Rozprawy doktorskie na temat „Proton sensing”

Thesis (M.S.)--Dept. of Physics. University of Missouri--Kansas City, 2006.
"A thesis in physics." Typescript. Advisor: Da-Ming Zhu. Vita. Title from "catalog record" of the print edition Description based on contents viewed Nov. 12, 2007. Includes bibliographical references (leaves 50-51). Online version of the print edition.

Activation of the intrinsic aqueous water-wire proton conductance (GAQ) in Hv1 channels is controlled by changes in membrane potential and the transmembrane pH gradient (ΔpH). The mechanism by which changes in ΔpH affect the apparent voltage dependence of GAQ activation is not understood. In order to measure voltage sensor (VS) activation in Hv1, we mutated a conserved Arg residue in the fourth helical segment (S4) to His and measured H+ currents under whole-cell voltage clamp in transfected HEK-293 cells. Consistent with previous studies in VS domain containing proteins, we find that Hv1 R205H mediates a robust resting-state H+ ‘shuttle’ conductance (GSH) at negative membrane potentials. Voltage-dependent GSH gating is measured at more negative voltages than the activation GAQ, indicating that VS activation is thermodynamically distinct from opening of the intrinsic H+ permeation pathway. A hallmark biophysical feature of Hv1 channels is a ~-40 mV/pH unit shift in the apparent voltage dependence of GAQ gating. We show here that changes pHO are sufficient to cause similar shifts in GSH gating, indicating that GAQ inherits its pH dependence from an early step in the Hv1 activation pathway. Furthermore, we show for the first time that Hv1 channels manifest a form of electromechanical coupling VS activation and GAQ pore opening. Second-site mutations of D185 markedly alter GAQ gating without affecting GSH gating, indicating that D185 is required for a late step in the activation pathway that controls opening of the aqueous H+ permeation pathway. In summary, this work demonstrates that the Hv1 activation pathway contains multiple transitions with distinct voltage and pH dependencies that have not been previously identified. The results reported here novel insight into the mechanism of VS activation in Hv1 and raise fundamental questions about the nature of pH-dependent gating and electromechanical coupling in related VS domain-containing ion channels and phosphatases.

Parathyroid hormone (PTH) secretion maintains free-ionised extracellular calcium (Ca2+o) homeostasis under the control of the calcium-sensing receptor (CaR). In humans and dogs, blood acidosis and alkalosis is associated with increased or suppressed PTH secretion respectively. Furthermore, large (1.0 pH unit) changes in extracellular pH (pHo) alter Ca2+o sensitivity of the CaR in CaR-transfected HEK-293 cells (CaR-HEK). Indeed, it has been found in this laboratory that even pathophysiological acidosis (pH 7.2) renders CaR less sensitive to Ca2+o while pathophysiological alkalosis (pH 7.6) increases its Ca2+o sensitivity, both in CaR-HEK and parathyroid cells. If true in vivo, then CaR’s pHo sensitivity might represent a mechanistic link between metabolic acidosis and hyperparathyroidism in ageing and renal disease. However, in acidosis one might speculate that the additional H+ could displace Ca2+ bound to plasma albumin, thus increasing free-Ca2+ concentration and so compensating for the decreased CaR responsiveness. Therefore, I first demonstrated that a physiologically-relevant concentration of albumin (5% w/v) failed to overcome the inhibitory effect of pH 7.2 or stimulatory effect of pH 7.6 on CaR-induced intracellular Ca2+ (Ca2+i) mobilisation. Determining the molecular basis of CaR pHo sensitivity would help explain cationic activation of CaR and permit the generation of experimental CaR models that specifically lack pHo sensitivity. With extracellular histidine and free cysteine residues the most likely candidates for pHo sensing (given their sidechains’ pK values), all 17 such CaR residues were mutated to non-ionisable residues. However, none of the resulting CaR mutants exhibited significantly decreased CaR pHo sensitivity. Even co-mutation of the two residues whose individual mutation appeared to elicit modest reductions (CaRH429V and CaRH495V) failed to exhibit any change in CaR pHo sensitivity. I conclude therefore, that neither extracellular histidine nor free cysteine residues account for CaR pHo sensitivity. Next, it is known that cytosolic cAMP drives PTH secretion in vivo and that cAMP potentiates Ca2+o-induced Ca2+i mobilisation in CaR-HEK cells. Given the physiological importance of tightly controlled PTH secretion and Ca2+o homeostasis, here I investigated the influence of cAMP on CaR signalling in CaR-HEK cells. Agents that increase cytosolic cAMP levels such as forskolin and isoproterenol potentiated Ca2+o-induced Ca2+i mobilisation and lowered the Ca2+o threshold for Ca2+i mobilisation. Indeed, forskolin lowered the EC50 for Ca2+o on CaR (2.3 ± 0.1 vs. 3.0 ± 0.1 mM control, P<0.001). Forskolin also potentiated CaR-induced ERK phosphorylation; however protein kinase A activation appeared uninvolved in any of these effects. Pertussis toxin, used to block CaR-induced suppression of cAMP accumulation, also lowered the Ca2+o threshold for Ca2+i mobilisation though appeared to do so by increasing efficacy (Emax). Furthermore, mutation of the CaR’s two putative PKA consensus sequences (CaRS899 and CaRS900) to a non-phosphorylatable residue (alanine) failed to alter the potency of Ca2+o for CaR or attenuate the forskolin response. In contrast, phosphomimetic mutation of CaRS899 (to aspartate) did increase CaR sensitivity to Ca2+o. Together this suggests that PKA-mediated CaRS899 phosphorylation could potentiate CaR activity but that this does not occur following Ca2+o treatment in CaR-HEK cells. Together, these data show that cAMP regulates the Ca2+o threshold for Ca2+i mobilisation, thus helping to explain differential efficacy between CaR downstream signals. If true in vivo, this could help explain how multiple physiological signal inputs may be integrated in parathyroid cells.

