Ring artifacts are a common problem with the use of photon-counting detectors and commercial deployment rests on being able to compensate for them. Deep learning has been proposed as a candidate for tackling the inefficiency or high cost of traditional techniques. In that spirit, we propose a new approach to ring artifact reduction, namely one that employs Residual Networks in sinogram domain. We train them on data simulated via a realistic photon-counting CT model based on numerical phantoms of real scans acquired by the KiTS19 Challenge dataset. By exploring various architectures we find that shallow ResNets achieve a significant artifact reduction by staying more true to the ground truth in terms of not introducing new artifacts. All networks introduce a smoothing effect which is attributed to the use of MSE as a loss function. An alternative training scheme using patches instead of whole sinograms is tested and it shows a slightly improved model stability. Lastly, we demonstrate via a performance metric study that common metrics are not suitable for quantifying the performance in this problem, save for a potential new approach in the virtual mono-energetic domain. / Ringartefakter är ett vanligt problem vid användning av fotonräknande detektorer och kommersiell introduktion kräver att man kan kompensera för dem. Djupinlärning har föreslagits som en kandidat för att hantera ineffektiviteten eller de höga kostnaderna för traditionella tekniker. I den andan föreslår vi ett nytt tillvägagångssätt för att reducera ringartefakter, nämligen en som använder sig av residualnätverk i sinogramdomänen. Vi tränar dem på data simulerad via en realistisk fotonräkning CT modell baserad på numeriska fantomer av verkliga skanningar från datamängen KiTS19 Challenge. Genom att utforska olika arkitekturer finner vi att grunda ResNet uppnår en betydande minskning av artefakter genom bevara en större likhet med den sanna bilden när det gäller att inte introducera nya artefakter. Alla nätverk introducerar en utsmetningseffekt som tillskrivs användningen av MSE som en förlustfunktion. Ett alternativt träningsschema med utsnitt istället för hela sinogram testas och det visar en något förbättrad modellstabilitet. Slutligen visar vi genom en prestandamåttstudie att vanliga prestandamått inte är lämpliga för att kvantifiera prestandan i detta problem med undantag för ett potentiellt nytt tillvägagångssätt i den virtuella monoenergetiska domänen.

Computed tomography angiography (CTA) of the brain poses challenges on the imaging system; the contrast between blood vessels and surrounding soft tissue is very low, and to render small intricate vessel structures high spatial resolution is needed. Higher precision angiography would facilitate more accurate diagnosis of pathological conditions. The aim of this work was to analyze the factors which contribute to the image quality in cerebrovascular imaging contexts and make a comparison between state-of-the-art energy-integrating and photon counting CT systems. A geometrical model was devised to mimic the conditions of cerebral angiography. Different parameters and detectors were used to reconstruct images of a spherical head phantom. Compton noise was added to several image acquisitions after a Monte Carlo study was used to estimate the scatter to primary ratio (SPR) with a spherical phantom. The images were evaluated qualitatively and quantitatively. A real phantom was scanned with an experimental photon counting detector and compared with the simulated approach. The work resulted in qualitative reconstructed images, a decrease in SPR when introducing air gaps and improved resolution but worsened contrast as a result of smaller detector sizes. The SPR was shown to be higher in cone-beam geometry than fan-beam geometry. Electronic noise present with energy integrating detectors was shown to degrade image quality significantly in low dose imaging, reducing contrast when imaging vascular-like structures. Photon counting detectors without electronic noise could provide greater image quality and better diagnostic information.

