Eye Fundus Image Analysis for Automatic Detection of Diabetic Retinopathy

Diabetes is a rapidly increasing worldwide problem which is characterised by defective metabolism of glucose that causes long-term dysfunction and failure of various organs. The most common complication of diabetes is diabetic retinopathy (DR), which is one of the primary causes of blindness and visual impairment in adults. The rapid increase of diabetes pushes the limits of the current DR screening capabilities for which the digital imaging of the eye fundus (retinal imaging), and automatic or semi-automatic image analysis algorithms provide a potential solution. In this work, the use of colour in the detection of diabetic retinopathy is statistically studied using a supervised algorithm based on one-class classification and Gaussian mixture model estimation. The presented algorithm distinguishes a certain diabetic lesion type from all other possible objects in eye fundus images by only estimating the probability density function of that certain lesion type. For the training and ground truth estimation, the algorithm combines manual annotations of several experts for which the best practices were experimentally selected. By assessing the algorithm’s performance while conducting experiments with the colour space selection, both illuminance and colour correction, and background class information, the use of colour in the detection of diabetic retinopathy was quantitatively evaluated. Another contribution of this work is the benchmarking framework for eye fundus image analysis algorithms needed for the development of the automatic DR detection algorithms. The benchmarking framework provides guidelines on how to construct a benchmarking database that comprises true patient images, ground truth, and an evaluation protocol. The evaluation is based on the standard receiver operating characteristics analysis and it follows the medical practice in the decision making providing protocols for image- and pixel-based evaluations. During the work, two public medical image databases with ground truth were published: DIARETDB0 and DIARETDB1. The framework, DR databases and the final algorithm, are made public in the web to set the baseline results for automatic detection of diabetic retinopathy. Although deviating from the general context of the thesis, a simple and effective optic disc localisation method is presented. The optic disc localisation is discussed, since normal eye fundus structures are fundamental in the characterisation of DR.

Real-time Musical Pair Improvisation on Mobile Devices

This thesis discusses the design and implementation of a real-time musical pair improvisation scenario for mobile devices. In the scenario transferring musical information over a network connection was required. The suitability of available wireless communication technologies was evaluated and communication was analyzed and designed on multiple layers of TCP/IP protocol stack. Also an application layer protocol was designed and implemented for the scenario. The implementation was integrated into a mobile musical software for children using available software components and libraries although the used platform lead to hardware and software constraints. Software limitations were taken into account in design. The results show that real-time musical improvisation can be implemented with wireless communication and mobile technology. The results also show that link layer had the most significant effect on real-time communication in the scenario.

Diskreettiaikainen liukuvan moodin säätö kiinteäoksidipolttokennon DC-tasonnostossa

Huoli ympäristön tilasta ja fossiilisten polttoaineiden hinnan nousu ovat vauhdittaneet tutkimusta uusien energialähteiden löytämiseksi. Polttokennot ovat yksi lupaavimmista tekniikoista etenkin hajautetun energiantuotannon, varavoimalaitosten sekä liikennevälineiden alueella. Polttokenno on tehonlähteenä kuitenkin hyvin epäideaalinen, ja se asettaa tehoelektroniikalle lukuisia erityisvaatimuksia. Polttokennon kytkeminen sähköverkkoon on tavallisesti toteutettu käyttämällä galvaanisesti erottavaa DC/DC hakkuria sekä vaihtosuuntaajaa sarjassa. Polttokennon kulumisen estämiseksi tehoelektroniikalta vaaditaan tarkkaa polttokennon lähtövirran hallintaa. Perinteisesti virran hallinta on toteutettu säätämällä hakkurin tulovirtaa PI (Proportional and Integral) tai PID (Proportional, Integral and Derivative) -säätimellä. Hakkurin epälineaarisuudesta johtuen tällainen ratkaisu ei välttämättä toimi kaukana linearisointipisteestä. Lisäksi perinteiset säätimet ovat herkkiä mallinnusvirheille. Tässä diplomityössä on esitetty polttokennon jännitettä nostavan hakkurin tilayhtälökeskiarvoistusmenetelmään perustuva malli, sekä malliin perustuva diskreettiaikainen integroiva liukuvan moodin säätö. Esitetty säätö on luonteeltaan epälineaarinen ja se soveltuu epälineaaristen ja heikosti tunnettujen järjestelmien säätämiseen.