Reconstruire des champs géophysiques à partir d'observations bruitées et partielles est un problème classique bien étudié dans la littérature. L'assimilation de données est une méthode populaire pour aborder ce problème, et se fait par l'utilisation de techniques classiques, comme le filtrage de Kalman d’ensemble ou des filtres particulaires qui procèdent à une évaluation online du modèle physique afin de fournir une prévision de l'état. La performance de l'assimilation de données dépend alors fortement de du modèle physique. En revanche, la quantité de données d'observation et de simulation a augmenté rapidement au cours des dernières années. Cette thèse traite l'assimilation de données d'une manière data-driven et ce, sans avoir accès aux équations explicites du modèle. Nous avons développé et évalué l'assimilation des données par analogues (AnDA), qui combine la méthode des analogues et des méthodes de filtrage stochastiques (filtres Kalman, filtres à particules, chaînes de Markov cachées). Des applications aux modèles chaotiques simplifiés et à des études de cas de télédétection réelle (température de surface de lamer, anomalies du niveau de la mer), nous démontrons la pertinence d'AnDA pour l'interpolation de données manquantes des systèmes dynamiques non linéaires et à haute dimension à partir d'observations irrégulières et bruyantes.Motivé par l'essor du machine learning récemment, la dernière partie de cette thèse est consacrée à l'élaboration de modèles deep learning pour la détection et de tourbillons océaniques à partir de données de sources multiples et/ou multi temporelles (ex: SST-SSH), l'objectif général étant de surpasser les approches dites expertes
Reconstructing geophysical fields from noisy and partial remote sensing observations is a classical problem well studied in the literature. Data assimilation is one class of popular methods to address this issue, and is done through the use of classical stochastic filtering techniques, such as ensemble Kalman or particle filters and smoothers. They proceed by an online evaluation of the physical modelin order to provide a forecast for the state. Therefore, the performanceof data assimilation heavily relies on the definition of the physical model. In contrast, the amount of observation and simulation data has grown very quickly in the last decades. This thesis focuses on performing data assimilation in a data-driven way and this without having access to explicit model equations. The main contribution of this thesis lies in developing and evaluating the Analog Data Assimilation(AnDA), which combines analog methods (nearest neighbors search) and stochastic filtering methods (Kalman filters, particle filters, Hidden Markov Models). Through applications to both simplified chaotic models and real ocean remote sensing case-studies (sea surface temperature, along-track sea level anomalies), we demonstrate the relevance of AnDA for missing data interpolation of nonlinear and high dimensional dynamical systems from irregularly-sampled and noisy observations. Driven by the rise of machine learning in the recent years, the last part of this thesis is dedicated to the development of deep learning models for the detection and tracking of ocean eddies from multi-source and/or multi-temporal data (e.g., SST-SSH), the general objective being to outperform expert-based approaches

The calcium-sensing receptor (CaR) is a G-protein coupled receptor that controls mammalian extracellular calcium (Ca2+o) homeostasis. CaR downstream signalling involves intracellular calcium (Ca2+i) mobilisation which can be negatively modulated by protein kinase C (PKC)-mediated phosphorylation of CaR residue Thr-888 (CaRT888). The nature of this regulation was investigated here using siRNA-based knockdown of individual PKC isotypes. Knocking down PKCα expression increased CaR-induced Ca2+i mobilisation in CaR-HEK cells, significantly lowering the EC50 for Ca2+o relative to control siRNA-transfected cells. In accordance, PKCα knockdown also decreased CaRT888 phosphorylation which also permitted the triggering of Ca2+i mobilisation in CaR-HEK cells at sub-threshold Ca2+o concentrations. Interestingly, PKCε knockdown attenuated CaR-induced Ca2+i mobilisation in CaR-HEK cells, significantly increasing the EC50 for Ca2+o. However, this knockdown was also also found to inhibit CaRT888 phosphorylation and this is the first time that CaRT888 phosphorylation has been shown to be dissociate from Ca2+i mobilisation. The results show the complexity of the interactions that potentially underlie the CaR’s pleiotropic signalling and provides novel targets for examining signal bias. Classically an increase in cAMP is known to trigger PTH seceretion. The observation in this study shows that raising intracellular cAMP levels with forskolin also decreased CaRT888 phosphorylation permitting increased Ca2+i mobilisation. This suggests that cAMP may stimulate the phosphatase (most likely protein phosphatase 2A (PP2A)). Nevertheless, knocking down Gα12, which has been shown to activate PP2A, resulted in increased CaRT888 phosphorylation and lower Ca2+i mobilisation (increased EC50 for Ca2+o). This suggests the possibility of CaR as a cAMP sensor that can detect an increase in intracellular cAMP in order to stop PTH serection. Three novel CaR effectors, P70 ribosamal protein S6 kinase, insulin-like growth factor receptor-1 and nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, were identified in CaR-HEK cells. It was shown that a) high Ca2+o stimulated the activation of these effectors and b) each effector was inhibited by knockdown of PKCα and Gα12, which further confirmed the association of these signals with CaR. These data show that CaR also plays an important role outside Ca2+o homeostasis, such as growth and inflammation. Finally, five CaR mutations associated with autosomal dominant hypocalcaemia (ADH) were found to increase Ca2+o-induced Ca2+i mobilisation, as well as ERK and p38MAPK activation, when transfected stably in HEK-293 cells. Cotreatment with the calcilytic NPSP795 inhibited ERK and p38MAPK phosphorylation in all 5 gain-of-function mutants and in the wild type CaR cells, with IC50s for the compound in the nanomolar range. These data highlight the potential utility of CaR negative allosteric modulators in the treatment of gain-of-function CaR mutations. Together these data enhance our understanding of CaRT888 phosphorylation and CaR signalling.