Semiconductor detectors are attracting interest for use in photon-counting spectral computed tomography. In order to obtain a high spatial resolution, it is of interest to find the photon interaction position. In this work we investigate if machine learning can be used to obtain a sub-pixel spatial resolution in a photon-counting silicon strip detector with pixels of 10 µm. Simulated charge distributions from events in one, three, and seven positions in each of three pixels were investigated using the MATLAB® Classification Learner application to determine the correct interaction position. Different machine learning models were trained and tested in order to maximize performance. With pulses originating from one and seven positions within each pixel, the model was able to find the originating pixel with an accuracy of 100% and 88.9% respectively. Further, the correct position within a pixel was found with an accuracy of 54.0% and 29.4% using three and seven positions per pixel respectively. These results show the possibility of improving the spatial resolution with machine learning. / Halvledardetektorer är av stigande intresse inom forskning för användning i fotonräknande datortomografi med spektral upplösning. För att erhålla en hög spatiell upplösning är det av intresse att hitta fotonens ursprungliga interaktionsposition. I detta arbete undersöks om maskininlärning kan användas för att erhålla en spatiell upplösning på subpixelnivå i en fotonräknande kiselstrippdetektor med 10 µm pixlar. Laddningsfördelningen från simulerade interaktioner i en, tre, och sju positioner inom var och en av tre pixlar undersöktes med hjälp av applikationen Classification Learner i MATLAB® för att bestämma den korrekta interaktionspositionen. Olika maskininlärningsmodeller tränades och testades för att maximera prestandan. När pulser från en och sju positioner inom pixeln användes, kunde modellen hitta den korrekta pixeln med en noggrannhet på 100% respektive 88.9%. Vidare kunde den korrekta positionen inom en pixel bestämmas med en noggrannhet på 54.0% och 29.4% när tre respektive sju positioner inom varje pixel användes. Resultaten visar att det skulle vara möjligt att förbättra den spatiella upplösningen med hjälp av maskininlärning.

In recent years, the usage of 3D deep learning techniques has seen a surge,mainly driven by advancements in autonomous driving and medical applications.This thesis investigates the applicability of existing state-of-the-art 3Ddeep learning network architectures to dense voxel grids from single photoncounting 3D LiDAR. This work also examine the choice of loss function asa means of dealing with extreme data imbalance, in order to segment peopleand vehicles in outdoor forest scenes. Due to data similarities with volumetricmedical data, such as computer tomography scans, this thesis investigates ifa model for 3D deep learning used for medical applications, the commonlyused 3D U-Net, can be used for photon counting data. The results showthat segmentation of people and vehicles is possible in this type of data butthat performance depends on the segmentation task, light conditions, and theloss function. For people segmentation the final models are able to predictall targets, but with a significant amount of false positives, something that islikely caused by similar LiDAR responses between people and tree trunks.For vehicle detection, the results are more inconsistent and varies greatlybetween different loss functions as well as the position and orientation of thevehicles. Overall, we consider the 3D U-Net model a successful proof-ofconceptregarding the applicability of 3D deep learning techniques to this kindof data. / Under de senaste åren har användningen för djupinlärningstekniker för 3Dsett en kraftig ökning, främst driven av framsteg inom autonoma fordon ochmedicinska tillämpningar. Denna avhandling undersöker befintliga modernadjupinlärningsnätverk för 3D i täta voxelgriddar från fotonräknande 3D LiDARför att segmentera människor och fordon i skogsscener. Vidare undersöksvalet av målfunktion som ett sätt att hantera extrem dataobalans. På grundav datalikheter med volymetriska medicinska data, såsom datortomografi,kommer denna avhandling att undersöka om en modell för 3D-djupinlärningsom används för medicinska applikationer, nämligen 3D U-Net, kan användasför fotonräknande data. Resultaten visar att segmentering av människor ochfordon är möjligt men att prestanda varier avsevärt med segmenteringsuppgiften,ljusförhållanden, och målfunktioner. För segmentering av människorkan de slutgiltiga modellerna segmentera alla mål men med en betydandemängd falska utslag, något som sannolikt orsakas av liknande LiDAR-svarmellan människor och trädstammar. För segmentering av fordon är resultatenmer oberäkneliga och varierar kraftigt mellan olika målfunktioner såväl somfordonens position och orientering. Sammantaget anser vi att 3D U-Netmodellenvisar på en framgångsrik konceptvalidering när det gäller tillämpningav djupinlärningstekniker för 3D på denna typ av data.

No description available.

[pt] Neste trabalho são apresentados dois reflectômetros ópticos por contagem de fótons no domínio do tempo para o monitoramento de fibras ópticas. O primeiro foi projetado para obter faixas dinâmicas altas. Demonstrou-se a sua capacidade de sintonização no monitoramento de redes passivas WDM-PON durante o tráfego de dados. 32 dB de faixa dinâmica com 6 m de resolução foram atingidos. O segundo reflectômetro foi projetado para atingir resoluções ultra altas. As aplicações neste caso, além do monitoramento de uma rede TDM-PON de curto alcance, foram na caracterização de redes de Bragg dispersivas e na descrição e modelagem de um fenômeno não reportado antes na literatura, chamado nesta tese de reflexão por curvatura. Foi demonstrada uma resolução menor que 3 cm com faixa dinâmica maior que 14.0 dB. / [en] This thesis presents the development of two photon counting optical time domain reflectometers for fiber optic links monitoring. The first one was focused on high dynamic range. It is demonstrated its tunable capability for a WDM-PON in-service monitoring. 32 dB on dynamic range and a two-point resolution of 6 m is achieved. The second reflectometer was design to accomplish an ultra high resolution. The monitoring of a short TDM-PON is performed. Moreover, due to its high resolution, a chirped fiber Bragg grating is characterized and a non previous reported phenomena, the bend reflection, is shaped and described. It is demonstrated 3 cm two-point resolution and more than 14 dB on dynamic range.

