La Tomographie par Emission de Positons (TEP) au Fluoro-déoxyglucose marqué au Fluor 18 (¹⁸F-FDG), analogue du glucose, permet d'obtenir une image de la consommation de glucose dans l'organisme. La plupart des foyers tumoraux présentant une consommation excessive de glucose, son utilisation en oncologie permet d'améliorer la prise en charge des patients en diminuant le temps nécessaire pour évaluer l'efficacité des traitements tels que la chimiothérapie et la radiothérapie. Mon projet de recherche visait à proposer et améliorer des méthodes de quantification en TEP au ¹⁸F-FDG afin de caractériser au mieux l'évolution métabolique des volumes tumoraux.De nombreux facteurs biaisent la quantification en TEP. Parmi eux, l'Effet de Volume Partiel (EVP) reste difficile à corriger, notamment à cause de la faible résolution spatiale des images TEP. Afin de déterminer l'impact de la correction de l’EVP sur l’évaluation des réponses des tumeurs, une étude sur données simulées par Monte Carlo a tout d’abord été effectuée. Cette étude a été complétée par l’analyse de données TEP/TDM (Tomodensitométrie) acquises chez 40 patients atteints de cancers colorectaux métastatiques (CCM), traités par chimiothérapie à l'Institut Jules Bordet (Bruxelles). L’analyse de 101 tumeurs a montré que les critères tels que le SUV, n’incluant pas de correction de l'EVP, et qui reflètent alors le volume tumoral et son activité, prédisaient mieux l’évolution tumorale que les critères corrigés de l’EVP. Compte tenus des résultats prometteurs récents de méthodes de caractérisation de l’hétérogénéité de la fixation du FDG dans les tumeurs, un second volet de notre travail a consisté à étudier l’intérêt de la prise en compte de la texture dans le cadre du suivi thérapeutique. L’application de l’analyse de texture aux cas de CCM étudiés précédemment n’a pas permis de démontrer une valeur ajoutée des indices de texture par rapport aux index quantitatifs couramment employés en clinique. Nous avons montré que cette conclusion s’expliquait en partie par la non-robustesse des indices de texture vis-à-vis des paramètres impliqués dans leur mesure. Nous avons enfin cherché à évaluer une méthode d’Analyse Factorielle de Séquences d’Images Médicales (AFSIM), appliquée au contexte du suivi thérapeutique, pour caractériser l’évolution tumorale tout au long du traitement. Cette étude a porté sur 9 séries de 4 à 6 examens TEP/TDM de patients traités par radiothérapie au Centre Hospitalier Universitaire Henri Becquerel de Rouen. Outre l’information visuelle globale apportée par cette méthode, l’analyse quantitative des résultats obtenus a permis de caractériser l’hétérogénéité de la réponse vue par l’AFSIM. L’échec des index classiques, provenant entre autres de leur incapacité à distinguer les processus inflammatoires de l’activité métabolique tumorale, a permis de monter la valeur ajoutée de l’AFSIM par rapport aux index tels que le SUV maximal ou moyen. / Positron Emission Tomography (PET) using ¹⁸F-FluoroDeoxyGlucose (¹⁸F-FDG), a radiolabelled analogue of the glucose, is used to get an image of the glucose consumption in the body. As most tumor masses show a high glucose consumption, PET is widely used in oncology for diagnosis and patient monitoring. In the context of patient monitoring, the motivation is to decrease the time interval needed to assess treatment (radiotherapy or chemotherapy efficieny) compared to therapeutic follow-up based only on anatomic imaging only (Computed Tomography or Magnetic Resonance Imaging). My research project aimed at proposing and improving quantitative methods in FDG-PET to better characterize tumor evolution.In PET, many factors affect the accuracy of parameters estimated from the images. Among them, Partial Volume Effect (PVE) remains difficult to correct, mainly due to the low spatial resolution of PET images. To determine the impact of PVE on treatment response evaluation, a preliminary study was performed using Monte Carlo simulated PET scans. An additional study was conducted based on the analysis of the PET/CT (Computed Tomography) data of 40 Metastatic Colorectal Cancer (MCC) patients treated with chemotherapy at the Jules Bordet Institute (Brussels, Belgium). The analysis of the 101 tumors showed that criteria such as the Standardized Uptake Value (SUV), which does not include PVE correction, were better predictors of tumors evolutions than PVE corrected criteria. This is because without PVE correction, SUV includes information on both metabolic volume and metabolic activity, which are two relevant pieces of information to characterize the tumor. A second part of our work was to study the potential of tumor texture analysis in patient monitoring, following promising results reported in the literature. Texture analysis was applied to the MMC patients data previously mentioned but did allow to a better segregation of tumors responses as compared to indices currently used in the clinics. We found that this was partly due to the lack of robustness of the textures indices.Finally, we evaluated a Factor Analysis in Medical Images Series (FAMIS) method to characterize tumor evolution during treatment. This study focused on 9 series of 4 to 6 PET/CT scans acquired all along the radiotherapy/radio-chemotherapy of patients treated at the Centre Hospitalier Universitaire Henri Becquerel (Rouen, France). In addition to the rich visual information brought by this method, a quantitative analysis of the results made it possible to characterize response heterogeneity as seen using FAMIS. In particular, FAMIS clearly demonstrated the occurrence of inflammatory processes. In addition, due to the low metabolic activity of the tumors at the end of the treatment, many conventional indices could not describe the tumor changes, while FAMIS gave a full assessment of the tumor change over time.

