Dissertations / Theses on the topic 'Gastrointestinal Imaging'

Positron Emission Tomography (PET) with 18F-FDG has emerged as a powerful tool in oncology. Furthermore, recent advent of PET/CT and novel tracers are continually expanding its role. This thesis investigates its application in two solid cancer models. In the diagnosing of primary pancreatic cancer, 18F-FDG PET/CT was shown to be more accurate than conventional CT. It did not add information to locoregional staging of disease but impacted management of patients with potentially operable tumours, by accurately confirming the presence / absence of metastases. In the pre-operative staging of patients with colorectal liver metastases (CLM), 18F- FDG PET/CT was also superior to CT in assessing extrahepatic disease, where it again impacted management. The accuracy of detecting hepatic disease was similar for both. Compared to PET alone, PET/CT improved the accuracy of lesions localization and interpretation. Next, the feasibility of imaging with the novel thymidine analogue tracer 18F-FLT was investigated. Overall, 18F-FLT PET was less accurate than 18F-FDG in detecting lesions in both cancer types, thus suggesting it to be an unsuitable tracer for routine diagnosis and staging. In the cohort of pancreatic cancer patients, 18F-FLT uptake (SUVs) were found to strongly correlate with the immunohistochemical proliferation marker, Ki-67 antigen. This supported 18F-FLT‟s potential role as a surrogate marker of proliferation. The prognostic implications of these require further investigation. Finally, an in vitro model was use to examine early changes in 18F-FLT uptake in response to treatment with cytotoxics. At 2 hours following pulse treatment with 5-fluorouracil, (and before changes in cell numbers and cell cycle phase were seen), a dose dependent increase in 18F-FLT uptake was seen. No change was observed with 18F-FDG nor following Cisplatin treatment. This adaptive response may have a role as an early predictor of response to 5-FU (and potentially other antimetabolites), which requires further investigation.

Medical imaging using Terahertz frequency radiation is in its infancy. Optical adjuncts to enhance the diagnostic precision of white light endoscopy are not new and almost all wavelengths of the electromagnetic spectrum have, at some point, been investigated to this end. The ultimate aim of an endoscopic technique is to identify 100% of mucosal lesions and to classify them with 100% specificity. Methods: This thesis examined the sensitivity and specificity of current technology to accurately identify colonic pathology using a diagnostic precision analysis of published data. The effect of biomaterials such as blood, mucus and faeces on Terahertz radiation was assessed using human tissue samples from health volunteers. The ability to discriminate pathological from normal colonic mucosa was assessed using terahertz radiation to interrogate excised samples of human colon. The physical nature of variation between pathological and normal colonic mucosa was assessed using histological markers in an attempt to identify the cause of terahertz radiation contrast. Results: Current technology has a sensitivity of 95% and a specificity of 78% in a non- inflamed colon. Terahertz pulsed imaging is a sensitive and specific technique for in vitro classification of colonic mucosal pathology. TPI may be a useful adjunct in the presence of inflamed mucosal tissue. Although there were too few data from the present study to quantify any potential benefit. The effects of blood, stool and mucus on TPI are similar but less pronounced than water. Immunohistochemical analysis has demonstrated an association between vascular and lymphatic density with neoplasia. This may be a mechanism for TPI discrimination of colonic pathology. TPI could in the future contribute to a minimally invasive method of in vivo diagnosis of colonic dysplasia or malignancy. Larger scale and in vivo trials of the technology are necessary to further investigate the potential clinical benefit.

The relationship between meal structure and composition can modulate gastrointestinal processing and the resulting sense of satiety. This applies also to the fat component of meals and particularly to the surface area available for digestion. The main hypothesis underpinning this thesis work was that fat emulsion droplet size has a profound effect on fat digestion and, in turn, on the gastrointestinal and satiety responses. To test this hypothesis two fat emulsion meal systems were used. They had exactly the same composition but a small (termed the Fine emulsion, with a droplet size of 400 nm) or a large (termed the Coarse emulsion, with a droplet size of 8 μm) emulsified fat droplet size. The two fat emulsion systems were manufactured and characterised using a range of bench techniques, in vitro digestion models and MRI techniques in vitro. The difference in microstructure caused different temporal creaming characteristics for the emulsions and different percentage hydrolysis profiles in a gastric digestion model in vitro. The Fine emulsion showed initial rapid hydrolysis whilst the Coarse emulsion showed an initial slow hydrolysis phase with the hydrolysis rate increasing at later stages. This indicated that there was indeed a droplet size effect on fat hydrolysis whereby the smaller droplet size with a larger surface area hydrolysed faster than a larger droplet size. The emulsions’ performance was finally tested in vivo in healthy volunteers using MRI in a series of pilot studies leading to a main physiological study. Creaming differences in the gastric lumen were addressed by redesigning the meals using a locust bean gum (LBG) thickener that made them stable throughout the gastric emptying process. A main three-way physiological and satiety study in healthy volunteers showed that a highly emulsified, intragastrically stable emulsion delayed gastric emptying, increased small bowel water content and reduced consumption of food at the end of the study day. Finally, magnetic resonance imaging, relaxometry and spectroscopy were further evaluated to assess fat emulsion parameters in vitro and in vivo in the gastric lumen. Main static magnetic field and droplet size effects on T2 relaxation times of the Fine and the Coarse emulsions were observed. There was reasonable correlation between m-DIXON and spectroscopy methods to quantify fat fraction both in vitro and in vivo. Differences in T2 relaxation times for different droplet sizes of 20% fat emulsions were detected in vitro. These changes were however difficult to separate from creaming effects in vivo with a view of drawing meaningful inferences on droplet sizes. The main conclusion from this work was that manipulating food microstructure especially intragastric stability and fat emulsion droplet size can influence human gastrointestinal physiology and satiety responses and that MRI and MRS provide unique non invasive insights into these processes. This improved knowledge could help designing foods with desired health-promoting characteristics which could help to fight the rising tide of obesity.

Gastrointestinal endoscopy is a vital tool for the detection and treatment of early neoplasia in the upper and lower gastrointestinal tracts. Survival from gastrointestinal cancer is largely dependent on the stage at diagnosis – thus detection of early lesions, or better still treatment of pre-cancerous lesions is vital to improve outcomes. The world of endoscopy is changing rapidly with the development of dye-based and digital enhancement techniques with the aim of improving the detection and characterisation of neoplastic lesions. This thesis reviews the development of all the main advanced imaging techniques and comprehensively reviews the evidence for the use of these in the colon. Chapter 4 describes the development of a new classification system for characterising small colonic polyps using the Pentax i-Scan digital enhancement system. Chapter 5 describes a prospective cohort study using i-Scan for the in-vivo characterisation of small colonic polyps in 87 patients. No differences in the accuracy of polyp characterisation between high-definition white light endoscopy, i-Scan and chromoendoscopy were found when performed by an expert endoscopist. All 3 modalities met the ASGE criteria for management based on optical diagnosis. Chapter 6 describes a retrospective study looking at factors influencing polyp and adenoma detection in a large UK Bowel cancer screening cohort. Endoscope definition (standard vs high definition) was examined in particular. In this group of patients, endoscope definition was found to have no impact on any outcome measure, but endoscopist and bowel preparation were consistent predictors of key quality outcomes. Chapter 7 describes a randomised controlled clinical trial recruiting 126 patients of a pre-endoscopy drink containing water, n-acetyl cysteine and simeticone. This was shown to significantly improve mucosal visibility compared to water alone, or no preparation, and also significantly reduced the need for procedural fluid flushes. Chapters 8 & 9 describes two studies examining the baseline performance and impact of training modules on the accuracy of colonic lesion and polyp characterisation amongst non-endoscopists and endoscopists with varying degrees of experience. Baseline performance of experienced endoscopists was found to be no different to that of inexperienced endoscopists and novices in both studies, but training improved accuracy in all groups. These studies highlight the need for training in lesion characterisation to become part of the formal training programme for all endoscopists.