Antal datortomografiundersökningar har ökat under flera år i Sverige tack vare tekniska utvecklingar och ökad tillgänglighet på sjukvård. Södertälje sjukhus röntgenavdelningen är i behov av att byta ut en utdaterad datortomograf (eng: Computed tomography, CT) och avdelningen för medicinsk teknik har föreslagit ett inköp av en CT med fotonräknande-detektor. Bilddata framställdes genom en fantomstudie för att påvisa förhållandet mellanstråldosparametern CTDIvol och kontrast-brus-förhållandet (CNR) hos CT-systemen: SOMATOM Drive och NAEOTOM Alpha. Den genererade datan påvisade att det finns en väsentlig skillnad i CNR-CTDIvol-förhållandet mellan SOMATOM Drive och NAEOTOM Alpha. Resultaten tydliggör att NAEOTOM Alpha kan producera bilder med betydligt mindre brus vid lägre stråldoser. Ett inköp av en fotonräknande detektor CT skulle kunna rättfärdigas utifrån bildkvalitéförbättringen som systemet kan erbjuda. / The number of computed tomography (CT) scans has increased during the past years in Sweden due to technical advancements and increased availability of healthcare. The x-ray department at Södertälje hospital is in need of replacing an outdated computed tomography and the departmentof clinical engineering has proposed a purchase of a photon-counting detector CT. Image data was produced through a phantom study to demonstrate the relationship between the parameter CTDIvol radiation dose and the contrast-to-noise ratio (CNR) of the CT systems: SOMATOM Drive and NAEOTOM Alpha. The generated data demonstrated that there is a substantial difference in the CNR-CTDIvol relationship between SOMATOM Drive and NAEOTOM Alpha. The results entail that NAEOTOM Alpha can produce images with considerably less noise at lower radiation doses. The purchase of a photon-counting CT could be justified by the improved image quality it can offer.

Research on photon counting detectors (PCDs) is focused on semiconductor materials, and silicon is a strong candidate to use in PCDs for photon counting computer tomography (CT). In a silicon detector, a significant portion of the counts is due to Compton scattering events. Since only part of the incident photon energy is deposited in a Compton interaction, Compton interactions lead to a loss of spectral information. By using Compton coincidence detection, i.e., combining information from multiple Compton events caused by the same incident photon, it is possible to obtain more spectral information from Compton scattered photons, increasing the energy resolution of the detector. The goal of this thesis is to develop and evaluate a method for Compton coincidence detection for photon counting CT. In this thesis, a method for Compton coincidence detection based on Compton kinematics and a χ2 test is presented and compared to a previously developed method based on maximum likelihood estimation. The χ2 method utilised the connection between the energy before vs after a Compton interaction, and the scattering angle. The possible scattering angles due to deposited energy in each interaction were called the energy angles. The spatial angles between the interaction positions in the detector were calculated and compared to the energy angles through a χ2 test in order to find the correct order of interaction and the incident photon energy. The χ2 method correctly identified the interaction order of 85.8% of simulated interaction chains ending in photoelectric effect and 64.1% of simulated interaction chains containing only Compton interactions. The energy estimation was 100% correct for all chains ending in photoelectric effect, since all of the incident energy was deposited in the detector. For chains of only Comptoninteractions, the energy was estimated with an RMS error of 21.2 keV. Combining the results from chains ending in a photoelectric interaction and chains of only Comptoninteractions, the total RMS error of the energy estimation was 11.5 keV. / Datortomografi (CT) är en viktig del av dagens sjukvård, och fotonräknande detektorer för CT är på väg från forskning till klinisk användning. Forskningen inom fotonräknande detektorer fokuserar på att använda halvledande material, och kisel är en stark kandidat till att användas för fotonräknande detektorer. I en kiseldetektor interagerar en betydande andel av fotonerna genom Compton-spridning. Då endast en del av fotonenergin deponeras i detektorn när en Compton-interaktion sker leder det till en förlust av spektral information. Genom att kombinera information från flera Compton-interaktioner som orsakats av samma infallande foton, så kallad sammanfallsdetektering, är det möjligtatt erhålla en ökad mängd spektral information från Compton-spridna fotoner. Målet med detta examensarbete är att utveckla och utvärdera en metod för sammanfallsdetektering för att erhålla spektral information från Compton-spridda fotoner i en detektortill fotonr¨aknande CT. I detta arbete presenteras en metod baserad på kinematiken bakom en Compton-interaktion och ett χ2-test. Metoden jämförs sedan med en tidigare utvecklad metod baserad på maximum likelihood-uppskattning. χ2-metoden utnyttjade sambandet mellan deponerad energi i en Compton-interaktion och möjliga spridningsvinklar, här kallade energivinklar. De spatiella vinklarna mellan interaktionerna i detektorn mättes och jämfördes genom ett χ2-test för att hitta interaktionsordningen och den infallande energin. χ2-metoden identifierade interaktionsordningen korrekt för 85.5% av alla simulerade interaktionskedjor som slutade i fotoelektrisk effekt och 64.1% av alla simulerade interaktionskedjorsom endast innehöll Compton-interaktioner. Uppskattningen av infallande energi var 100% korrekt för alla interaktionskedjor som slutade med en fotoelektrisk interaktion,eftersom all infallande energi deponerats i detektorn. För kejdor som endast bestod av Compton-interaktioner uppskattades den infallande energin med ett RMS-fel på 21.2 keV. Genom att kombinera resultaten från kedjor som slutade med en fotoelektrisk interaktion och resultaten från kejdor som endast bestod av Compton-interaktioner blev det totala RMS-felet för energi-uppskattningen 11.5 keV.

