京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第22326号 / 医博第4567号 / 新制||医||1041(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 滝田 順子, 教授 小川 誠司, 教授 遊佐 宏介 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

京都大学 / 新制・課程博士 / 博士(理学) / 甲第23004号 / 理博第4681号 / 新制||理||1671(附属図書館) / 京都大学大学院理学研究科物理学・宇宙物理学専攻 / (主査)准教授 WENDELL Roger, 教授 中家 剛, 准教授 久徳 浩太郎 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

Sonic logs are essential tools for reliably identifying interval velocities which, in turn, are used in many seismic processes. One problem that arises, while logging, is irregularities due to washout zones along the borehole surfaces that scatters the transmitted energy and hence weakens the signal recorded at the receivers. To alleviate this problem, I have extended the theory of super-virtual refraction interferometry to enhance the signal-to-noise ratio (SNR) sonic waveforms. Tests on synthetic and real data show noticeable signal-to-noise ratio (SNR) enhancements of refracted P-wave arrivals in the sonic waveforms. The theory of super-virtual interferometric stacking is composed of two redatuming steps followed by a stacking procedure. The first redatuming procedure is of correlation type, where traces are correlated together to get virtual traces with the sources datumed to the refractor. The second datuming step is of convolution type, where traces are convolved together to dedatum the sources back to their original positions. The stacking procedure following each step enhances the signal to noise ratio of the refracted P-wave first arrivals. Datuming with correlation and convolution of traces introduces severe artifacts denoted as correlation artifacts in super-virtual data. To overcome this problem, I replace the datuming with correlation step by datuming with deconvolution. Although the former datuming method is more robust, the latter one reduces the artifacts significantly. Moreover, deconvolution can be a noise amplifier which is why a regularization term is utilized, rendering the datuming with deconvolution more stable. Tests of datuming with deconvolution instead of correlation with synthetic and real data examples show significant reduction of these artifacts. This is especially true when compared with the conventional way of applying the super-virtual refraction interferometry method.

In this thesis, I tackle the problem of facial image super-resolution using convolutional neural networks with focus on preserving identity. I propose a method consisting of DPNet architecture and training algorithm based on state-of-the-art super-resolution solutions. The model of DPNet architecture is trained on Flickr-Faces-HQ dataset, where I achieve SSIM value 0.856 while expanding the image to four times the size. Residual channel attention network, which is one of the best and latest architectures, achieves SSIM value 0.858. While training models using adversarial loss, I encountered problems with artifacts. I experiment with various methods trying to remove appearing artefacts, which weren't successful so far. To compare quality assessment with human perception, I acquired image sequences sorted by percieved quality. Results show, that quality of proposed neural network trained using absolute loss approaches state-of-the-art methods.

In healthcare, a large number of medical data has emerged. To effectively use these data to improve healthcare outcomes, clinicians need to identify the relevant measures and apply the correct analysis methods for the type of data at hand. In this dissertation, we present various machine learning (ML) and data mining (DM) methods that could be applied to the type of data sets that are available in the healthcare area. The first part of the dissertation investigates DM methods on healthcare or medical data to find significant information in the form of rules. Class association rule mining, a variant of association rule mining, was used to obtain the rules with some targeted items or class labels. These rules can be used to improve public awareness of different cancer symptoms and could also be useful to initiate prevention strategies. In the second part of the thesis, ML techniques have been applied in healthcare or medical data to build a predictive model. Three different classification techniques on a real-world breast cancer risk factor data set have been investigated. Due to the imbalance characteristics of the data set various resampling methods were used before applying the classifiers. It is shown that there was a significant improvement in performance when applying a resampling technique as compared to applying no resampling technique. Moreover, super learning technique that uses multiple base learners, have been investigated to boost the performance of classification models. Two different forms of super learner have been investigated - the first one uses two base learners while the second one uses three base learners. The models were then evaluated against well-known benchmark data sets related to the healthcare domain and the results showed that the SL model performs better than the individual classifier and the baseline ensemble. Finally, we assessed cancer-relevant genes of prostate cancer with the most significant correlations with the clinical outcome of the sample type and the overall survival. Rules from the RNA-sequencing of prostate cancer patients was discovered. Moreover, we built the regression model and from the model rules for predicting the survival time of patients were generated.