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2018.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 139-149).
Diseases of the gastrointestinal (GI) tract are typically diagnosed by random biopsy of tissue, which samples only a small area and often misses focal neoplasias. Existing endoscopic visualization tools including white light endoscopy, narrowband imaging and confocal laser endomicroscopy have enabled in vivo assessment to guide biopsies, but suffer from technical limitations and have demonstrated suboptimal sensitivity and specificity to neoplasia. Optical Coherence Tomography (OCT) can generate in vivo, 3-dimensional microscopic imaging. Recent efforts in ultrahigh-speed OCT systems for endoscopic applications have shown promise, but devices had limited fields of view and imprecise beam scanning mechanisms, limiting image quality and coverage. This thesis develops a wide range of new fiber optic devices that substantially extend OCT capabilities in the GI tract, either by greatly increasing field of view for wide field mapping of entire luminal organs, or achieving high precision 2-D beam scanning with compact actuators for in vivo microscopy. Piezoelectrically actuated fiber scanning devices enable forward viewing for focal inspection, while micromotor actuators combined with pneumatic or piezoelectric mechanisms enclosed in tethered capsules generate side viewing over large areas. The work also advances the emerging paradigm of gastrointestinal screening without use of sedation, which promises to lower costs of screening and improve access for a broader population. Design, fabrication and benchtop evaluation of devices, as well as pre-clinical and clinical imaging protocols are reported. Results from validation studies in living swine, and human patients in collaboration with the Veterans Affairs Boston Healthcare System are discussed. The thesis work demonstrates new imaging modalities for in vivo detection and diagnosis of GI pathology that could have important applications in disease screening, surveillance, and therapeutic procedures.
by Kaicheng Liang.
Ph. D.

La fistule anastomotique (FA) est une complication grave de la chirurgie. Une perfusion locale adéquate est fondamentale pour réduire le risque de FA. Cependant, les critères cliniques ne sont pas fiables pour évaluer la perfusion intestinale. À cet égard, l'angiographie par fluorescence (AF) a été explorée. Malgré des résultats prometteurs dans les essais cliniques, l'évaluation de l'AF est subjective, d'où l'incertitude quant à son efficacité. L'AF quantitative a déjà été introduite. Cependant, elle est limitée par la nécessité d'injecter un fluorophore. L'imagerie hyperspectrale (HSI) est une technique d'imagerie optique prometteuse couplant un spectroscope à une caméra photo, permettant une analyse quantitative des tissus en temps réel et sans contraste. L'utilisation intraopératoire de l'HSI est limitée par la présence d'images statiques. Nous avons développé la hyperspectral-based enhanced reality (HYPER), pour permettre une évaluation précise de la perfusion intraopératoire. Cette thèse décrit les étapes du développement et de la validation d'HYPER
Anastomotic leak (AL) is a severe complication in surgery. Adequate local perfusion is fundamental to promote anastomotic healing, reducing the risk of AL. However, clinical criteria are unreliable to evaluate bowel perfusion. Consequently, a tool allowing to objectively detect intestinal viability intraoperatively is desirable. In this regard, fluorescence angiography (FA) has been explored. In spite of promising results in clinical trials, FA assessment is subjective, hence the efficacy of FA is unclear. Quantitative FA has been previously introduced. However, it is limited by the need of injecting a fluorophore. Hyperspectral imaging (HSI) is a promising optical imaging technique coupling a spectroscope with a photo camera, allowing for a contrast-free, real-time, and quantitative tissue analysis. The intraoperative usability of HSI is limited by the presence of static images. We developed hyperspectral-based enhanced reality (HYPER), to allow for precise intraoperative perfusion assessment. This thesis describes the steps of the development and validation of HYPER

In the medical field, the detection and diagnosis of diseases continue to improve. Developments in diagnostic techniques have helped to improve treatment in the early stages and avoid many risks to patients. One relatively new diagnostic technique is optical coherence tomography (OCT), which is used in many medical applications to perform internal microstructure imaging of the human body at high resolution (typically 10 micro metre), at high speed and in real time. OCT is non-invasive and can be used as a contact or non-contact technique to obtain an image. In medicine, there are many applications that involve OCT, such as in ophthalmology, gastroenterology, cardiology and oncology. This work demonstrates the design, development and implementation of a high resolution swept laser OCT system for the imaging and diagnosis of tissues in laboratory and clinical experiments. It reports an investigation to measure the thickness of the peritoneal membrane and the use of optical imaging contrast agents such as gold nanorods. There is also an account of the design of an endoscope-catheter fast scanning OCT system for biomedical studies of the gastrointestinal tract and gynaecological areas. These results were achieved by using a swept tuneable laser source with a very high tuning speed of 16 kHz over a wide range of wavelengths: 1260 nm to 1390 nm. The laser sweeps across 110 nm at a 16 kHz repetition rate. The real axial line speed is limited by the source that is used in the OCT system. The axial resolution of the system is 7 µm and its transverse resolution is 15 µm. The bandwidth of the source is up to DeltaGamma = 110 nm, centred at Gamma0 = 1325 nm, and the coherent length is 7 µm. On the sample arm of the interferometer, the swept laser OCT technique is combined with an optical probe and endoscope in order to develop a novel diagnostic imaging device to visualize tissue in vivo for animal and human experimental trials.

Purpose: Hypoxia in colorectal cancer (CRC) and gastroesophageal cancer (GEC) decreases tumour responsiveness to radio and chemotherapy leading to cancer progression and poor prognosis. This is the first study to utilise [18F]FAZA hypoxia radiotracer in patients with CRC and GEC. Methods: Six patients (mean age 68±8 years, 2 males and 4 females) with CRC and 4 patients diagnosed with GEC (mean age 65 years, 3 males and 1 female) were included in the study. [18F]FAZA was synthesised at the John Mallard Scottish PET Centre. After injection with 370 MBq of [18F]FAZA, PET/CT images with 60 min dynamic scan were acquired. In addition, 15 min static scans 2 hr post injection were performed. 3D PET images were reconstructed iteratively using an ordered subset expectation maximization (OSEM) method and fused to the corresponding low-dose CT images. [18F]FAZA uptake parameters including maximum standard uptake value (SUVmax), tumour-to-muscle ratio (T/M), tumour-to-bowel ratio (T/B) and volume of interest (VOI) were measured. Results: 4 out 6 patients with CRC (66%) showed clear uptake of [18F]FAZA in the primary tumour. The mean tumour SUVmax was 2.2±0.91 (range 1.12 - 3.71). The tumour SUVmax was significantly higher compared with muscle and bowel (t(5) =3.11, P=0.03), (t(5) =3.08, P=0.03), respectively. However, tumour SUVmean didn't differ significantly compared with muscle and bowel (t(5) =2.41 , P=0.06), (t(5) =2.46 , P=0.06) respectively. The mean tumour to muscle ratio (T/M) ratio was 1.89±0.64 (range 1.10 - 2.87), while the mean tumour to normal bowel (T/B) was 1.92±0.64 (range 1.08 - 2.74). However, [18F]FAZA did not accumulate in any of the tumours found in patients with GEC. Conclusions: [18F]FAZA PET/CT imaging is suitable and feasible for detecting CRC hypoxic tumour regions with image quality that can be used in clinical practice.