La scintillation organique correspond au phénomène d’émission de lumière par un matériau moléculaire à la suite de l’excitation de celui-ci par un rayonnement externe d’énergie donnée. Lors de l’interaction, le dépôt d’énergie induit des transitions électroniques peuplant des états dont la plupart se désexcite de manière non radiative, à l’exception d’une, entre le premier état électronique singulet et l’état fondamental de la molécule. Lors de cette relaxation, un photon de fluorescence est émis. Cette émission a deux origines : i) l’excitation directe par le rayonnement primaire et les électrons secondaires ; elle donne lieu à une émission dite rapide ou prompte ; ii) l’ionisation par le rayonnement primaire et les électrons secondaires ; elle donne lieu à une émission dite lente ou différée. Ce travail de recherche fondamentale, à la fois théorique et expérimental, fait l’analyse de toutes les étapes du processus, de l’interaction primaire à l’émission de fluorescence, de manière à relier la dose déposée à la quantité de lumière émise, à des fins d’applications en dosimétrie médicale. Il repose sur la mesure des déclins de fluorescence de deux molécules modèles, l’anthracène et le paraterphényle, excitées par un flux continu de rayons X, et la séparation des contributions rapide et lente de la lumière émise, aux énergies médicales. Une modélisation analytique des processus physiques conduisant à l’émission de lumière, au regard de la dose déposée, a ensuite été effectuée, faisant apparaître de nombreux résultats originaux. Dans un premier temps, un dispositif expérimental original a été développé, basé sur la technique TCSPC (Time-Correlated Single Photon Counting), afin de pouvoir mesurer des déclins temporels de fluorescence en résolution nanoseconde et sous flux d’irradiation continu. Dans un second temps, nous avons développé une nouvelle approche mathématique permettant d’extraire finement les composantes rapides et lentes du signal. L’analyse des résultats a montré, pour la première fois, l’existence d’un rapport R constant et uniquement fonction du matériau, entre les rendements d’excitation et d’ionisation. Le caractère constant de ce rapport ne peut être attribué qu’à un mécanisme d’autoionisation moléculaire au sein d’un matériau se comportant intrinsèquement comme une chambre d’ionisation proportionnelle pour l’ionisation secondaire de basse énergie. Ceci est en accord total avec la linéarité observée entre l’intensité totale de lumière différée (ionisation) et la dose mesurée par une chambre d’ionisation proportionnelle. Une étude plus approfondie des mécanismes d’excitation, au regard du rapport R, a également permis de montrer, pour la première fois, une proportionnalité directe entre l’intensité totale de la lumière prompte et le dépôt d’une dose que nous avons baptisé dose d’excitation. Cette dose a été observée comme étant de 4 à 14 fois supérieure à celle mesurée par une chambre d’ionisation. Ce résultat original majeur devra impérativement conduire à des études futures afin de mieux comprendre les dégâts infligés à la matière organique et biologique par les excitations. / Organic scintillation is the emission of light by an organic scintillator when irradiated by an external source of radiation depositing enough energy to excite the molecule. Electronic states are populated by the electronic transitions generated by the deposited energy. The states de-excite through radiationless transitions, except for one, the transition between the first electronic state and the ground state where a photon of fluorescence is emitted. This light has two different origins: i) direct excitation caused by primary radiation or secondary electrons which leads to an emission knows as prompt; ii) ionization caused by primary radiation or secondary electrons generate what is known as the delayed component. This fundamental research was based on both theoretical and experimental work. We studied all the different processes in organic scintillation, from the interaction between the incident radiation and matter to the emission of light in order to find the relationship between fluorescence and the deposited dose, to the application to medical dosimetry. Two well known organic scintillators, anthracene and p-terphenyl, were excited using an X-ray source set at typical medical imaging parameters. The light emitted was acquired and an analytical model was used to describe the different processes that led to light emission revealing interesting new results.An experimental setup, based on the Time Correlated Single Photon Counting (TCSPC) technique, was developed to acquire fluorescence decay curves with nanosecond resolution using a continuous X-ray source. Afterwards, these curves were analyzed using an innovative mathematical approach in order to determine the prompt and delayed components.Results showed the ratio, defined as R, between the prompt and delayed components of fluorescence was constant and independent of the energy of the incident X-rays and that the response of the delayed component of fluorescence was linear to an ionization chamber. These observations were explained by considering that the only process taking place within the molecule after excitation was autoionization. Hence, the response of organic scintillator was the same as the one of an ionization chamber. Furthermore, due to the constant ration R, the response of prompt component of fluorescence was linear to the ionization chamber as well. This was the first time this behavior was observed and we referred to it as excitation dose. This dose was between 4 and 14 times bigger than the one measured with the ionization chamber. These original results suggested that energy is deposited mainly through excitation processes, suggesting the need for further studies to better understand the damage caused by excitation to the living.