L’analyse des images satellite et aériennes figure parmi les sujets fondamentaux du domaine de la télédétection. Ces dernières années, les avancées technologiques ont permis d’augmenter la disponibilité à large échelle des images, en comprenant parfois de larges étendues de terre à haute résolution spatiale. En plus des questions évidentes de complexité calculatoire qui en surgissent, un de plus importants défis est l’énorme variabilité des objets dans les différentes régions de la terre. Pour aborder cela, il est nécessaire de concevoir des méthodes de classification qui dépassent l’analyse du spectre individuel de chaque pixel, en introduisant de l’information contextuelle de haut niveau. Dans cette thèse, nous proposons d’abord une méthode pour la classification avec des contraintes de forme, basée sur l’optimisation d’une structure de subdivision hiérarchique des images. Nous explorons ensuite l’utilisation des réseaux de neurones convolutionnels (CNN), qui nous permettent d’apprendre des descripteurs hiérarchiques profonds. Nous étudions les CNN depuis de nombreux points de vue, ce qui nous permettra de les adapter à notre objectif. Parmi les sujets abordés, nous proposons différentes solutions pour générer des cartes de classification à haute résolution et nous étudions aussi la récolte des données d’entrainement. Nous avons également créé une base de données d’images aériennes sur des zones variées, pour évaluer la capacité de généralisation des CNN. Finalement, nous proposons une méthode pour polygonaliser les cartes de classification issues des réseaux de neurones, afin de pouvoir les intégrer dans des systèmes d’information géographique. Au long de la thèse, nous conduisons des expériences sur des images hyperspectrales, satellites et aériennes, toujours avec l’intention de proposer des méthodes applicables, généralisables et qui passent à l’échelle
The analysis of airborne and satellite images is one of the core subjects in remote sensing. In recent years, technological developments have facilitated the availability of large-scale sources of data, which cover significant extents of the earth’s surface, often at impressive spatial resolutions. In addition to the evident computational complexity issues that arise, one of the current challenges is to handle the variability in the appearance of the objects across different geographic regions. For this, it is necessary to design classification methods that go beyond the analysis of individual pixel spectra, introducing higher-level contextual information in the process. In this thesis, we first propose a method to perform classification with shape priors, based on the optimization of a hierarchical subdivision data structure. We then delve into the use of the increasingly popular convolutional neural networks (CNNs) to learn deep hierarchical contextual features. We investigate CNNs from multiple angles, in order to address the different points required to adapt them to our problem. Among other subjects, we propose different solutions to output high-resolution classification maps and we study the acquisition of training data. We also created a dataset of aerial images over dissimilar locations, and assess the generalization capabilities of CNNs. Finally, we propose a technique to polygonize the output classification maps, so as to integrate them into operational geographic information systems, thus completing the typical processing pipeline observed in a wide number of applications. Throughout this thesis, we experiment on hyperspectral, atellite and aerial images, with scalability, generalization and applicability goals in mind

Les failles sismogéniques sont la source des séismes. L'étude de leurs propriétés nous informe donc sur les caractéristiques des forts séismes qu'elles peuvent produire. Les failles sont des objets 3D qui forment des réseaux complexes incluant une faille principale et une multitude de failles et fractures secondaires qui "découpent" la roche environnante à la faille principale. Mon objectif dans cette thèse a été de développer des approches pour aider à étudier cette fracturation secondaire intense. Pour identifier, cartographier et mesurer les fractures et les failles dans ces réseaux, j'ai adressé deux défis :1) Les failles peuvent former des escarpements topographiques très pentus à la surface du sol, créant des "couloirs" ou des canyons étroits et profond où la topographie et donc, la trace des failles, peut être difficile à mesurer en utilisant des méthodologies standard (comme des acquisitions d'images satellites optiques stéréo et tri-stéréo). Pour répondre à ce défi, j'ai utilisé des acquisitions multi-stéréos avec différentes configurations (différents angles de roulis et tangage, différentes dates et modes d'acquisitions). Notre base de données constituée de 37 images Pléiades dans trois sites tectoniques différents dans l'Ouest américain (Valley of Fire, Nevada ; Granite Dells, Arizona ; Bishop Tuff, California) m'a permis de tester différentes configurations d'acquisitions pour calculer la topographie avec trois approches différentes. En utilisant la solution photogrammétrique open-source Micmac (IGN ; Rupnik et al., 2017), j'ai calculé la topographie sous la forme de Modèles Numériques de Surfaces (MNS) : (i) à partir de combinaisons de 2 à 17 images Pléiades, (ii) en fusionnant des MNS calculés individuellement à partir d'acquisitions stéréo et tri-stéréo, évitant alors l'utilisant d'acquisitions multi-dates et (iii) en fusionnant des nuages de points calculés à partir d'acquisitions tri-stéréos en suivant la méthodologie multi-vues développée par Rupnik et al. (2018). J’ai aussi combiné, dans une dernière approche (iv), des acquisitions tri-stéréos avec la méthodologie multi-vues stéréos du CNES/CMLA (CARS) développé par Michel et al. (2020), en combinant des acquisitions tri-stéréos. A partir de ces quatre approches, j'ai calculé plus de 200 MNS et mes résultats suggèrent que deux acquisitions tri-stéréos ou une acquisition stéréo combinée avec une acquisition tri-stéréo avec des angles de roulis opposés permettent de calculer les MNS avec la surface topographique la plus complète et précise.2) Couramment, les failles sont cartographiées manuellement sur le terrain ou sur des images optiques et des données topographiques en identifiant les traces curvilinéaires qu'elles forment à la surface du sol. Néanmoins, la cartographie manuelle demande beaucoup de temps ce qui limite notre capacité à produire cartographies et des mesures complètes des réseaux de failles. Pour s'affranchir de ce problème, j'ai adopté une approche d'apprentissage profond, couramment appelé un réseau de neurones convolutifs (CNN) - U-Net, pour automatiser l'identification et la cartographie des fractures et des failles dans des images optiques et des données topographiques. Volontairement, le modèle CNN a été entraîné avec une quantité modérée de fractures et failles cartographiées manuellement à basse résolution et dans un seul type d'images optiques (photographies du sol avec des caméras classiques). A partir d'un grand nombre de tests, j'ai sélectionné le meilleur modèle, MRef et démontre sa capacité à prédire des fractures et des failles précisément dans données optiques et topographiques de différents types et différentes résolutions (photographies prises au sol, avec un drone et par satellite). Le modèle MRef montre de bonnes capacités de généralisations faisant alors de ce modèle un bon outil pour cartographie rapidement et précisément des fractures et des failles dans des images optiques et des données topographiques
Seismogenic faults are the source of earthquakes. The study of their properties thus provides information on some of the properties of the large earthquakes they might produce. Faults are 3D features, forming complex networks generally including one master fault and myriads of secondary faults and fractures that intensely dissect the master fault embedding rocks. I aim in my thesis to develop approaches to help studying this intense secondary faulting/fracturing. To identify, map and measure the faults and fractures within dense fault networks, I have handled two challenges:1) Faults generally form steep topographic escarpments at the ground surface that enclose narrow, deep corridors or canyons, where topography, and hence fault traces, are difficult to measure using the available standard methods (such as stereo and tri-stereo of optical satellite images). To address this challenge, I have thus used multi-stéréo acquisitions with different configuration such as different roll and pitch angles, different date of acquisitions and different mode of acquisitions (mono and tri-stéréo). Our dataset amounting 37 Pléiades images in three different tectonic sites within Western USA (Valley of Fire, Nevada; Granite Dells, Arizona; Bishop Tuff, California) allow us to test different configuration of acquisitions to calculate the topography with three different approaches. Using the free open-source software Micmac (IGN ; Rupnik et al., 2017), I have calculated the topography in the form of Digital Surface Models (DSM): (i) with the combination of 2 to 17 Pleiades images, (ii) stacking and merging DSM built from individual stéréo or tri-stéréo acquisitions avoiding the use of multi-dates combinations, (iii) stacking and merging point clouds built from tri-stereo acquisitions following the multiview pipeline developped by Rupnik et al., 2018. We used the recent multiview stereo pipeling CARS (CNES/CMLA) developped by Michel et al., 2020 as a last approach (iv), combnining tri-stereo acquisitions. From the four different approaches, I have thus calculated more than 200 DSM and my results suggest that combining two tri-stéréo acquisitions or one stéréo and one tri-stéréo acquisitions with opposite roll angles leads to the most accurate DSM (with the most complete and precise topography surface).2) Commonly, faults are mapped manually in the field or from optical images and topographic data through the recognition of the specific curvilinear traces they form at the ground surface. However, manual mapping is time-consuming, which limits our capacity to produce complete representations and measurements of the fault networks. To overcome this problem, we have adopted a machine learning approach, namely a U-Net Convolutional Neural Network, to automate the identification and mapping of fractures and faults in optical images and topographic data. Intentionally, we trained the CNN with a moderate amount of manually created fracture and fault maps of low resolution and basic quality, extracted from one type of optical images (standard camera photographs of the ground surface). Based on the results of a number of performance tests, we select the best performing model, MRef, and demonstrate its capacity to predict fractures and faults accurately in image data of various types and resolutions (ground photographs, drone and satellite images and topographic data). The MRef predictions thus enable the statistical analysis of the fault networks. MRef exhibits good generalization capacities, making it a viable tool for fast and accurate extraction of fracture and fault networks from image and topographic data