O uso de agentes de contraste em técnicas de diagnóstico por imagem é uma prática médi-ca rotineira. Certos compostos, presentes em agentes de contraste, possuem propriedades paramagnéticas que podem afetar os sinais da tomografia por ressonância magnética, \"Mag-netic Resonance Imaging\"(MRI). Em estudos aplicados ao trato gastrintestinal (GI), os meios de contraste são amplamente utilizados por via endovenosa, mas também podem ser admi-nistrados oralmente. Porém, a adoção do uso oral é limitada, principalmente, porque os agentes convencionais são caros e causam, geralmente, efeitos colaterais. Desta forma, a-presentamos a caracterização e os resultados preliminares da implementação da polpa do fruto da Euterpe olerácea para um possível uso clínico como agente de contraste oral em MRI do trato GI. A polpa da Euterpe olerácea, conhecida como Açaí, de origem amazônica, apre-senta um aumento de sinal de MRI ponderado em T1 equivalente ao do Gd-DTPA e, tam-bém, um decréscimo de sinal em imagens ponderadas em T2. Investigamos propriedades intrínsecas que possam estar correlacionadas com o aumento de sinal em T1 e à opacidade em T2. O espectro de absorção atômica revelou a presença de íons Fe, Mn e Cu no Açaí, o que contribui para o valor susceptométrico encontrado de -4,83 . 10-6. Essa medida fomen-ta a hipótese de que as mudanças de contraste nas imagens são devido à presença de mate-rial paramagnético, revelando um contraste clinicamente satisfatório nas porções superiores do trato GI. Estudos preliminares indicaram que a homogeneidade e a intensidade do sinal da polpa do Açaí (Euterpe olerácea), no estômago, e duodeno, são próximos daquele encon-trado em agentes convencionais. Além disso, ele não apresentou efeito colateral algum. Devido ao aumento de contraste associado ao Açaí, podemos observar a parede gástrica de forma singular. Ainda, este agente contribuiu para o diagnóstico das vias pancreática e biliar em exames de colangiopancreatografia por ressonância magnética, Magnetic Resonance Colan-giopancreatography (MRCP), com seqüências ponderadas em T2, por reduzir o sinal das alças intestinais.
The use of contrast agents is a common practice in medical imaging protocols. Paramagnetic properties of certain compounds present in contrast agents can affect Magnetic Resonance Imaging (MRI) signals. For abdominal applications, they are usually injected, but may also be administered orally. However, their use as a routine technique is limited, mainly due to the lack of appropriate oral contrast agent. Standard agents are expensive and cause, generally, some kind of side effect. We herein present the preliminary characterization and results for implementation of Euterpe olerácea (popularly named Açaí) as a possible clinical oral contrast agent for MRI of the gastrointestinal (GI) tract. The pulp of Açaí, a fruit from the Amazon area, presented an increase in T1-weighted MRI signal, equivalent to that of Gd-DTPA, and a signal decrease in T2-weighted images. We looked for intrinsic properties that could be responsible for the T1 signal enhancement and T2 opacification. Atomic Absorption spectra revealed the presence of Fe, Mn and Cu ions in Açai. The presence of such ions contribute to the susceptometric value found of -4.83 x 10-6 . This finding assents with the hypothesis that image contrast changes were due to the presence of paramagnetic material. The first measurements in vivo demonstrate a clear increase of contrast due to signal intensity and homogeneity in stomach and bowel walls with the pulp of Açaí, which look like the effects related to standard agents. Consistently, the increase in T1-weighted and the opacification in a T2-weighted acquisition was evident, revealing a biphasic contrast on gastric tissues. Besides, the pulp does not present any side effect. It still has contributed to the diagnostic of pancreatobiliary system at Magnetic Resonance Cholangiopancreatography (MRCP), by reducing overlap of the surround tissues and those structures.

Il manque actuellement aux médecins une nouvelle méthode qui rationalise le traitement peu invasif pour en faire des procédures à opérateur unique, assistées par une caractérisation précise des tissus in situ et en temps réel, en situation de prise de décisions dans la gestion du cancer colorectal. Une solution prometteuse à ce problème a été développée par l'équipe AVR (Automatique, Vision et Robotique) du laboratoire ICube, au sein de laquelle l'endoscope interventionnel flexible (fabriqué par Karl Storz) a été entièrement robotisé, permettant ainsi à un seul opérateur de télémanipuler indépendamment l'endoscope et deux instruments thérapeutiques insérables, grâce à unité de contrôle commune. Cependant, l'endoscope flexible assisté par robot est soumis aux mêmes limites de précision diagnostique que les systèmes d'endoscopie standards. Il a été démontré que l'OCT endoscopique présente un potentiel pour l'imagerie des troubles de la voie gastro-intestinale et pour la différenciation de tissus sains des tissus malades. Actuellement, l'OCT se limite à l'imagerie de l'œsophage humain, qui présente une géométrie simple et un accès facile. Ni l'OCT, ni l'endoscope robotisé ne peuvent résoudre à eux seuls les limites de la norme actuelle de soins pour la prise en charge d’un cancer du côlon. La combinaison de ces deux technologies et le développement d'une nouvelle plate-forme pour la détection et le traitement précoce du cancer constituent l'objet principal de cette thèse, avec la vision de développer une console d'imagerie OCT et une sonde de haute technologie intégrée à l'endoscope robotisé. Ce système permet d'obtenir des images de l'intérieur du gros intestin pour la caractérisation des tissus et l'assistance au traitement, permettant ainsi à un seul opérateur d'effectuer une intervention peu invasive en mode télémanipulation
There exists an unmet clinical need to provide doctors with a new method that streamlines minimally invasive endoscopic treatment of colorectal cancer to single operator procedures assisted by in-situ and real-time accurate tissue characterization for informed treatment decisions. A promising solution to this problem has been developed at the ICube laboratory, in which the flexible interventional endoscope (Karl Storz) was completely robotized, so allowing a single operator to independently telemanipulate the endoscope and two insertable therapeutic instruments with a joint control unit. However, the robot-assisted flexible endoscope is subject to the same diagnostic accuracy limitations as standard endoscopy systems. It has been demonstrated that endoscopic optical coherence tomography (OCT) has a good potential for imaging disorders in the gastrointestinal tract and differentiating healthy tissue from diseased. Neither OCT, nor the robotized endoscope can solve the limitations of current standard of care for colon cancer management alone. Combining these two technologies and developing a new platform for early detection and treatment of cancer is the main interest of this work, with the aim of developing a state-of-the-art OCT imaging console and probe integrated with the robotized endoscope. The capabilities of this new technology for imaging of the interior of the large intestine were tested in pre-clinical experiments showing potential for improvement in margin verification during minimally invasive endoscopic treatment in the telemanipulation mode