Tuulivoimalan staattoripakan puristusrakenteen kehittäminen

Diplomityössä tutkittiin voidaanko tuulivoimalan generaattorin staattoripakan puristamisessa hyödyntää komposiittista rakenneratkaisua. Tyypillisesti generaattorissa staattorin teräslevyt puristetaan erilaisilla teräsrakenteilla toisiaan vasten. Tavoitteena oli selvittää, voidaanko puristavan komposiittirakenteen osana hyödyntää liimaliitosta tai laminoitua liitosta. Tarkoitus oli etsiä rakenteeseen soveltuva liima ja liimaliitoksen arvot tai laminoitu rakenne ja sille soveltuvat materiaalit ja suoritustapa. Työssä on perehdytty erilaisiin tuulivoimalatyyppeihin, sekä niissä käytettäviin kesto- ja vierasmagnetoituihin generaattorityyppeihin. Tämän lisäksi on tarkasteltu niissä käytettävien staattorien valmistusvaihtoehtoja ja syitä miksi niissä olevat teräslevyt on puristettava toisiaan vasten. Samalla on luotu katsaus nykyisin käytössä oleviin rakenteisiin, joilla puristus voidaan toteuttaa. Liimauksesta on käsitelty perusteoriaa, sekä seikkoja jotka vaikuttavat liimaliitoksen kestoon. Työssä tutkittavaan liitokseen soveltuvien liimojen ominaisuuksia on käsitelty. Myös laminoituun liitokseen jo aiemmin kovettuneeseen komposiittiin on perehdytty. Tutkittavaan rakenteeseen soveltuvia hartsi- ja lasikuitutyyppejä on esitelty. Komposiittien mekaaniseen liittämiseen on lyhyesti perehdytty. Työssä suoritettiin useita vetokokeita, joilla selvitettiin puristusrakenteen tutkimista varten valmistettujen koekappaleiden suurin vetokuormankesto. Vetokokeiden perusteella voitiin valita soveltuvin rakenne staattorin puristamiseksi.

Design Aspects of Megawatt-Range Direct-Driven Permanent MagnetWind Generators

Direct-driven permanent magnet synchronous generator is one of the most promising topologies for megawatt-range wind power applications. The rotational speed of the direct-driven generator is very low compared with the traditional electrical machines. The low rotational speed requires high torque to produce megawatt-range power. The special features of the direct-driven generators caused by the low speed and high torque are discussed in this doctoral thesis. Low speed and high torque set high demands on the torque quality. The cogging torque and the load torque ripple must be as low as possible to prevent mechanical failures. In this doctoral thesis, various methods to improve the torque quality are compared with each other. The rotor surface shaping, magnet skew, magnet shaping, and the asymmetrical placement of magnets and stator slots are studied not only by means of torque quality, but also the effects on the electromagnetic performance and manufacturability of the machine are discussed. The heat transfer of the direct-driven generator must be designed to handle the copper losses of the stator winding carrying high current density and to keep the temperature of the magnets low enough. The cooling system of the direct-driven generator applying the doubly radial air cooling with numerous radial cooling ducts was modeled with a lumped-parameter-based thermal network. The performance of the cooling system was discussed during the steady and transient states. The effect of the number and width of radial cooling ducts was explored. The large number of radial cooling ducts drastically increases the impact of the stack end area effects, because the stator stack consists of numerous substacks. The effects of the radial cooling ducts on the effective axial length of the machine were studied by analyzing the crosssection of the machine in the axial direction. The method to compensate the magnet end area leakage was considered. The effect of the cooling ducts and the stack end area effects on the no-load voltages and inductances of the machine were explored by using numerical analysis tools based on the three-dimensional finite element method. The electrical efficiency of the permanent magnet machine with different control methods was estimated analytically over the whole speed and torque range. The electrical efficiencies achieved with the most common control methods were compared with each other. The stator voltage increase caused by the armature reaction was analyzed. The effect of inductance saturation as a function of load current was implemented to the analytical efficiency calculation.