Créer un double numérique de la Terre sous forme de cartes a de nombreuses applications comme la conduite autonome, la planification urbaine, les télécommunications, la gestion des catastrophes naturelles, etc. Les systèmes d'information géographique (SIG) sont utilisés pour intégrer des données géolocalisées sous forme de cartes. Les SIG utilisent une représentation vectorielle pour les objets, prenant peu d'espace mémoire et rendant leur modification plus facile que des données raster. Avec la quantité croissante d'images satellites et aériennes capturées chaque jour, des méthodes automatiques sont en cours de développement pour extraire les informations de ces images de télédétection. Les méthodes d'apprentissage profond pour la segmentation d'images sont capables de délimiter les formes des objets, mais elles le font avec une représentation raster, sous la forme d'une carte de probabilité. Des méthodes de vectorisation post-traitement convertissent ensuite cette représentation raster en une représentation vectorielle compatible avec les SIG. Un autre défi de la télédétection est de gérer un certain type de bruit dans les données, qui est le désalignement entre différentes couches d'informations géolocalisées (par exemple entre les images et les cadastres des bâtiments). Ce type de bruit est fréquent en raison de diverses erreurs introduites lors du traitement des données de télédétection. Cette thèse développe des approches combinées d'apprentissage et géométriques dans le but d'améliorer l'automatisation du processus de cartographie SIG à partir d'images de télédétection.Nous proposons d'abord une méthode pour corriger une carte mal alignée sur une image, pur faire correspondre ces deux données géolocalisées, et aussi pour créer des jeu de données de télédétection pour la segmentation d'images avec une vérité terrain corrigé. En effet, entraîner un modèle sur une vérité terrain mal alignée ne mènerait pas à de bonnes segmentations. Au cours de ce travail, nous avons également observé un effet de débruitage par notre modèle d'alignement et l'avons utilisé pour débruiter un jeu de données mal aligné de manière auto-supervisée, ce qui signifie que seul le jeu de données mal aligné a été utilisé pour l'apprentissage.Nous proposons ensuite une approche simple pour utiliser un réseau de neurones produisant directement une représentation vectorielle de l'objet à détecter, afin de contourner l'étape de vectorisation post-traitement. Nous démontrons qu'il est possible d'apprendre à régresser les coordonnées de polygones (avec un nombre de sommets fixes dans notre cas), produisant directement des sorties cartographiques vectorielles.Bien que les méthodes plus récentes d'apprentissage directement en représentation vectorielle sont maintenant plus évoluées, elles ont encore d'autres limitations en termes de type de formes d'objets qu'elles peuvent prédire. Des cas topologiques plus complexes tels que des objets avec des trous ou des bâtiments se touchant ayant un mur mitoyen ne sont pas gérés par ces méthodes d'apprentissage. Nous proposons ainsi une approche hybride palliant ces limitations en entraînant un réseau de neurones pour produire une carte de probabilité de segmentation comme usuellement, mais aussi pour produire un “frame field” (4 champs vectoriels superposés) aligné avec les contours des objets détectés. Ce “frame field” encode des informations géométriques supplémentaires apprises par le réseau. Nous proposons ensuite notre méthode de polygonisation parallélisable pour exploiter ce “frame field” pour vectoriser efficacement la carte de probabilité de segmentation. Notre méthode de polygonisation ayant accès à des informations supplémentaires sous la forme d'un “frame field” elle peut être moins complexe que d'autres méthodes de vectorisation avancées et donc plus rapide. De plus calculer ce “frame field” n'augmente pratiquement pas le temps d'inférence, il n'est que bénéfique
Creating a digital double of the Earth in the form of maps has many applications in e.g. autonomous driving, automated drone delivery, urban planning, telecommunications, and disaster management. Geographic Information Systems (GIS) are the frameworks used to integrate geolocalized data and represent maps. They represent shapes of objects in a vector representation so that it is as sparse as possible while representing shapes accurately, as well as making it easier to edit than raster data. With the increasing amount of satellite and aerial images being captured every day, automatic methods are being developed to transfer the information found in those remote sensing images into Geographic Information Systems. Deep learning methods for image segmentation are able to delineate the shapes of objects found in images however they do so with a raster representation, in the form of a mask. Post-processing vectorization methods then convert that raster representation into a vector representation compatible with GIS. Another challenge in remote sensing is to deal with a certain type of noise in the data, which is the misalignment between different layers of geolocalized information (e.g. between images and building cadaster data). This type of noise is frequent due to various errors introduced during the processing of remote sensing data. This thesis develops combined learning and geometric approaches with the purpose to improve automatic GIS mapping from remote sensing images.We first propose a method for correcting misaligned maps over images, with the first motivation for them to match, but also with the motivation to create remote sensing datasets for image segmentation with alignment-corrected ground truth. Indeed training a model on misaligned ground truth would not lead to great performance, whereas aligned ground truth annotations will result in better models. During this work we also observed a denoising effect of our alignment model and use it to denoise a misaligned dataset in a self-supervised manner, meaning only the misaligned dataset was used for training.We then propose a simple approach to use a neural network to directly output shape information in the vector representation, in order to by-pass the post-processing vectorization step. Experimental results on a dataset of solar panels show that the proposed network succeeds in learning to regress polygon coordinates, yielding directly vectorial map outputs. Our simple method is limited to predicting polygons with a fixed number of vertices though.While more recent methods for learning directly in the vector representation do not have this limitation, they still have other limitations in terms of the type of object shapes they can predict. More complex topological cases such as objects with holes or buildings touching each other (with a common wall which is very typical of European city centers) are not handled by these fully deep learning methods. We thus propose a hybrid approach alleviating those limitations by training a neural network to output a segmentation probability map as usual and also to output a frame field aligned with the contours of detected objects (buildings in our case). That frame field constitutes additional shape information learned by the network. We then propose our highly parallelizable polygonization method for leveraging that frame field information to vectorize the segmentation probability map efficiently. Because our polygonization method has access to additional information in the form of a frame field, it can be less complex than other advanced vectorization methods and is thus faster. Lastly, requiring an image segmentation network to also output a frame field only adds two convolutional layers and virtually does not increase inference time, making the use of a frame field only beneficial