L'Endomicrocopie Confocale par Minisondes (ECM) permet l'observation dynamique des tissus au niveau cellulaire, in vivo et in situ, pendant une endoscopie. Grâce à ce nouveau système d'imagerie, les médecins endoscopistes ont la possibilité de réaliser des "biopsies optiques" non invasives. Les biopsies traditionnelles impliquent le diagnostic ex vivo d'images histologiques par des médecins pathologistes. Le diagnostic in vivo d'images ECM est donc un véritable challenge pour les endoscopistes, qui ont en général seulement un peu d'expertise en anatomopathologie. Les images ECM sont néanmoins de nouvelles images, qui ressemblent visuellement aux images histologiques. Cette thèse a pour but principal d'assister les endoscopistes dans l'interprétation in vivo des séquences d'images ECM. Lors de l'établissement d'un diagnostic, les médecins s'appuient sur un raisonnement par cas. Afin de mimer ce processus, nous explorons les méthodes de Reconnaissance d'Images par le Contenu (CBIR) pour l'aide au diagnostique. Notre premier objectif est le développement d'un système capable d'extraire de manière automatique un certain nombre de vidéos ECM qui sont visuellement similaires à la vidéo requête, mais qui ont en plus été annotées avec des métadonnées comme par exemple un diagnostic textuel. Un tel système de reconnaissance devrait aider les endoscopistes à prendre une décision éclairée, et par là-même, à établir un diagnostic ECM plus précis. Pour atteindre notre but, nous étudions la méthode des Sacs de Mots Visuels, utilisée en vision par ordinateur. L'analyse des propriétés des données ECM nous conduit à ajuster la méthode standard. Nous mettons en œuvre la reconnaissance de vidéos ECM complètes, et pas seulement d'images ECM isolées, en représentant les vidéos par des ensembles de mosaïques. Afin d'évaluer les méthodes proposées dans cette thèse, deux bases de données ECM ont été construites, l'une sur les polypes du colon, et l'autre sur l'œsophage de Barrett. En raison de l'absence initiale d'une vérité terrain sur le CBIR appliquée à l'ECM, nous avons d'abord réalisé des évaluations indirectes des méthodes de reconnaissance, au moyen d'une classification par plus proches voisins. La génération d'une vérité terrain éparse, contenant les similarités perçues entre des vidéos par des experts en ECM, nous a ensuite permis d'évaluer directement les méthodes de reconnaissance, en mesurant la corrélation entre la distance induite par la reconnaissance et la similarité perçue. Les deux évaluations, indirecte et directe, démontrent que, sur les deux bases de données ECM, notre méthode de reconnaissance surpasse plusieurs méthodes de l'état de l'art en CBIR. En termes de classification binaire, notre méthode de reconnaissance est comparable au diagnostic établi offline par des endoscopistes experts sur la base des Polypes du Colon. Parce que diagnostiquer des données ECM est une pratique de tous les jours, notre objectif n'est pas seulement d'apporter un support pour un diagnostique ponctuel, mais aussi d'accompagner les endoscopistes sans leurs progrès. A partir des résultats de la reconnaissance, nous estimons la difficulté d'interprétation des vidéos ECM. Nous montrons l'existence d'une corrélation entre la difficulté estimée et la difficulté de diagnostic éprouvée par plusieurs endoscopistes. Cet estimateur pourrait ainsi être utilisé dans un simulateur d'entraînement, avec différents niveaux de difficulté, qui devrait aider les endoscopistes à réduire leur courbe d'apprentissage. La distance standard basée sur les mots visuels donne des résultats adéquats pour la reconnaissance de données ECM. Cependant, peu de connaissance clinique est intégrée dans cette distance. En incorporant l'information a priori sur les similarités perçues par les experts en ECM, nous pouvons apprendre une distance de similarité qui s'avère être plus juste que la distance standard. Dans le but d'apprendre la sémantique des données ECM, nous tirons également profit de plusieurs concepts sémantiques utilisés par les endoscopistes pour décrire les vidéos ECM. Des signatures sémantiques basées mots visuels sont alors construites, capables d'extraire, à partir de caractéristiques visuelles de bas niveau, des connaissances cliniques de haut niveau qui sont exprimées dans le propre langage de l'endoscopiste.

In questo lavoro di tesi viene affrontato il tema della fluidodinamica computazionale mediante un modello in 3D basato su geometria e motilità dello stomaco umano al fine di analizzare il comportamento del contenuto gastrico, le cui differenti proprietà reologiche influenzano non solo il flusso dei fluidi gastrici, ma anche la distribuzione delle particelle di cibo all’interno dello stomaco, organo cavo a forma di J: il fine ultimo è dunque quello di fornire uno sguardo introspettivo, con un punto di vista unico, sull’argomento, chiarendo i meccanismi insiti nel processo della digestione gastrica.