Une précision à l’échelle de la milliseconde dans le transfert d'informations entre les neurones est essentielle pour la synchronisation et la plasticité des circuits neuronaux dans le cerveau. Les axones sont des prolongements neuronaux qui assurent la communication via des impulsions électriques ou des potentiels d’action (PA). A cause du manque de myéline et de leur diamètre très fin, les axones de l'hippocampe propagent les PA lentement et ainsi générer des délais de conduction très long (jusqu’à 100 ms) qui sont traditionnellement considérés comme invariants. Cependant, plusieurs études ont montré que l'activité change la morphologie des axones et module le temps de latence de la transmission. Il convient donc de se demander si le diamètre des axones varie en fonction de l'activité pouvant influencer lapropagation des PA.Les diamètres des axones non-myélinisés de l’hippocampe (compris entre 100-350 nm) sont généralement trop petits pour être résolu par la microscopie photonique conventionnelle. Le développement récent de l’imagerie super résolution STED permet désormais l'observation de la dynamique de leur morphologie détaillée dans le tissu vivant. En combinant la microscopie STED, l’électrophysiologie avec enregistrements en champs et patch-clamp dans des tranches de cerveau de souris et des simulations informatiques, nous avons découvert que les axones du CA3 subissent un élargissement de leur diamètre après l'induction de la potentialisation à long terme (PLT). Nous démontrons que cet élargissementde diamètre augmente la vitesse de conduction des PA. Dans l'ensemble, nos résultats indiquent que les axones peuvent réguler leur diamètre de manière dynamique changeant le délai de conduction des PA, ce qui modifie le timing du transfert d’information dans les circuits neuronaux. Cette étude suggère l’existence d’un nouveau type de mécanisme structurel dans le compartiment axonal jouant un rôle pour la plasticité neuronale. / Millisecond timing precision in the transfer of information between neurons is essential for the synchrony and plasticity of neural circuits in the brain. Axons are neuronal extensions that ensure the communication via brief electrical impulses called action potentials (AP). Because they are unmyelinated and are extremely thin, hippocampal axons propagate APsslowly and thus generate long delays of conduction (up to 100 ms) that are traditionally considered invariant. However, recent studies have shown that activity changes the morphology of axons and modulate the latency of transmission, thus raising the question whether axons undergo activity-dependent structural changes that could influence the propagation of APs. The diameter of hippocampal axons (ranging between 100-350 nm) are usually too thin to be properly resolved by conventional light microscopy. However, the development of super resolution STED imaging now enables the observation of their detailed morphological dynamics in living tissue. Using a novel combination of STED microscopy, field recordings, patch-clamp electrophysiology in mouse brain slices and computer simulations we discovered that CA3 axons undergo long-lasting enlargement in their diameter after the induction of long term potentiation (LTP). We provide strong evidence that this diameter enlargement increases AP conduction velocity. Taken together, our findings indicate that axons can dynamically tune AP propagation delays by changing their diameters, thereby altering the timing of information transfer in neural circuits. This study suggests a novel and powerful structural mechanism for neural plasticity.