Performance and scalability of isolated DC-DC converter topologies in low voltage, high current applications

Fuel cells are a promising alternative for clean and efficient energy production. A fuel cell is probably the most demanding of all distributed generation power sources. It resembles a solar cell in many ways, but sets strict limits to current ripple, common mode voltages and load variations. The typically low output voltage from the fuel cell stack needs to be boosted to a higher voltage level for grid interfacing. Due to the high electrical efficiency of the fuel cell, there is a need for high efficiency power converters, and in the case of low voltage, high current and galvanic isolation, the implementation of such converters is not a trivial task. This thesis presents galvanically isolated DC-DC converter topologies that have favorable characteristics for fuel cell usage and reviews the topologies from the viewpoint of electrical efficiency and cost efficiency. The focus is on evaluating the design issues when considering a single converter module having large current stresses. The dominating loss mechanism in low voltage, high current applications is conduction losses. In the case of MOSFETs, the conduction losses can be efficiently reduced by paralleling, but in the case of diodes, the effectiveness of paralleling depends strongly on the semiconductor material, diode parameters and output configuration. The transformer winding losses can be a major source of losses if the windings are not optimized according to the topology and the operating conditions. Transformer prototyping can be expensive and time consuming, and thus it is preferable to utilize various calculation methods during the design process in order to evaluate the performance of the transformer. This thesis reviews calculation methods for solid wire, litz wire and copper foil winding losses, and in order to evaluate the applicability of the methods, the calculations are compared against measurements and FEM simulations. By selecting a proper calculation method for each winding type, the winding losses can be predicted quite accurately before actually constructing the transformer. The transformer leakage inductance, the amount of which can also be calculated with reasonable accuracy, has a significant impact on the semiconductor switching losses. Therefore, the leakage inductance effects should also be taken into account when considering the overall efficiency of the converter. It is demonstrated in this thesis that although there are some distinctive differences in the loss distributions between the converter topologies, the differences in the overall efficiency can remain within a range of a few percentage points. However, the optimization effort required in order to achieve the high efficiencies is quite different in each topology. In the presence of practical constraints such as manufacturing complexity or cost, the question of topology selection can become crucial.

Tilannekuvaa liikkuvasta maalista

Kirjallisuusarvostelu

Scalable algorithms for height field illumination

Global illumination algorithms are at the center of realistic image synthesis and account for non-trivial light transport and occlusion within scenes, such as indirect illumination, ambient occlusion, and environment lighting. Their computationally most difficult part is determining light source visibility at each visible scene point. Height fields, on the other hand, constitute an important special case of geometry and are mainly used to describe certain types of objects such as terrains and to map detailed geometry onto object surfaces. The geometry of an entire scene can also be approximated by treating the distance values of its camera projection as a screen-space height field. In order to shadow height fields from environment lights a horizon map is usually used to occlude incident light. We reduce the per-receiver time complexity of generating the horizon map on N N height fields from O(N) of the previous work to O(1) by using an algorithm that incrementally traverses the height field and reuses the information already gathered along the path of traversal. We also propose an accurate method to integrate the incident light within the limits given by the horizon map. Indirect illumination in height fields requires information about which other points are visible to each height field point. We present an algorithm to determine this intervisibility in a time complexity that matches the space complexity of the produced visibility information, which is in contrast to previous methods which scale in the height field size. As a result the amount of computation is reduced by two orders of magnitude in common use cases. Screen-space ambient obscurance methods approximate ambient obscurance from the depth bu er geometry and have been widely adopted by contemporary real-time applications. They work by sampling the screen-space geometry around each receiver point but have been previously limited to near- field effects because sampling a large radius quickly exceeds the render time budget. We present an algorithm that reduces the quadratic per-pixel complexity of previous methods to a linear complexity by line sweeping over the depth bu er and maintaining an internal representation of the processed geometry from which occluders can be efficiently queried. Another algorithm is presented to determine ambient obscurance from the entire depth bu er at each screen pixel. The algorithm scans the depth bu er in a quick pre-pass and locates important features in it, which are then used to evaluate the ambient obscurance integral accurately. We also propose an evaluation of the integral such that results within a few percent of the ray traced screen-space reference are obtained at real-time render times.