博士
國立陽明大學
生物醫學影像暨放射科學系
104
Cross-sectional imaging scan plus quantitative measurement data increases an additional value; it can provide a more accurate or early diagnosis in the field of abdominal diseases (including liver and gastrointestinal tracts). Hepatocellular carcinoma (HCC) is one of the frequently occurred cancers worldwide; each year there are around 500,000 new cases of patients. In Taiwan, HCC is the 2nd most common cause of cancer death, and it is quite malignant cancer, and the average age of onset is between about 50-60 years old. If there is no effective treatment, the duration is usually only 4 to 6 months long from the symptoms onset to death. In most cases, HCC is a multistage disease whose occurrence is linked to environmental, dietary and life-style factors. Unlike other cancers, HCC usually arises on a previously damaged organ, mostly in the setting of chronic hepatopathy, cirrhosis, or in association with hereditary diseases such as hemochromatosis, Wilson’s disease and a-1-antitrypsin deficiency. HCC is based on pathological diagnosis, and its risk factors include: HBV or HCV hepatitis carriers, liver cirrhosis, chronic alcoholism, family history, exposure to aflatoxin, and etc. Although histopathologic criteria for diagnosing classical HCC have not recently changed, new diagnostic guidelines for the suspected cases of HCC has updated in 2010. The full version of the new guidelines is available on the American Association for the Study of Liver Diseases (AASLD) website at http://www.aasld.org/Publications /Practice Guidelines/Hepatocellular Carcinoma, management/HCCUpdate2010.pdf. Based on the AASLD guidelines, biopsy is not needed for tumors with characteristic magnetic resonance (MR) or computed tomography (CT) images. Therefore, accurate pretreatment prediction of histological grade of HCCs based on MR images could aid in the identification of candidates for transplantation. Worse histological grade of HCC correlates with poor patient survival after liver transplantation and is an independent predictor of postoperative recurrence. Patients with a large low-grade hepatic tumor (well-differentiated HCC to moderately-differentiated HCC) have good survival rates after orthotopic liver transplantation, but patients with poorly-differentiated HCC have a dismal prognosis. Several groups have observed that long-term survival may be predicted from the histological grade of HCCs before local treatment or liver transplantation. For this reason, it has been suggested that prognostication and triage should ideally include histological grade along with other factors, such as tumor size, tumor stage, or the presence of portal vein tumor thrombus. Liver transplantation is performed either from a cadaveric or living donor. The incidence of early graft failure has been reported to be in the range of 5 to 10%, and it represents the critical cause of early death following transplantation. Previous studies have shown that the donor characteristics, including age > 65 years, steatosis > 40%, positive viral hepatitis, intensive care unit stay, and history of malignancy, have been presumed to increase the risk of initial graft dysfunction. Being considered separately, prolonged cold ischemia time (>12 hours) has been recognized as an important independent risk factor for early graft failure. However, prior studies did not focus on early postoperative assessment. Occurring in the early postoperative period, graft failure represents a severe form of ischemia-reperfusion injury to the allograft, with clinical and radiographic features similar to the hepatic parenchymal infarction. Though CT is commonly used to evaluate anatomic postsurgical complications, the prognostic value of specific CT findings after liver transplantation has not been described. Therefore, we performed an exploratory study to retrospectively review all CT scans obtained within 3 months after liver transplantation to identify CT imaging findings that predict graft failure within 1 year of transplantation. Finally, in the series of PhD study regarding abdominal images, the methodology is also applied to the small bowel imaging. CT imaging features plus quantitative measurement helps to diagnosis the gastrointestinal diseases (including obscure acute gastrointestinal bleeding and small bowel obstruction). Although intestinal diseases often exhibit similar imaging features, increasing relevant objective data of CT measurement results can increase diagnosis rate and predict whether the patient needs surgery and other active treatments. Specific Aim 1: To investigate the value of gadoxetic acid–enhanced dynamic MRI images and DWI in differentiating the histological grades of HCC in patients with chronic liver disease. In our serial studies, we reviewed the MR images of 79 patients with 141 surgically resected HCCs. The signal intensity and its relationship with histopathological grade were assessed. We measured the apparent diffusion correlation (ADC) values and calculated arterial enhancement ratios, washout ratios, and relative intensity ratios of HCCs relative to the surrounding liver parenchyma in gadoxetic-enhanced MR images in order to determine their relationship to the histological grade. MR imaging for liver was performed with a 1.5-T MR scanner (Philips, the Achieva) and a phased-array body coil. Turbo spin-echo (TSE) T2WI (TR/TE: 1000-1800/110; TSE factor, 24) without and with fat saturation, and coronal T2WI were obtained under respiratory trigger. Dual T1WI (TR/TE: 180-210/2.3 and 4.6; slices thickness 8mm, gap 0.8mm) and fat sat T1WIwere performed during one breath hold. Automatic shimming is used for fat suppression imaging to maximize magnetic field homogeneity. (1) Primovist, 0.1ml/kg for patients but not used for patients less than 18 years old, injected as a bolus, through peripheral vein. Patients underwent dynamic 3D T1-weighted fast field echo (FFE) sequence (5-10/3.3; flip angle, 10°) before, using bolus tracking method for arterial and portovenous phases, about 18~20mesec and 50~55 sec after the contrast enhancement, T1WI/T1WI fat sat 3 minutes later. The postcontrast TIWI and fat sat T1WI performed at about 20 minuets after the injection. (2) DWI sequences: Diffusion-weighted imaging was obtained with a single-shot, spin-echo echo-planar imaging sequence using a parallel technique in the axial plane, a slice thickness of 8 mm, an interslice gap of 0.8 mm, voxel size of 2.5 × 2.5 mm, the number of excitations was 4 and the b-factors=0, 100, 500, 1000 s/mm2. The motion-probing gradients with three orthogonal directions (x, y and z) were applied sequentially. (3) The ADC value was automatically calculated by a computer program. The HCCs were identified on the T2-weighted FSE images and the contrast-enhanced T1-weighted GRE images. The ADC value of the HCC was measured on an ADC map, and the slice's location was identical to that of the selected image on the T2-weighted FSE images and the contrast-enhanced T1-weighted GRE images, respectively. The ADC mapping and eADC mapping was depicted. In our 1st study, we concluded that the quantitative MRI parameters are benefit to predict the histological grade of HCC preoperatively, and relative decrease arterial enhancement on gadoxetic acid-enhanced MR images and low ADC are predictive of worse histological grades of HCC. Specific Aim 2: To identify post-liver transplant CT imaging findings, which predict graft failure within 1 year. And then, we evaluated the CT scans of 202 adult liver transplants performed who underwent CT within the first 3 months after transplantation in our 2nd study. We recorded the CT findings of liver perfusion defect, patchy liver parenchymal hypoattenuation, the diameters of the portal veins, hepatic veins, and inferior vena cava at the anastomosis, periportal edema, and hematoma. Ascites was rated as 0=none to 4=severe. Findings were correlated to graft failure within 1 year of transplantation. Our 2nd study was designed to identify specific CT findings that may predict early graft failure. We found that, for patients who have CT scans within 3 months of liver transplantation, liver perfusion defects and small portal vein diameters independently predict one-year graft failure. RPV diameters less than 11.5 mm and left less than 10.0 mm correlate with poor graft outcome. Specific Aim 3: To compare the treatment outcome in patients with advanced HCCs: Stereotatic body radiotherapy (SBRT) with and without sorafenib administration SBRT is not included as a curative option in the European Association for the Study of the Liver practical guidelines. Of note, concerns have been raised about the possibility of radiation-induced liver disease for patients treated with external-beam radiotherapy, which encumbered the inclusion of this kind of radiotherapy in the therapeutic strategies for cirrhotic patients. This is despite almost 20 years have now passed since Blomgren et al. have pioneered SBRT for liver malignancies. However, the value of the combination therapy of the two treatments (SBRT and sorafenib) in patients with advanced HCC was never been investigated and the additional effect of sorafenib should be further validated. The purpose of our study was to retrospectively compare the treatment outcome of patients with advanced HCC who received SBRT alone or a combined therapy with SBRT and sorafenib in a follow-up study. Most studies published about SBRT for liver tumors have contained both liver metastases and HCC. Articles about SBRT for HCC alone are scarce. Most of these studies are retrospective, like ours, or with a very limited number of patients. We present here the largest study in terms of both patients and the number of lesions treated. Despite its unfavorable prognosis, our study supports that SBRT is feasible in treating advanced HCC, tumor response rate of 50.0% (CR + PR) over the first 6 months. And, the 1-year overall survival rate of 67% in patients with combined therapy was encouraging. Compared to patients who did not receive sorafenib, the patients who underwent SBRT plus sorafenib had significantly prolonged survival time. While patients with a score of Child-Pugh A and those receiving doses of at least 40 Gy were able to achieved a 1-yr local control rate of 66.67% as well as 71.43% were promising. Acute toxicities were mild and tolerable. However, local and regional recurrence remained the major problem. Prospective studies of combination of SBRT with other treatment modalities may be suggested. Specific Aim 4: To use CT features analyses for prediction and localization of obscure acute gastrointestinal bleeding and small bowel obstruction In the evaluation of patients with suspected small-bowel disease, an accurate radiologic examination is important both for recognizing possible small-bowel disease and to help reliably document normal morphology. Small-bowel follow-through and enteroclysis are widely used for small-bowel imaging; however, these examinations provide only indirect information about the bowel wall and surrounding structures and are prone to problems caused by overlapping bowel loops. Compared with the stomach and the colon, the small intestine is an uncommon site for intestinal bleeding, and unless the bleeding is massive, it is often difficult to diagnose. Therefore, such patients may present with prolonged, chronic occult blood loss or recurrent episodes of melena without a specific diagnosis. Although various radiologic studies have been used, a bleeding site cannot be localized in approximately 5-20% of these patients who are therefore classified with obscure gastrointestinal bleeding. Nevertheless, every effort should be made to determine the source of their gastrointestinal bleeding because adequate diagnosis is followed by an improved patient outcome and a decreased need for transfusion. Although CT is still not comparable in sensitivity to scintigraphy using 99mTc-labeled RBCs or to conventional angiography, it may be an alternative to more invasive procedures when routine workup fails to determine the cause of active intestinal bleeding. In our study, we found that obscure gastrointestinal bleeding, unexpected bleeding foci, unexpected tumors, and inflammatory bowel disease can all be easily observed on MDCT. Furthermore, the small bowel is involved in 60-80% of cases of intestinal obstruction. Because of the possibility of strangulation, misdiagnosis or delayed diagnosis can result in lethal or life-threatening complications such as ischemia and perforation. Therefore, in patients with suspected small-bowel obstruction, the role of imaging is to determine the presence or absence of an obstruction; to identify the site, severity, and cause of an existing obstruction; and to detect the possible presence of strangulation.