In this study, the regular expressions forming abstract states in Brzozowski’s algorithm are not remapped to sequential state transition table addresses as would be the case in the classical approach, but are hashed to integers. Two regular expressions that are hashed to the same hash code are assigned the same integer address in the state transition table, reducing the number of states in the automaton. This reduction does not necessarily lead to the construction of a minimal automaton: no restrictions are placed on the hash function hashing two regular expressions to the same code. Depending on the quality of the hash function, a super-automaton, previously referred to as an approximate automaton, or an exact automaton can be constructed. When two regular expressions are hashed to the same state, and they do not represent the same regular language, a super-automaton is constructed. A super-automaton accepts the regular language of the input regular expression, in addition to some extra strings. If the hash function is bad, many regular expressions that do not represent the same regular language will be hashed together, resulting in a smaller automaton that accepts extra strings. In the ideal case, two regular expressions will only be hashed together when they represent the same regular language. In this case, an exact minimal automaton will be constructed. It is shown that, using the hashing approach, an exact or super-automaton is always constructed. Another outcome of the hashing approach is that a non-deterministic automaton may be constructed. A new version of the hashing version of Brzozowski’s algorithm is put forward which constructs a deterministic automaton. A method is also put forward for measuring the difference between an exact and a super-automaton: this takes the form of the k-equivalence measure: the k-equivalence measure measures the number of characters up to which the strings of two regular expressions are equal. The better the hash function, the higher the value of k, up to the point where the hash function results in regular expressions being hashed together if and only if they have the same regular language. Using the k-equivalence measure, eight generated hash functions and one hand coded hash function are evaluated for a large number of short regular expressions, which are generated using G¨odel numbers. The k-equivalence concept is extended to the average k-equivalence value in order to evaluate the hash functions for longer regular expressions. The hand coded hash function is found to produce good results. Copyright / Dissertation (MEng)--University of Pretoria, 2009. / Computer Science / unrestricted

The novel method herein proposed, aims to study Deoxyribonucleic acid (DNA) and Rad51 repair protein in its resting state after their interaction by using a combination of biological preparation and physical measures. Rad51 is a highly conserved protein; it is involved in eukaryotes genome stability and can interact with single strand (ss) and double strands (ds) DNA. In our work, a droplet of the solution containing the dsDNA/Rad51 complexes was deposited on micro-fabricated super-hydrophobic substrates (SHS) to obtain self-organized and suspended fibers. The silicon-based SHS were designed to incorporate a regular circular array of pillars and to maintain a high contact angle with the drop. The samples were let dehydrate at controlled temperature and humidity conditions. At the end of the buffer evaporation process, non-suspended material and salt excess are concentrated on the top of a few micro-pillars in a limited area (drop residual) of the device while ordered and self-assembled DNA/Rad51 fibers are suspended between micro-pillars. To find the ideal conditions to obtain and suspend the nucleic acid/protein complexes, several parameters were investigated: saline buffer, DNA and protein concentrations were widely titrated and showed a significant effect on the biomolecule suspension on SHS. The samples were then preliminarily checked by microscopy techniques and then described by the Raman spectra acquired. Several techniques were used: optical microscopy, Energy Dispersive X-Ray Spectroscopy (EDAX), Scanning Electron Microscopy (SEM) and Raman Spectroscopy. Protein expressions, DNA suspension, micro-fabrication and characterization were all performed in KAUST Core Labs and Structural Molecular Imaging Light Enhanced Spectroscopies (SMILEs) Lab. The novel approach presented in this work is highly multidisciplinary and comprises physical measurements (Raman spectroscopy and EM imaging), chemistry and biology. In future the method can be used further expanded supporting the data with HRTEM direct imaging to elucidate the nucleic acids/proteins behavior in the multiple phases of the genome repair processes. Also, it and can serve as a fingerprint of the biological molecules involved in biological interactions, their localization and structural characterization, providing a new tool for structural analysis, screening and diagnostics.

This thesis is focused on the impact of using soil conditioners in growing grapevine. In the experimental part the effect of Super Hume was evaluated at selected locations. It is impact on the quality and production was evaluated. The results were used to create the economic analysis when using Super Hume.

Automatic target recognition (ATR) is the ability for a computer to discriminate between different objects in a scene. ATR is often performed on point cloud data from a sensor known as a Ladar. Increasing the resolution of this point cloud in order to get a more clear view of the object in a scene would be of significant interest in an ATR application. A technique to increase the resolution of a scene is known as super resolution. This technique requires many low resolution images that can be combined together. In recent years, however, it has become possible to perform super resolution on a single image. This thesis sought to apply Gabor Wavelets and Compressive Sensing to single image super resolution of digital images of natural scenes. The technique applied to images was then extended to allow the super resolution of a point cloud.