Islandora Archipelago: Update on the Project & Community

Presentation at Open Repositories 2014, Helsinki, Finland, June 9-13, 2014

Along-scan radiometric gradient in Landsat images: understanding the relation between floristic and remote sensing information in tropical forests

The along-scan radiometric gradient causes severe interpretation problems in Landsat images of tropical forests. It creates a decreasing trend in pixel values with the column number of the image. In practical applications it has been corrected assuming the trend to be linear within structurally similar forests. This has improved the relation between floristic and remote sensing information, but just in some cases. I use 3 Landsat images and 105 floristic inventories to test the assumption of linearity, and to examine how the gradient and linear corrections affect the relation between floristic and Landsat data. Results suggest the gradient to be linear in infrared bands. Also, the relation between floristic and Landsat data could be conditioned by the distribution of the sampling sites and the direction in which images are mosaicked. Additionally, there seems to be a conjunction between the radiometric gradient and a natural east-west vegetation gradient common in Western Amazonia. This conjunction might have enhanced artificially correlations between field and remotely-sensed information in previous studies. Linear corrections may remove such artificial enhancement, but along with true and relevant spectral information about floristic patterns, because they can´t separate the radiometric gradient from a natural one.

Pinoamislaitteen suunnittelu vetokoneen jälkeiselle linjalle

Työn tavoitteena oli suunnitella vetokoneen linjalle pinoamislaite. Laitteen tehtävänä on pinota yksittäin vetokoneelta tulevista tuotteista asiakasvaatimusten mukaisia pinoja. Laitteen suunnittelu toteutettiin käyttämällä systemaattisen koneensuunnittelun metodia. Diplomityö rajattiin koskemaan laitteen mekaanisten ratkaisujen suunnittelua. Suunnittelussa huomioitiin linjan asettamat vaatimukset, jotka koskevat sekä tuotantomäärää että tilojen käyttöä. Laitteen suunnittelussa ja mallinnuksessa käytettiin Solidworksin 3D-mallinnusohjelmaa. Mallinnettuja ratkaisuvaihtoehtoja verrattiin keskenään. Parhaiten vaatimusluettelossa esitetyt vaatimukset täyttänyt ratkaisumalli valittiin kehitettäväksi. Suunnitellusta laitteesta valmistettiin työpiirustukset sekä kokoonpanokuvat.

Deep generative models for facial keypoints detection

A new area of machine learning research called deep learning, has moved machine learning closer to one of its original goals: artificial intelligence and general learning algorithm. The key idea is to pretrain models in completely unsupervised way and finally they can be fine-tuned for the task at hand using supervised learning. In this thesis, a general introduction to deep learning models and algorithms are given and these methods are applied to facial keypoints detection. The task is to predict the positions of 15 keypoints on grayscale face images. Each predicted keypoint is specified by an (x,y) real-valued pair in the space of pixel indices. In experiments, we pretrained deep belief networks (DBN) and finally performed a discriminative fine-tuning. We varied the depth and size of an architecture. We tested both deterministic and sampled hidden activations and the effect of additional unlabeled data on pretraining. The experimental results show that our model provides better results than publicly available benchmarks for the dataset.

Nichesourcing The Uralic Languages For The Benefit Of Linguistic Research And Lingual Societies