A feature extraction approach based on color is firstly proposed. Exploiting color histogram of an image, we can obtain distribution of different colors in images. Then we employ minimum distance classifier based on a new distance criterion to judge status of regions. In this section, we also validate benefits of WCE image enhancement to the proposed CAD system.
Finally, we propose a new approach of chrominance moment as another kind of feature to discriminate normal regions from abnormal regions, which makes full use of Tchebichef polynomials and HSI color space. This new feature extraction scheme preserves illumination invariance without numerical approximation.
In conclusion, this thesis investigates several major and challenging problems such as WCE images enhancement and feature extractions in CAD for WCE images, and proposes several novel schemes to solve those problems. Extensive experiments are reported to demonstrate effectiveness of the proposed algorithms.
Next, we investigate automatic diseases detection for WCE images to partially solve the second problem. In this part we explore different features that are suitable for detection of diseases from three viewpoints, i.e., color, texture and chromaticity, because clinicians mainly use these clues to diagnose. At the same time, we introduce their corresponding classifiers.
We further advance a new texture feature extraction method, curvelet based local binary pattern, to detect abnormal regions in WCE images. This method takes advantage of curvelet transform and local binary pattern to describe textural features of WCE images.
Wireless capsule endoscopy (WCE) is a state-of-the-art technology to diagnose gastrointestinal (GI) tract diseases without invasiveness. However, there exist two major problems concerning WCE images. One problem is that many images for diagnosis have rather low contrast and are noisy, which causes difficulties to diagnosis and also to computer-aided detection, so it is necessary to enhance these images. The other one is that the viewing process of video data per examination is very time consuming because of the great amount of video data. If we can use computerized methods to help the physicians detect some abnormal regions in WCE images, it will certainly reduce the burden of physicians. Focusing on these two goals, this thesis mainly studies some main challenging problems in computer-aided diagnosis (CAD) system for WCE images. To solve the first problem, we put forward an adaptive curvature strength diffusion method to enhance WCE images. Based on local characteristics analysis of WCE images, we propose a new concept of curvature strength. Then, we employ curvature strength diffusion to enhance WCE images with an adaptive choice of conductance parameter. Finally, we extend the curvature strength diffusion to color space since WCE images are color images.
Li, Baopu.
Adviser: Max Q. H. Meng.
Source: Dissertation Abstracts International, Volume: 70-06, Section: B, page: 3640.
Thesis (Ph.D.)--Chinese University of Hong Kong, 2008.
Includes bibliographical references (leaves 126-150).
Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web.
Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web.
Abstracts in English and Chinese.
School code: 1307.

無線膠囊內窺鏡（CE）是一種用於檢查整個胃腸道，尤其是小腸的無創技術。它極大地改善了許多小腸疾病的診斷和管理方式，如不明原因的消化道出血，克羅恩病，小腸腫瘤，息肉綜合征等。儘管膠囊內窺鏡有很好的臨床表現，但它仍然有一定的局限性。主要問題是每次檢查產生約50,000 幅低質量的圖像，對於醫生來說，評估如此大量的圖像是一項非常耗時、耗力的工作。
到目前為止，對於膠囊內窺鏡的分析和評估，學者們都把膠囊內窺鏡圖像視為單獨的，獨立的觀測對象。事實並非如此，因為圖像之間往往有顯著的重疊。特別是當膠囊內窺鏡在被小腸蠕動緩緩推動時，它可以捕捉同一病灶的多個視圖。我們的研究目的是使用所有可用的資訊，包括多幅圖像，研究對於膠囊內窺鏡的電腦輔助診斷（CAD）系統。
在這篇論文中，我們提出了一個嵌入分類器的多類隱馬爾可夫模型（HMM）的方案，它可以融合多幅相鄰圖像的時間資訊。由於膠囊內窺鏡圖像的品質比較低，我們首先進行預處理，以加強膠囊內窺鏡圖像，增加其對比度，消除噪聲。我們調查研究了多種圖像增強的方法，並調整了它們的參數使其適用於膠囊內窺鏡圖像。
對於基於單幅圖像的有監督的分類，AdaBoost 作為一個集成分類器來融合多個分類器，即本論文中的支持向量機（SVM），k-近鄰（k-NN），貝葉斯分類。在分類之前，我們提取和融合了顏色，邊緣和紋理特徵。
對於無線膠囊內窺鏡的視頻摘要，我們提出了有監督和無監督的兩類方法。對於有監督方法，我們提出了一個基於隱馬爾可夫模型的，靈活的，可擴展的框架，用於整合膠囊內窺鏡中連續圖像的時間資訊。它可以擴展到多類別，多特徵，多狀態。我們還提出了聯合隱馬爾可夫模型和並行隱馬爾可夫（PHMM）模型對系統進行改進，它們可以被看作是決策級的資訊融合。聯合隱馬爾可夫模型通過多層次的隱馬爾可夫模型，結合不同的資訊來源，對膠囊內窺鏡視頻進行分類和視頻摘要。 並行隱馬爾可夫模型採用貝葉斯推理，在決策時融合多個不同來源的資訊。對於無監督的方法，我們首先提出了一種基於顏色的特徵提取方法。在反色顏色空間中對亮度不變的色度不變矩用來表示膠囊內窺鏡圖像的顏色特徵。接著，我們又提出了一種基於輪廓元（Contourlet）變換的局部二元模式（LBP）作為紋理特徵。在特徵空間中，我們測量了相鄰圖像的距離，並把它視為一個位於二維平面上的開放輪廓上的點。 然後，我們採用一個無參數的關鍵點檢測方法檢測在視頻片段上的突變關鍵點。基於這些突變關鍵點，我們對膠囊內窺鏡視頻進行分割。最後，在每段被分割的視頻片段上，我們通過提取有代表性的關鍵幀來實現膠囊內窺鏡視頻摘要。我們分別用模擬和真實的病人數據進行實驗，對提出的方法進行驗證，結果表明了我們所提出的方案的有效性。它在實現自動評估膠囊內窺鏡圖像上具有很大的潛力。
Wireless Capsule Endoscopy (CE) is a non-invasive technology to inspect the whole gastrointestinal (GI) tract, especially the small intestine. It has dramatically changed the way of diagnosis and management of many diseases of the small intestine, such as obscure gastrointestinal bleeding, Crohn’s disease, small bowel tumors, polyposis syndromes, etc. Despite its promising clinical findings, it still has some limitations. The main problem is that it requires manual assessment of approximately 50,000 low quality images per examination which is highly time-consuming and labor-intense.
CE analysis and assessment so far treated CE images as individual and independent observations. It is obviously not the case as there is often significant overlap among images. In particular, CE captures multiple views of the same anatomy as the capsule is slowly propelled by peristalsis. Our broader work aims to perform computer aided diagnosis (CAD) in endoscopy using all available information, including multiple images.
In this dissertation, a framework of multi-class Hidden Markov Models (HMM) embedded with statistical classifiers for combining information from multiple CE images is proposed. Due to the low quality of CE image, pre-processing is performed to enhance CE images by increasing the contrast and removing noises. Several image enhancement methods are investigated and customized for CE images. For frame-based supervised classification, AdaBoost is used as the ensemble classifier to combine multiple classifiers, i.e. support vector machine (SVM), k-nearest neighbor (k-NN), and Bayes classifier. Before classification, color, edge and texture features are extracted and fused. Finally, both supervised and unsupervised methods are proposed for CE study synopsis. For supervised method, a flexible and extensible framework based on HMM is developed to integrate temporal information in CE images. It can be extended to multi-class, multi-features, and multi-states. Improvements can be made by combined HMM and Parallel HMM (PHMM) which are introduced as decision-level fusion schemes. Combined HMM considers different sources via a multi-layer HMM model to perform classification and video synopsis. PHMM employs Bayesian inference to combine the recognition results at decision level. For unsupervised method, illumination-independent opponent color moment invariants and local binary pattern (LBP) based on Contourlet transform are explored as color and texture features, respectively. Pair-wise image dissimilarity is measured in the feature space and treated as points on an open contour in a 2-D plane. CE video is segmented based on sudden change points which are detected using a non-parametric key-point detection method. From each segment, representative frames are extracted to summarize the CE video. Validation results on simulated and real patient data show promising performance of the proposed framework. It has great potential to achieve automatic assessment for CE images.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Zhao, Qian.
Thesis (Ph.D.)--Chinese University of Hong Kong, 2012.
Includes bibliographical references (leaves 142-175).
Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web.
Abstract also in Chinese.
Abstract --- p.ii
Acknowledgments --- p.vii
List of Tables --- p.xiii
List of Figures --- p.xv
Chapter 1 --- The Relevance of Synopsis --- p.1
Chapter 1.1 --- Problem Statement --- p.1
Chapter 1.2 --- Application - Capsule Endoscopy Assessment --- p.4
Chapter 1.3 --- Literature Review --- p.9
Chapter 1.3.1 --- Methods Based on Frame Classification --- p.11
Chapter 1.3.2 --- Methods Integrating Temporal Information --- p.14
Chapter 1.4 --- Contributions --- p.19
Chapter 1.5 --- Organization --- p.23
Chapter 2 --- Preliminary --- p.25
Chapter 2.1 --- Hidden Markov Model (HMM) --- p.25
Chapter 2.2 --- Factorial HMM --- p.35
Chapter 3 --- Temporal Integration in Capsule Endoscopy Image Analysis --- p.37
Chapter 3.1 --- Pre-processing --- p.38
Chapter 3.2 --- Feature Extraction --- p.43
Chapter 3.3 --- Frame-based Supervised Classification --- p.47
Chapter 3.3.1 --- Supervised Classification using Individual Frames --- p.47
Chapter 3.3.2 --- Ensemble Learning Based on AdaBoost --- p.50
Chapter 3.4 --- Sequence-based Supervised Classification --- p.52
Chapter 3.5 --- Experiments --- p.58
Chapter 3.5.1 --- Capsule Endoscopy Image Enhancement --- p.60
Chapter 3.5.2 --- Frame-based Supervised Classification --- p.67
Chapter 3.5.3 --- Image Sequence Classification --- p.68
Chapter 3.6 --- Discussion --- p.80
Chapter 3.7 --- Summary --- p.82
Chapter 4 --- Capsule Endoscopy Study Synopsis --- p.98
Chapter 4.1 --- Supervised Synopsis Using Statistical Models --- p.98
Chapter 4.2 --- Unsupervised Synopsis via Representative Frame Extraction --- p.100
Chapter 4.2.1 --- Feature Extraction --- p.100
Chapter 4.2.2 --- Non-parametric Key-point Detection --- p.111
Chapter 4.2.3 --- Representative Frame Extraction --- p.112
Chapter 4.3 --- Experiments --- p.119
Chapter 4.3.1 --- Supervised Synopsis Based on HMM --- p.119
Chapter 4.3.2 --- Unsupervised Synopsis --- p.125
Chapter 4.4 --- Discussion --- p.132
Chapter 4.5 --- Summary --- p.133
Chapter 5 --- Conclusions and Future Work --- p.138
Chapter 5.1 --- Conclusions --- p.138
Chapter 5.2 --- Future Work --- p.141
Bibliography --- p.142