Le suivi d’objets et la prédiction de mouvement sont des aspects importants pour les véhicules autonomes. Tout d'abord, nous avons développé une méthode de suivi mono-objet en utilisant le compressive tracking, afin de corriger le suivi à base de flux optique et d’arriver ainsi à un compromis entre performance et vitesse de traitement. Compte tenu de l'efficacité de l'extraction de caractéristiques comprimées (compressive features), nous avons appliqué cette méthode de suivi au cas multi-objets pour améliorer les performances sans trop ralentir la vitesse de traitement. Deuxièmement, nous avons amélioré la méthode de suivi mono-objet basée sur DCF en utilisant des caractéristiques provenant d’un CNN multicouches, une analyse de fiabilité spatiale (via un masque d'objet) ainsi qu’une stratégie conditionnelle de mise à jour de modèle. Ensuite, nous avons appliqué la méthode améliorée au cas du suivi multi-objets. Les VGGNet-19 et DCFNet pré-entraînés sont testés respectivement en tant qu’extracteurs de caractéristiques. Le modèle discriminant réalisé par DCF est pris en compte dans l’étape d'association des données. Troisièmement, deux modèles LSTM (seq2seq et seq2dense) pour la prédiction de mouvement des véhicules et piétons dans le système de référence de la caméra sont proposés. En se basant sur des données visuelles et un nuage de points 3D (LiDAR), un système de suivi multi-objets basé sur un filtre de Kalman avec un détecteur 3D sont utilisés pour générer les trajectoires des objets à tester. Les modèles proposées et le modèle de régression polynomiale, considéré comme méthode de référence, sont comparés et évalués
Object tracking and motion prediction are important for autonomous vehicles and can be applied in many other fields. First, we design a single object tracker using compressive tracking to correct the optical flow tracking in order to achieve a balance between performance and processing speed. Considering the efficiency of compressive feature extraction, we apply this tracker to multi-object tracking to improve the performance without slowing down too much speed. Second, we improve the DCF based single object tracker by introducing multi-layer CNN features, spatial reliability analysis (through a foreground mask) and conditionally model updating strategy. Then, we apply the DCF based CNN tracker to multi-object tracking. The pre-trained VGGNet-19 and DCFNet are tested as feature extractors respectively. The discriminative model achieved by DCF is considered for data association. Third, two proposed LSTM models (seq2seq and seq2dense) for motion prediction of vehicles and pedestrians in the camera coordinate are proposed. Based on visual data and 3D points cloud (LiDAR), a Kalman filter based multi-object tracking system with a 3D detector are used to generate the object trajectories for testing. The proposed models, and polynomial regression model, considered as baseline, are compared for evaluation

La multiplication des sources de données et la mise à disposition de systèmes d'imagerie à haute résolution a fait rentrer l'observation de la Terre dans le monde du big data. Cela a permis l'émergence de nouvelles applications (étude de la répartition des sols par data mining, etc.) et a rendu possible l'application d'outils statistiques venant des domaines de l'apprentissage automatique et de la vision par ordinateur. Cette thèse cherche à concevoir et implémenter un modèle de classification bénéficiant de l'existence de grande bases de données haute résolution (si possible, annotées) et capable de générer des cartes sémantiques selon diverses thématiques. Les applications visés incluent la cartographie de zones urbaines ainsi que l'étude de la géologie et de la végétation à des fins industrielles.L'objectif de la thèse est de développer de nouveaux outils statistiques pour la classification d'images aériennes et satellitaires. Des approches d'apprentissage supervisé telles que les réseaux de neurones profonds, surpassant l'état-de-l'art en combinant des caractéristiques locales des images et bénéficiant d'une grande quantité de données annotées, seront particulièrement étudiées. Les principales problématiques sont les suivantes : (a) la prédiction structurée (comment introduire la structure spatial et spectral dans l'apprentissage ?), (b) la fusion de données hétérogènes (comment fusionner des données SAR, hyperspectrales et Lidar ?), (c) la cohérence physique du modèle (comment inclure des connaissances physiques a priori dans le modèle ?) et (d) le passage à l'échelle (comment rendre les solutions proposées capables de traiter une quantité massive de données ?)
Thanks to high resolution imaging systems and multiplication of data sources, earth observation(EO) with satellite or aerial images has entered the age of big data. This allows the development of new applications (EO data mining, large-scale land-use classification, etc.) and the use of tools from information retrieval, statistical learning and computer vision that were not possible before due to the lack of data. This project is about designing an efficient classification scheme that can benefit from very high resolution and large datasets (if possible labelled) for creating thematic maps. Targeted applications include urban land use, geology and vegetation for industrial purposes.The PhD thesis objective will be to develop new statistical tools for classification of aerial andsatellite image. Beyond state-of-art approaches that combine a local spatial characterization of the image content and supervised learning, machine learning approaches which take benefit from large labeled datasets for training classifiers such that Deep Neural Networks will be particularly investigated. The main issues are (a) structured prediction (how to incorporate knowledge about the underlying spatial and contextual structure), (b) data fusion from various sensors (how to merge heterogeneous data such as SAR, hyperspectral and Lidar into the learning process ?), (c) physical plausibility of the analysis (how to include prior physical knowledge in the classifier ?) and (d) scalability (how to make the proposed solutions tractable in presence of Big RemoteSensing Data ?)