The emerging technologies have recently challenged the libraries to reconsider their role as a mere mediator between the collections, researchers, and wider audiences (Sula, 2013), and libraries, especially the nationwide institutions like national libraries, haven’t always managed to face the challenge (Nygren et al., 2014). In the Digitization Project of Kindred Languages, the National Library of Finland has become a node that connects the partners to interplay and work for shared goals and objectives. In this paper, I will be drawing a picture of the crowdsourcing methods that have been established during the project to support both linguistic research and lingual diversity. The National Library of Finland has been executing the Digitization Project of Kindred Languages since 2012. The project seeks to digitize and publish approximately 1,200 monograph titles and more than 100 newspapers titles in various, and in some cases endangered Uralic languages. Once the digitization has been completed in 2015, the Fenno-Ugrica online collection will consist of 110,000 monograph pages and around 90,000 newspaper pages to which all users will have open access regardless of their place of residence. The majority of the digitized literature was originally published in the 1920s and 1930s in the Soviet Union, and it was the genesis and consolidation period of literary languages. This was the era when many Uralic languages were converted into media of popular education, enlightenment, and dissemination of information pertinent to the developing political agenda of the Soviet state. The ‘deluge’ of popular literature in the 1920s to 1930s suddenly challenged the lexical orthographic norms of the limited ecclesiastical publications from the 1880s onward. Newspapers were now written in orthographies and in word forms that the locals would understand. Textbooks were written to address the separate needs of both adults and children. New concepts were introduced in the language. This was the beginning of a renaissance and period of enlightenment (Rueter, 2013). The linguistically oriented population can also find writings to their delight, especially lexical items specific to a given publication, and orthographically documented specifics of phonetics. The project is financially supported by the Kone Foundation in Helsinki and is part of the Foundation’s Language Programme. One of the key objectives of the Kone Foundation Language Programme is to support a culture of openness and interaction in linguistic research, but also to promote citizen science as a tool for the participation of the language community in research. In addition to sharing this aspiration, our objective within the Language Programme is to make sure that old and new corpora in Uralic languages are made available for the open and interactive use of the academic community as well as the language societies. Wordlists are available in 17 languages, but without tokenization, lemmatization, and so on. This approach was verified with the scholars, and we consider the wordlists as raw data for linguists. Our data is used for creating the morphological analyzers and online dictionaries at the Helsinki and Tromsø Universities, for instance. In order to reach the targets, we will produce not only the digitized materials but also their development tools for supporting linguistic research and citizen science. The Digitization Project of Kindred Languages is thus linked with the research of language technology. The mission is to improve the usage and usability of digitized content. During the project, we have advanced methods that will refine the raw data for further use, especially in the linguistic research. How does the library meet the objectives, which appears to be beyond its traditional playground? The written materials from this period are a gold mine, so how could we retrieve these hidden treasures of languages out of the stack that contains more than 200,000 pages of literature in various Uralic languages? The problem is that the machined-encoded text (OCR) contains often too many mistakes to be used as such in research. The mistakes in OCRed texts must be corrected. For enhancing the OCRed texts, the National Library of Finland developed an open-source code OCR editor that enabled the editing of machine-encoded text for the benefit of linguistic research. This tool was necessary to implement, since these rare and peripheral prints did often include already perished characters, which are sadly neglected by the modern OCR software developers, but belong to the historical context of kindred languages and thus are an essential part of the linguistic heritage (van Hemel, 2014). Our crowdsourcing tool application is essentially an editor of Alto XML format. It consists of a back-end for managing users, permissions, and files, communicating through a REST API with a front-end interface—that is, the actual editor for correcting the OCRed text. The enhanced XML files can be retrieved from the Fenno-Ugrica collection for further purposes. Could the crowd do this work to support the academic research? The challenge in crowdsourcing lies in its nature. The targets in the traditional crowdsourcing have often been split into several microtasks that do not require any special skills from the anonymous people, a faceless crowd. This way of crowdsourcing may produce quantitative results, but from the research’s point of view, there is a danger that the needs of linguists are not necessarily met. Also, the remarkable downside is the lack of shared goal or the social affinity. There is no reward in the traditional methods of crowdsourcing (de Boer et al., 2012). Also, there has been criticism that digital humanities makes the humanities too data-driven and oriented towards quantitative methods, losing the values of critical qualitative methods (Fish, 2012). And on top of that, the downsides of the traditional crowdsourcing become more imminent when you leave the Anglophone world. Our potential crowd is geographically scattered in Russia. This crowd is linguistically heterogeneous, speaking 17 different languages. In many cases languages are close to extinction or longing for language revitalization, and the native speakers do not always have Internet access, so an open call for crowdsourcing would not have produced appeasing results for linguists. Thus, one has to identify carefully the potential niches to complete the needed tasks. When using the help of a crowd in a project that is aiming to support both linguistic research and survival of endangered languages, the approach has to be a different one. In nichesourcing, the tasks are distributed amongst a small crowd of citizen scientists (communities). Although communities provide smaller pools to draw resources, their specific richness in skill is suited for complex tasks with high-quality product expectations found in nichesourcing. Communities have a purpose and identity, and their regular interaction engenders social trust and reputation. These communities can correspond to research more precisely (de Boer et al., 2012). Instead of repetitive and rather trivial tasks, we are trying to utilize the knowledge and skills of citizen scientists to provide qualitative results. In nichesourcing, we hand in such assignments that would precisely fill the gaps in linguistic research. A typical task would be editing and collecting the words in such fields of vocabularies where the researchers do require more information. For instance, there is lack of Hill Mari words and terminology in anatomy. We have digitized the books in medicine, and we could try to track the words related to human organs by assigning the citizen scientists to edit and collect words with the OCR editor. From the nichesourcing’s perspective, it is essential that altruism play a central role when the language communities are involved. In nichesourcing, our goal is to reach a certain level of interplay, where the language communities would benefit from the results. For instance, the corrected words in Ingrian will be added to an online dictionary, which is made freely available for the public, so the society can benefit, too. This objective of interplay can be understood as an aspiration to support the endangered languages and the maintenance of lingual diversity, but also as a servant of ‘two masters’: research and society.