The stomach is a digestive organ in the gastrointestinal tract that regulates food intake and paces digestion of nutrients and fluids. The emptying and motility patterns of the stomach are crucial rate-determining processes in maintaining energy homeostasis in the body. Dysregulation of gastric functions often leads to distressing conditions such as gastroesophageal reflux diseases, functional dyspepsia, gastroparesis and obesity. Gastric disorders affect more than 60 million people in the US, producing significant medical and economic burden. These diseases are often chronic and greatly compromise quality of life. As the causes of these diseases remain largely unknown, effects of current pharmacological, dietary, or surgical treatments are often dismal. In this regard, neuromodulation of peripheral nerves emerges as a promising electroceutical therapy for remedying gastric disorders. However, therapeutic effects were shown to be modest, largely due to the inability to validate or calibrate the efficacy and stability of neuromodulation methods with appropriate physiological readouts. To address these problems, here I developed a non-invasive, repeatable online high-resolution magnetic resonance imaging protocol, empowered with advanced image processing algorithms, to track gastric emptying, antral motility, pyloric motility, intestinal filling and absorption in a rat model. The protocol can be used to guide tuning and optimization of stimulation parameters of neuromodulation without perturbing ongoing and spontaneous physiology. The proposed technology and findings are expected to pave the way for the use of gastric MRI to evaluate the efficacy of therapeutics in treating gastric disorders under both preclinical and clinical settings.

Tese de Doutoramento em Informática
A cápsula endoscópica é atualmente o principal método para o diagnóstico do intestino delgado, uma área de difícil acesso devido a dificuldades anatómicas, pela endoscopia tradicional alta e pela colonoscopia. É um método fundamental para a deteção de lesões ou outras anomalias no sistema digestivo, para o acompanhamento prévio e/ou posterior a intervenções cirúrgicas. Resultante de cada diagnóstico, são recolhidas entre 8 a 10 horas de vídeo, que necessitam de posterior visualização por parte do médico. São disponíveis pelas companhias que comercializam cápsulas endoscópicas, sistemas de visualização e apoio à decisão, tendo estes como exemplo, ferramentas de melhoria de contraste da imagem, detetores automáticos de lesões vermelhas, entre outras. Para a deteção de sangramentos existe uma ferramenta no software comercial de visualização Rapid Reader (RR) da empresa Given (atualmente Medtronic) denominada de Suspected Blood Indicator (SBI). Esta explora a presença de determinados valores na componente vermelha da cor, o que permite uma busca rápida na procura de hemorragias. No entanto, esta ferramenta permanece limitada em termos de eficiência, apresentando valores variáveis entre 20% a 56% para a métrica sensibilidade [1]–[3]. Por outro lado, estas estas cápsulas não fornecem por si, a sua localização no interior do paciente. Atualmente os médicos especialistas estimam a localização da cápsula baseandose em referências anatómicas, como o duodeno ou a válvula ileocecal, e na perceção do deslocamento aparente entre imagens do vídeo. Devido à duração do vídeo e à escassez de referências anatómicas ao longo do intestino delgado, é difícil obter uma estimativa precisa da posição da cápsula sem informações adicionais. Os métodos de rastreamento mais usados são baseados em matrizes de sensores com transmissão por radiofrequência (RF), com valores de erro reportados de 3.7 a 11.4 cm [4], [5]. Estas soluções apresentam algumas desvantagens como o desconforto no paciente, o aumento de custos, os problemas de biocompatibilidade ou a interferência com o sensor da câmara da cápsula [6]. Existem algumas soluções baseadas somente de software que mostram precisões de localização semelhantes às que usam equipamentos externos, no entanto com a aplicação de aprendizagem profunda (AP) houve uma melhoria nos desempenhos [7]–[9]. Centrado nestas problemáticas, nesta tese são apresentadas algumas soluções. Em concreto para as lesões vermelhas, são apresentados dois métodos: O primeiro é baseado na mudança da perspetiva de visualização das lesões vermelhas, aplicando a mudança do espaço de cor alicerçado na rotação das cores. A ideia é promover a melhor visualização das características das lesões vermelhas, sob uma perspetiva diferente de visualização em relação espaço de cor RGB, que apresenta correlação entre as cores e que é o espaço de cor nativo dos vídeos endoscópicos. Este método foi testado para vários classificadores de aprendizagem profunda reportados na bibliografia, havendo a verificação dos desempenhos. O segundo método de deteção e segmentação de lesões vermelhas aplica uma abordagem de AP, onde são aplicadas a uma rede neural (RN) um conjunto de imagens anotadas para que esta “aprenda” as características das lesões vermelhas. Obteve-se um modelo que pode inferir, em conjuntos de imagens que não foram usados para treinar a RN, se estas possuem ou não lesões, inferindo também em caso positivo, os seus limites. Em relação à odometria visual para o problema de localização de cápsula endoscópica, foi proposta uma solução onde o deslocamento por imagem (em percentagem) é calculado para todo o intestino delgado. Para isso, primeiro são selecionadas as imagens informativas do conjunto de imagens, seguindo-se a aproximação da homografia entre pares de imagens consecutivas, com uma abordagem de AP para finalmente serem estimados os deslocamentos em percentagem. Esta tese contribui para a evolução de modelos em visão por computador, que auxiliem nas tarefas quotidianas dos médicos, nomeadamente na deteção e segmentação de lesões vermelhas e também na localização da cápsula no interior do trato gastrointestinal (TG). Estas soluções apresentam melhorias no desempenho das ferramentas que são usadas diariamente nos serviços médicos, de modo a melhorar o desempenho dos profissionais que poderão apoiar-se com confiança nos sistemas de apoio à decisão.
The endoscopic capsule is currently the main method for the diagnosis of the small intestine, an area of difficult access due to anatomical difficulties, through high endoscopy and colonoscopy. It is a fundamental method for lesion detection or other abnormalities in the digestive system, for the previous and/or subsequent follow-up of surgical interventions. As a result of each diagnosis, 8 to 10 hours of video data are collected, requiring further visualization by the physician. The companies that commercialize endoscopic capsules also make available visualization and decision support systems, such as image contrast enhancement tools, automatic red lesion detectors, among others. For the bleeding detection, there is a tool in the commercial Rapid Reader (RR) visualization software from Given (currently Medtronic) named Suspected Blood Indicator (SBI). It explores the presence of certain values in the red component of color, which allows a quick search for hemorrhages. However, this tool remains limited in terms of efficiency, presenting values between 20% and 56% for the Sensitivity metric [1]–[3]. On the other hand, these capsules do not provide by themselves their location within the patient. Currently, doctors estimate the location of the capsule based on anatomical references, such as the duodenum or ileocecal valve, and on the perception of the apparent displacement through video images. Due to the length of the video and the scarcity of anatomical references along the small intestine, it is difficult to obtain an accurate estimate of the position of the capsule without additional information. The most commonly used tracking methods are based on radiofrequency sensor arrays, with error values reported from 3.7 to 11.4 cm [4], [5]. These solutions present some disadvantages such as patient discomfort, increased costs, biocompatibility problems or interference with the sensor of the capsule chamber [6]. There are some software-only solutions that present location accuracies similar to those that use external equipment, however, with the application of deep learning, there was an improvement in performances [7]–[9]. Focusing on these issues, some solutions are presented in this thesis. Specifically, for red lesions, two methods are presented: The first one is based on the change of visualization perspective of red lesions, applying a color space, based on the rotation of colors, transformation. The idea is to promote the best visualization of the red lesions characteristics, under a different perspective of visualization in relation to RGB color space, which presents a correlation between the colors and which is the native color space of the endoscopic videos. This method was tested for several deep learning classifiers reported in the bibliography, with performance verification. The second method of detecting and segmenting red lesions applies a deep learning approach, where a set of annotated images are applied to a neural network so that it "learns" red lesions characteristics. It was obtained a model that can infer, in sets of images not used to train the neural network, whether they have lesions, also inferring, in a positive case, their boundaries. Regarding visual odometry for the endoscopic capsule localization problem, a solution was proposed where the image displacement (in percentage) is calculated for the whole small intestine. To achieve this, firstly were selected the informative frames from a set of images, followed by the homography approximation between consecutive pairs of frames applying a learning approach, to finally estimate the displacements in percentage. This thesis contributes to the evolution of models in computer vision, which aid in the physician’s daily tasks, namely in the detection and segmentation of red lesions, and the location of the capsule inside the gastrointestinal tract. These solutions present improvements in the performance of the tools that are used daily in medical services, to improve the professional’s performance, who will be able to rely confidently on decision support systems.

