwkeurigheid van 92%. Dit model werkte echter niet op nieuwe experimentele data. Een data kwaliteitsprobleem bleek 1 van de fouten te zijn. Data waarbij de breaches zich aan de rand van een sample afspeelde kwamen niet voor in de train data. Een ander model met een nauwkeurigheid van slechts 75% kon in combinatie met een gestandaardiseerde methode 86% van de 3D geprinte buisjes correct categoriseren. Hierbij waren er slechts 9% vals positieven. Dit in vergelijking met de 80% vals positieven van chirurgen in een studie door MD Sedory in 2010 over de betrouwbaarheid van een BTF toont een grote verbetering. In deze thesis is een BTF ontwikkelt die met een IMU en een weegcel doormiddel van AI breach detectie voorspelt. Deze is nog in proof of concept fase maar heeft potentieel om een goedkope en gebruiksvriendelijke oplossing te zijn. In de toekomst zijn een betere datakwaliteit en het verder uitbouwen van het neurale netwerk de sleutel voor de ontwikkeling van deze slimme BTF.