Les progrès des satellites d'observation de la Terre à haute résolution et la réduction des temps de revisite introduite par la création de constellations de satellites ont conduit à la création quotidienne de grandes quantités d'images (des centaines de Teraoctets par jour). Simultanément, la popularisation des techniques de Deep Learning a permis le développement d'architectures capables d'extraire le contenu sémantique des images. Bien que ces algorithmes nécessitent généralement l'utilisation de matériel puissant, des accélérateurs d'inférence IA de faible puissance ont récemment été développés et ont le potentiel d'être utilisés dans les prochaines générations de satellites, ouvrant ainsi la possibilité d'une analyse embarquée des images satellite. En extrayant les informations intéressantes des images satellite directement à bord, il est possible de réduire considérablement l'utilisation de la bande passante, du stockage et de la mémoire. Les applications actuelles et futures, telles que la réponse aux catastrophes, l'agriculture de précision et la surveillance du climat, bénéficieraient d'une latence de traitement plus faible, voire d'alertes en temps réel.Dans cette thèse, notre objectif est double : D'une part, nous concevons des architectures de Deep Learning efficaces, capables de fonctionner sur des périphériques de faible puissance, tels que des satellites ou des drones, tout en conservant une précision suffisante. D'autre part, nous concevons nos algorithmes en gardant à l'esprit l'importance d'avoir une sortie compacte qui peut être efficacement calculée, stockée, transmise au sol ou à d'autres satellites dans une constellation.Tout d'abord, en utilisant des convolutions séparables en profondeur et des réseaux neuronaux récurrents convolutionnels, nous concevons des réseaux neuronaux de segmentation sémantique efficaces avec un faible nombre de paramètres et une faible utilisation de la mémoire. Nous appliquons ces architectures à la segmentation des nuages et des forêts dans les images satellites. Nous concevons également une architecture spécifique pour la segmentation des nuages sur le FPGA d'OPS-SAT, un satellite lancé par l'ESA en 2019, et réalisons des expériences à bord à distance. Deuxièmement, nous développons une architecture de segmentation d'instance pour la régression de contours lisses basée sur une représentation à coefficients de Fourier, qui permet de stocker et de transmettre efficacement les formes des objets détectés. Nous évaluons la performance de notre méthode sur une variété de dispositifs informatiques à faible puissance. Enfin, nous proposons une architecture d'extraction de graphes routiers basée sur une combinaison de Fully Convolutional Networks et de Graph Neural Networks. Nous montrons que notre méthode est nettement plus rapide que les méthodes concurrentes, tout en conservant une bonne précision
The recent advances in high-resolution Earth observation satellites and the reduction in revisit times introduced by the creation of constellations of satellites has led to the daily creation of large amounts of image data hundreds of TeraBytes per day). Simultaneously, the popularization of Deep Learning techniques allowed the development of architectures capable of extracting semantic content from images. While these algorithms usually require the use of powerful hardware, low-power AI inference accelerators have recently been developed and have the potential to be used in the next generations of satellites, thus opening the possibility of onboard analysis of satellite imagery. By extracting the information of interest from satellite images directly onboard, a substantial reduction in bandwidth, storage and memory usage can be achieved. Current and future applications, such as disaster response, precision agriculture and climate monitoring, would benefit from a lower processing latency and even real-time alerts.In this thesis, our goal is two-fold: On the one hand, we design efficient Deep Learning architectures that are able to run on low-power edge devices, such as satellites or drones, while retaining a sufficient accuracy. On the other hand, we design our algorithms while keeping in mind the importance of having a compact output that can be efficiently computed, stored, transmitted to the ground or other satellites within a constellation.First, by using depth-wise separable convolutions and convolutional recurrent neural networks, we design efficient semantic segmentation neural networks with a low number of parameters and a low memory usage. We apply these architectures to cloud and forest segmentation in satellite images. We also specifically design an architecture for cloud segmentation on the FPGA of OPS-SAT, a satellite launched by ESA in 2019, and perform onboard experiments remotely. Second, we develop an instance segmentation architecture for the regression of smooth contours based on the Fourier coefficient representation, which allows detected object shapes to be stored and transmitted efficiently. We evaluate the performance of our method on a variety of low-power computing devices. Finally, we propose a road graph extraction architecture based on a combination of fully convolutional and graph neural networks. We show that our method is significantly faster than competing methods, while retaining a good accuracy

碩士
國立中央大學
系統生物與生物資訊研究所
98
G-protein-coupled receptors (GPCRs) that belong to seven transmembrane receptors, mediate a variety of extracellular signals to induce intracellular responses. Only 10% of GPCRs are targeted bt 50% of current marked drugs, emphasizing the potential of the remaining 90% of the GPCR superfamily for drug targets. The difficulties in large producing GPCRs in vitro and in generating crystal structures hinder structural studies of GPCRs. More complicatedly, GPCRs can form homo- or heteromers for function. The responses between homomers and heteromers or between oligomers and monomers are different while using the same agonist to stimulate. The oligomeric potential of GPCRs allows for more complex ligand-receptor relationships and signaling pathways. Four proton-sensing GPCRs (OGR1, GPR4, TDAG8, G2A) are identified to have fully activation at pH6.4~pH6.8 and important to pH homeostasis and acid-induced pain. Among these genes, primary sequence of G2A is less close to the others, and four of five critical histidine residues that are involved in pH sensing of OGR1 are replaced by other amino acids in G2A. Unexpectedly, G2A is the only proton-sensing GPCR that does not generate any significant responses after acid stimulation in cells overexpressing G2A gene, but did so in OGR1. Whether G2A does respond proton or forms heteromers with other family receptors (such as OGR1) to be functional, remains unclear. The objective of this thesis are (1) to purify proton-sensing GPCRs using bacteria expression system for structural analysis and (2) to explore whether G2A responds acid stimulation using ligand-mediated internalization technique and whether G2A form a functional heteromer with OGR1 to respond acid stimulation using FRET acceptor photobleaching technique. I have found that (1) bacteria expression system was not suitable for in vitro expression of proton-sensing GPCRs; (2) G2A did not internalize into the cytosol in response to acid stimulation, but OGR1 and TDAG8 did so; (3) G2A and OGR1 can form heteromers in resonse to acid stimulation.