Dissertação de Mestrado Integrado em Engenharia Electrotécnica e de Computadores apresentada à Faculdade de Ciências e Tecnologia
Com a introdução de novas tecnologias na área da saúde, os sistemas para captura e tratamento dos diferentes exames representam uma melhoria no diagnóstico e eficiência prestados ao utente. No mercado existe um conjunto de soluções que visam preencher estas necessidades. Contudo não satisfazem as exigência reais do público, ou são muito dispendiosas. Este trabalho apresenta uma framework de elevada qualidade, baixo custo e escalável, que oferece ao médico um conjunto de ferramentas de fácil utilização que auxiliam o processo de realização de relatório, possibilitando a dupla captura de exame (endoscopia e ecografia). Para isso foi realizada uma análise de requisitos, bem como uma seleção das melhores soluções para cada requisito pretendido. Neste contexto e tendo em conta os diferentes requisitos, desenvolveu-se em C# e Windows Forms (para interface gráfica) uma solução que integra uma base de dados em SQLite, um sistema de autenticação e diferenciação de utilizadores. Realiza também a compressão de vídeo proveniente de 2 dispositivos de captura ligados ao endoscópio e ecógrafo (através de directshow, Accord.NET e FFmpeg). Esta pressupõe uma solução que permita a reprodução do vídeo adquirido e adição de marcadores que auxiliarão o médico na função de diagnóstico, bem como edição de relatório e obtenção deste documento. Todos os resultados obtidos foram satisfatórios: o programa funciona de acordo com o esperado, a compressão realizada é eficaz não afetando o diagnóstico. Evidenciou-se apenas um problema pontual na reprodução de vídeo, cujo a sua resolução se encontra pendente da atualização de uma nova release do Accord.NET. Este problema, em nada influência o desempenho de diagnóstico bem como o resultado do mesmo.Todo o desenvolvimento, integração, interface gráfica e testes foram realizados pelo autor da presente tese.A plataforma desenvolvida encontra-se testada e preparada a ser colocada em produção.
With the introduction of new technologies in health, systems for capturing and processsing the different exams, have brought diferents improvements in the diagnosis and efficiency provided to patients. Presently in the market there are a set of solutions that aim to fill these needs. However, they either do not meet the actual requirements of the public or are expensive. This work presents a high quality, low-cost and scalable framework, which offers health providers a set of user-friendly tools that support the reporting process and enable double exam capture (endoscopy and ultrasound). To develop such framework, a requirement analysis was performed, as well a selection of the best solutions for each requirement. Taking into account the different requirements, a solution that integrates a database in SQLite was developed in C# and Windows Forms with system authentication and user differentiation. It also performs video compression from 2 capture devices connected to the endoscope and echograph (using Directshow, Accord.NET and FFmpeg). This system allows the reproduction of the acquired video and introduction of tags that will assist the physician with the patient diagnosis, as well as report editing and rendering. All the results are satisfactory: the program functions as expected and the compression performed is effective without affecting the diagnosis. There was only a sporadic problem in video playback. Resolution is pending of a new Accord.NET release. This is a minor issue that in no way influences the diagnostic performance as well as the results.All the development, integration, graphical interface and tests were performed by the author of this thesis. The developed platform was tested and is considered ready to production.

Informados con la literatura mundial acerca de que existe un bajo porcentaje de metástasis a ganglios linfáticos, cuando se logra identificar en el tracto gastrointestinal (TGI) la presencia de cáncer temprano, en el año 2001 se inició el estudio del cáncer temprano gastrointestinal y su tratamiento con técnica de mucosectomía con el objetivo de evaluar la utilidad de la misma, la cual ha demostrado tener buenos resultados en lesiones menores de 2cms, sin embargo existe una interrogante en cuanto a la utilidad de la misma en las lesiones rectocolónicas de gran tamaño (LCGT: ≥ 2.5cms) motivo por el cual, de las mucosectomías realizadas en el TGI a partir del esa fecha, se incluyeron en el presente estudio 50 pacientes (de 90 con lesiones rectocolónicas) que presentaron LCGT quienes cumplieron con los criterios de inclusión y exclusión, con el objetivo principal de evaluar su utilidad e identificar las recidivas así como su efectivo tratamiento durante el seguimiento. Los resultados se presentan mediante cuadros y gráficas, los cuales se discuten y finalmente se plantean las respectivas conclusiones derivadas de este estudio.
Estudio realizado en el período comprendido entre el 1 de noviembre de 2001 y el 30 de noviembre de 2007 en el Servicio de Gastroenterología del HIGA San Martín de La Plata, Cátedra de Gastroenterología de Postgrado FCM-UNLP, WGO Advanced Training Center de la Organización Mundial de Gastroenterología