Permanent magnet synchronous generator design solution for large directdrive wind turbines: Thermal behavior of the LC DD-PMSG

Wind is one of the most compelling forms of indirect solar energy. Available now, the conversion of wind power into electricity is and will continue to be an important element of energy self-sufficiency planning. This paper is one in a series intended to report on the development of a new type of generator for wind energy; a compact, high-power, direct-drive permanent magnet synchronous generator (DD-PMSG) that uses direct liquid cooling (LC) of the stator windings to manage Joule heating losses. The main param-eters of the subject LC DD-PMSG are 8 MW, 3.3 kV, and 11 Hz. The stator winding is cooled directly by deionized water, which flows through the continuous hollow conductor of each stator tooth-coil winding. The design of the machine is to a large degree subordinate to the use of these solid-copper tooth-coils. Both steady-state and timedependent temperature distributions for LC DD-PMSG were examined with calculations based on a lumpedparameter thermal model, which makes it possible to account for uneven heat loss distribution in the stator conductors and the conductor cooling system. Transient calculations reveal the copper winding temperature distribution for an example duty cycle during variable-speed wind turbine operation. The cooling performance of the liquid cooled tooth-coil design was predicted via finite element analysis. An instrumented cooling loop featuring a pair of LC tooth-coils embedded in a lamination stack was built and laboratory tested to verify the analytical model. Predicted and measured results were in agreement, confirming the predicted satisfactory operation of the LC DD-PMSG cooling technology approach as a whole.

Embedded Service Bus for Internet of Things Interoperability

Internet of Things (IoT) technologies are developing rapidly, and therefore there exist several standards of interconnection protocols and platforms. The existence of heterogeneous protocols and platforms has become a critical challenge for IoT system developers. To mitigate this challenge, few alliances and organizations have taken the initiative to build a framework that helps to integrate application silos. Some of these frameworks focus only on a specific domain like home automation. However, the resource constraints in the large proportion of connected devices make it difficult to build an interoperable system using such frameworks. Therefore, a general purpose, lightweight interoperability framework that can be used for a range of devices is required. To tackle the heterogeneous nature, this work introduces an embedded, distributed and lightweight service bus, Lightweight IoT Service bus Architecture (LISA), which fits inside the network stack of a small real-time operating system for constrained nodes. LISA provides a uniform application programming interface for an IoT system on a range of devices with variable resource constraints. It hides platform and protocol variations underneath it, thus facilitating interoperability in IoT implementations. LISA is inspired by the Network on Terminal Architecture, a service centric open architecture by Nokia Research Center. Unlike many other interoperability frameworks, LISA is designed specifically for resource constrained nodes and it provides essential features of a service bus for easy service oriented architecture implementation. The presented architecture utilizes an intermediate computing layer, a Fog layer, between the small nodes and the cloud, thereby facilitating the federation of constrained nodes into subnetworks. As a result of a modular and distributed design, the part of LISA running in the Fog layer handles the heavy lifting to assist the lightweight portion of LISA inside the resource constrained nodes. Furthermore, LISA introduces a new networking paradigm, Node Centric Networking, to route messages across protocol boundaries to facilitate interoperability. This thesis presents a concept implementation of the architecture and creates a foundation for future extension towards a comprehensive interoperability framework for IoT.