碩士
國立中央大學
生命科學研究所
100
T-cell death-associated gene 8 (TDAG8) is a member of proton-sensing G-protein coupled receptors, which are sensitive to acid stimulation. TDAG8 can also respond to the lysophospholipid, 1-β-D-Galactosylacyosylsphingosine (Psychosine), an intermediate of cerebrosides biosynthesis. A model was proposed in which the receptors have two ligand-binding sites, one for protons and the other for lipids. The liplids were suggested to interact with both sites, as agonist and antagonist, respectively. The aim of this study is toinvestigate whether psychosine acts on TDAG8 as an agonist and antagonized proton response. We stimulated cells with psychosine in the presence or absence of proton. Cells were stimulated with psychosine in presence or the absence of proton. The results have demonstrated that psychosine in presence or the absence of proton. The results have demonstrated that psychosine can act on TDAG8. Psychosine can inhibit proton-induced [Ca2+] increase in N2A cells, and vice versa.

碩士
國立中央大學
生命科學系
103
Chronic pain induced by nerve injury, cancer and arthritis is a serious clinical problem disturbing many people. Tissue acidosis is a major factor contributing to chronic pain. Although, a previous study in muscle pain model suggested that ASIC3 and TRPV1 are involved in hyperalgesic priming. However, the detailed mechanism for hyperalgesic priming or other factors involved in priming remains unclear. In this study, I used intraplantarly dual acid injection to explore the roles of ASIC3, TRPV1, and TDAG8 in hyperalgesic priming. Dual acid injection in wild-type mice induced a long-term (13 days) mechanical hyperalgesia. Both of ASIC3-/- and TRPV1-/- mice shortened mechanical hyperalgesia after dual acid injection. TDAG8 was involved in the initiation of mechanical hyperalgesia induced by acid. We also found that small molecule compounds, NSC745885 and 887 inhibited mechanical hyperalgesia induced by acid, CFA, or nerve injury. NSC745885 specifically inhibited TDAG8 expression and function. Accordingly, TDAG8 is involved in the initiation of hyperalgesia, whereas ASIC3 and TRPV1 are involved in hyperalgesic priming.

碩士
國立中央大學
生命科學研究所
95
Tissue injury and inflammation often cause an increase of hydrogen ion concentration in local tissues, called tissue acidosis. Tissue acidosis seems to be the dominant factor that leads to painful sensation. Two cation channels,acid-sensing ion channel(ASIC3) and vanilloid receptor 1, are activated by proton and involved in nociceptive transduction. Recently, a subfamily of G-protein-coupled receptor (GPCR) including OGR1, GPR4, TDAG8 and G2A has been identified as proton-sensing receptors. G2A is originally known as a lysophosphatidylcholine (LPC) receptor and is also actived by proton and 9-hydroxyoctadecadienoic. However, original studies for LPC and proton cannot be reproducible. Whether G2A is activated by proton and whether it is involved in nociception remain unclear. The objective of this thesis is to determine effects of pH on G2A and study its expression pattern. From RT-PCR results, G2A was expressed in many tissues including dorsal root ganglia (DRG). G2A was predominantly expressed in small-diameter, IB4-positive nociceptors. Interestingly, expression levels of G2A increased in ASIC3-/-DRG. This increase is due to an increase in G2A-expressing neurons, mainly in large diameter neurons. Accordingly, G2A may be involved in nociception. Consistent with previous studies, I have found G2A cannot be activated by proton.

博士
國立中央大學
生命科學系
104
Tissue injury and inflammation raise local proton concentration (called tissue acidosis) and accompany with painful sensations. Tissue acidosis is a dominant factor that contributes to pain. Transient receptor potential vanilloid 1 (TRPV1) and acid-sensing ion channel 3 (ASIC3), one member of ASIC family, are proved to be related to acid-induced pain. Proton-sensing G-protein-coupled receptors (GPCRs) consists of ovarian cancer G-protein-coupled receptor 1 (OGR1), GPR4, G2A and T-cell death associated gene 8 (TDAG8). Our previous study indicates that the OGR1 family are expressed in nociceptors of DRG, and are co-localized with TRPV1 and ASIC3. TDAG8 activation sensitize TRPV1 response to capsaicin. In complete Freund’s adjuvant (CFA)-induced inflammation, prolonged hyperalgesia in mice is regulated by PKA and PKC. The switch time for PKA and PKC dependency is about 3 to 4 hours. Acute hyperalgesia induced by acidic solution (pH 5.5 or 5.0) depended on both PKA and PKC, as for prolonged hyperalgesia induced by CFA. The switch time for PKA and PKC dependency is about 2 to 4 hours. Therefore, the switch of PKA and PKC dependency in prolonged hyperalgesia induced by CFA can be due to acidosis signals. The Gs-AC-PKA pathway may be responsible for the early phase of hyperalgesia and Gi- PLC- PKC pathway for the late phase. Taken together, proton-sensing GPCRs might be the candidate to be involved in these two pathways. In this study, I have found the dominant role of TDAG8 to mediate proton-induced calcium signals after 2 hours of CFA injection, and TDAG8 is involved in the case after 24 hours of CFA injection as well. Co-expression of OGR1 and G2A increases the sensitivity to proton, and the magnitude of intracellular calcium signals. Co-expression of OGR1 and G2A is involved in a Gi- PLC- PKC pathway. OGR1 and G2A heteromer is the candidate responsible to the Gi- PLC- PKC pathway observed in inflammation.

碩士
國立中央大學
生命科學研究所
94
Tissue injury and inflammation often raise local proton concentration (called tissue acidosis) and accompany with painful sensations. Tissue acidosis is a dominant factor that contributes to pain. Vanilloid receptor 1 (VR1) and acid-sensing ion channel 3 (ASIC3), one member of ASIC family, are proved to be related to acid-induced pain. However, acid-induced pain is not inhibited in ASIC3 or VR1 gene deletion. Therefore it would be interesting to know whether proton-sensing GPCRs are involved in acid-induced nociception. Proton-sensing GPCRs, including ovarian cancer G-protein-coupled receptor 1 (OGR1), GPR4, G2A, and T cell death associated gene 8 (TDAG8), are originally identified as lysophospholipid receptors. Using RT-PCR and quantitative PCR, I have found that mouse OGR1, GPR4, G2A, and TDAG8 are expressed in dordal root ganglion (DRG). Among the four genes, OGR1 has the highest expression levels in DRG, suggesting that OGR1 may have a role in sensory responses. The localization of OGR1 gene in DRG neurons was examined using in situ hybridization and the results show that 35% small-diameter and 21% large-diameter neurons have OGR1 expression. Since small-diameter neurons are related to nociception, the major function of mOGR1 is probably involved in nociception. Both of proton and sphingosylphosphatidylcholine (SPC) can activate OGR1 to increase intracellular calcium concentration, and they are competitive agonists for OGR1.