Ein Ziel der Arbeit war die Entwicklung eines Verfahrens zur 3D-Positionierung auf Basis elektrooptischer Pseudostreckenmessung. Ein zweites Ziel war die Entwicklung eines Reflexgoniometers zur zweiachsigen Winkelmessung. Im Rahmen der Arbeit wurden die Grundlagen zur instrumentellen Umsetzung beider Verfahren erarbeitet, die Genauigkeitspotentiale ermittelt und mögliche Anwendungen für die einzelnen Schlüsseltechnologien und deren Kombination abgeleitet. In einer Prototyp-Entwicklung wurden Vorschläge für die wesentlichen Funktionselemente des räumlichen Weg- bzw. Winkelmesssystems gemacht. Hierzu gehören das kardanisch aufgehängte Etalon, die temperaturstabilisierte Laseroptik und die temperaturkalibrierte CCD der Winkelmesseinheit (Reflexgoniometer), die Systeme zur Erzeugung und Detektion ultrakurzer Laserpulse, eine elastische Optik, Möglichkeiten zur Formung des transversalen Strahlprofils, das TCSPC-System und die zur Auswertung und Kalibrierung notwendigen Algorithmen, wie z.B. die Autokollimation eines Lasers. ("Veröffentlicht von der Deutschen Geodätischen Kommission Reihe C (Dissertationen) unter der Nummer C 614 (München 2008; ISBN 3 7696 5053 0; 2; 144 S.") / One goal of the thesis was the development of a method for three-dimensional positioning based on electro-optical measurement of pseudo ranges. Another goal was the development of a reflex goniometer for biaxial angle measurement. Within the scope of this thesis the basics for the instrumental realisation of both methods were developed, the accuracy potentials were determined and possible applications for the separate key technologies and their combination were deduced. In a prototype development proposals were made for the main functional elements of the spatial distance and angle measurement systems. These include the gimbal mounted etalon, the temperature stabilised laser optics and the temperature calibrated CCD of the angle measurement device (reflex goniometer), the systems for creation and detection of ultrashort laser pulses, an elastic optical device, possibilities of transversal beam shaping, the TCSPC system and the algorithms which are necessary for analysis and calibration, e.g. the autocollimation of a laser.