碩士
國立中央大學
生命科學研究所
95
Tissue injury, inflammation, or ischemia usually cause an increase in local proton concentration, which is called tissue acidosis. This phenomenon often companies with disease and painful sensation. The proton has been demonstrated as the main factor of acid-induced pain. Acid-sensing ion channel 3 (ASIC3), a member of ASICs family, and vanilloid receptor 1 (VR1) are the proton-sensing receptors. Deficiency of the two genes can not completely eliminate acid–induced pain. It is possible that other molecules involved in acid–induced pain. OGR1 family that belongs to G protein-coupled receptors including ovarian cancer G protein-coupled receptor 1 (OGR1), G protein-coupled receptor 4 (GPR4), G2A, and T cell death associated gene 8 (TDAG8), has been reported as the proton-sensing receptors. The previous study in our lab has found that OGR1 family members are expressed in neuronal tissues, including dorsal root ganglion (DRG). Among these, OGR1 has the highest gene expression levels. However, whether OGR1 and GPR4 are located in nociceptors and their function in DRG remain unclear. Therefore, the objective of the thesis is to determine localization of OGR1 and GPR4 and to study their signaling pathways in primary DRG culture. I have found that OGR1 and GPR4 were mainly expressed in non-peptidergic, small-diameter nociceptors. Approximately 31%~40% of total DRG neurons contain at least two receptors of OGR1 family. A subset of OGR1 family members were co-localized with ASIC3 or VR1. In primary culture experiments, no clear conclusion was found.

Polymer electrolyte membrane (PEM) fuel cells are being studied for use as high efficiency power plants in alternative fuel vehicles. To maintain high efficiency the operating temperatures of the membranes in these fuel cells must be closely monitored and controlled. However, the environment inside of the fuel cell is not favorable for traditional temperature sensing, so a new optical-fiber-based, distributed temperature sensor was proposed to address this need. This thesis investigates the properties of neodymium-doped optical fiber for use as a distributed temperature sensor for PEM fuel cells. The optical absorption spectrum was measured to identify the energy band structure and determine upconversion excitation schemes. The temperature coefficient of the Nd³⁺-doped silica fiber fluorescence decay time was measured for several bands of emission. Finally, two-photon upconversion was attempted from the lower excited states of Nd:YAG and Nd:silica.
Graduation date: 2012

碩士
國立中央大學
生命科學研究所
94
High concentrations of hydrogen ions are usually found in local area surrounding the insulted tissues during several inflammatory responses. The tissue acidosis phenomenon is thought to be associated with inflammatory pain because the proton molecule can activate nociceptors (pain-related neurons) directly through proton-sensing ion channels. Several proton-sensing receptors expressed in dorsal root ganglia neurons. Some of them, such as vallinoid receptor1 and acid-sensing ion channel 3 are thought to be involved in inflammatory hyperalgesia. However, it is still unclear that whether there are any other proton-sensing receptors involved in inflammation. Here we have demonstrated that a proton-sensing G-protein-coupled receptor, mouse T-cell Death Associated gene 8 (mTDAG8), was predominantly expressed in small-diameter neurons, which give rise to the majority of nociceptors. The transcripts of mTDAG8 were increased 24 hours after complete Freund’s Ajuvant (CFA)-induced inflammation, suggesting that the TDAG8 receptors might associate with inflammatory pain. Consistently, data of in situ showed that the total TDAG8-expressing neurons increased approximately 15% in DRG neurons after CFA-inflammation. Most of the increased TDAG8-expressing neurons are large-diameter neurons (9-10%), and the increased small-diameter neurons expressing TDAG8 are restricted in IB4-positive neurons (5-7%). Since the large-diameter neurons can be activated by non-noxious stimulations during inflammation, and the responses of IB4-positive neurons are enhanced after treating with inflammatory mediators, the increased number of neurons expressing TDAD8 receptors may associate with allodynia and hyperalgesia in inflammation. In addition, the mTDAG8-transfected HEK 293 cells accumulated the cAMP in responding to pH 6.0 buffers. Thus the TDAG8 may mediate certain responses through cAMP-pathway in neurons after inflammation.

碩士
國立勤益科技大學
冷凍空調系
106
The main purpose of this study is to make a controller for 500W oxyhydrogen fuel cell. The controller uses ARDUINO MEGA2560 control board for development and system construction. The controller can measure voltage, current, temperature and its power supply. The measurement data is stored in the computer via Bluetooth transmission technology, and the protection components are installed outside the control panel. The components are integrated by using a PCB (Printed circuit board) process. The self-made controller can avoid overloading the hydrogen-oxygen fuel cell. And instantly transfer the data to the user record. The oxyhydrogen fuel cell controller that developed in this study can control the proper opening/closing time of the exhaust valve according to the fuel battery under different electric load conditions. Therefore the fuel cell can maintain the performance of output power more stable and better.

In this Ph.D. dissertation, the feasibility investigation, design and characterization of different microwave and optical resonator devices with applications in the fields of medicine, such as cancer radiotherapy, and diagnostic, such as chemical/biological fluid sensing, is detailed. Different microwave and optical resonant structures have been considered, the common thread among them is related to the electromagnetic field theory and the exploitation of the resonance effect to improve their performance. Ad-hoc homemade computer codes have been developed, for accurate investigations, and validated via experimental data. Finally, the design and optimization of side-coupled proton linear accelerator microwave cavities via a novel hybrid numerical/analytical approach is reported. Such microwave cavities are typically used in proton linear accelerators devoted to hadron therapy applications. The design hybrid approach has been validated through measurements. An excellent agreement between simulation and experiment has been found in terms of accelerator frequency and accelerating field nonuniformity. By exploiting the same foregoing hybrid approach, the design and optimization of a novel proton linear accelerator based on on-axis coupled electromagnetic band-gap (EBG) cavities for hadron therapy applications is also reported. The use of EBG cavities allows a very strong reduction (by about 65%) of the peak surface electric field, paving the way to the design and fabrication of very high gradient proton linear accelerators. The design of optical whispering gallery mode (WGM) microresonators efficiently and selectively excited via tapered optical fibers and long period gratings is illustrated. The design has been well validated via experimental data. A microbubble-based set-up for chemical and biomedical fluid sensing has been also investigated. By proper coupling the WGMs with the tapered fiber modes, resonance shifts higher than i) −40 GHz/wt.% at 1550 nm and ii) −3 GHz/wt.% at 589 nm, have been calculated for a sodium chloride (NaCl) and glucose (C6H12O6) fluid sensing set-ups, respectively.