Background: Hibernomas are rare benign soft-tissue tumours arising from brown fat tissue. Although imaging characteristics are not specific certain imaging features, common locations and patient demographics may suggest hibernoma as a differential diagnosis. Case presentation: We report on two 48-year-old male patients with hibernoma. The tumour presented with local swelling of the inguinal region in the first patient and was an incidental imaging finding in the second patient. Imaging included magnetic resonance imaging in both patients and computed tomography as well as 18 F-fluorodeoxyglucose positron emission tomography-computed tomography in the second patient. In both cases histological diagnosis was initially based on excisional and needle core biopsy, respectively. Complete surgical resection confirmed the diagnosis of hibernoma thereafter. Conclusion: In soft tissue tumours with fatty components hibernoma may be included into the differential diagnosis. Because of the risk of sampling errors in hibernoma-like tissue components of myxoid and well-differentiated liposarcoma, complete resection is mandatory. This article also reviews the current imaging literature of hibernomas.

Introdução: O câncer do esôfago é uma das neoplasias do aparelho digestivo com maior gravidade e que apresenta grande morbimortalidade, mesmo quando o diagnóstico é precoce. A maioria dos pacientes é diagnosticado nos estágios avançados. O tratamento depende do estadiamento da neoplasia que avalia a profundidade de invasão do tumor (T), a disseminação linfonodal (N) e a presença de metástases a distância (M) e segue as orientações da União Internacional Contra o Câncer (UICC). Nas últimas décadas o estadiamento era realizado convencionalmente pela tomografia computadorizada (TC) e atualmente com a utilização de equipamentos que avaliam o metabolismo glicolítico do tumor como o 18F-FDG-PET e o PET-CT. Este estudo teve como Objetivo avaliar a relação entre a tomografia computadorizada e os métodos metabólicos como o 18F-FDG-PET e PET-CT, no estadiamento e tratamento do Adenocarcinoma e do Carcinoma Espinocelular (CEC) do esôfago. Método: Foram avaliados 331 pacientes com diagnóstico de Adenocarcinoma e CEC do esôfago entre 2008 e 2014. Destes, 55 pacientes (16,6%) apresentaram Adenocarcinoma e 276 (83,4%) apresentaram CEC. A idade variou de 38 a 92 anos, com média de 62,9 (+/- 9,8) anos. Inicialmente foram submetidos ao estadiamento com TC e proposta de conduta cirúrgica curativa ou tratamento paliativo. Posteriormente foram avaliados com a inclusão do 18F-FDG-PET ou do PET-CT e foi definida a conduta final. Resultados: A proporção de linfonodos positivos (N+) na tomografia foi de 71%, enquanto que nos métodos metabólicos foi de 70,1% (p=0,834), contudo, com pequena concordância (Kappa=0,339). A proporção de metástases (M1) encontradas na TC foi de 44,1% e no PET-CT 47,1%. Para metástases, o teste Kappa mostrou que os dois métodos apresentam uma concordância regular (0,452), apresentando mudanças de estadiamento em 36,5% dos indivíduos, sendo 19,3% com sobre estadiamento e 17,2% com subestadiamento. Entretanto, apenas 63 pacientes (19%) apresentaram mudança de conduta final e esta foi maior nos pacientes com sobre estadiamento (67,2%) (p < 0,005). Nos pacientes com Adenocarcinoma, observou-se um número maior de subestadiamento (32,7%), comparado àqueles com CEC (15,4%) (p < 0,0001), entretanto, sem apresentar diferença estatisticamente significativa quando avaliada a mudança de conduta. Avaliando individualmente os 140 pacientes estadiados com 18F-FDG-PET, 52,9% apresentaram linfonodos positivos (N+), valor semelhante à tomografia (p=0,053), entretanto com concordância pequena, cerca de 32,9% destes com metástases (M1) (p=0,749) e com concordância regular entre os métodos. Após o estadiamento, ocorreu uma mudança de conduta de 23,6% quando avaliado por equipe multidisciplinar. Com o uso do PET-CT, a proporção de tumores T4 foi de 27,2% (p=0,071), porém, com concordância boa com a tomografia (Kappa=0,616). A proporção de linfonodos positivos (N+) foi de 82,7%, com pequena concordância com a tomografia (Kappa=0,392). A proporção de metástases (M1) no PET-CT foi de 57,6%, com concordância regular (Kappa=0,465). Apresentaram mudança de estadiamento 34% dos indivíduos, sendo 19,3% com sobre estadiamento e 14,7% com subestadiamento. Entretanto, dos 191 pacientes, apenas 30 (15,7%) apresentaram mudança de conduta final, sendo que 67,6% ocorreu nos casos com sobre estadiamento, quando comparada aos casos com subestadiamento (17,9%) (p < 0,005). Pacientes com Adenocarcinoma apresentaram um número maior de subestadiamento (30%), comparado àqueles com CEC (11,8%), (p < 0,0001), entretanto, sem apresentar diferença estatisticamente significativa. A sobrevida global, quando avaliados com PET-CT, foi em torno de 30% após 30 meses, sendo semelhante tanto no grupo de pacientes em que houve mudança de conduta quanto no grupo em que esta mudança de conduta não ocorreu. Conclusão: Conclui-se, portanto, que no estadiamento tomográfico com 18F-FDG-PET e com PET-CT foi identificado um número expressivo de pacientes em estágios avançados, entretanto estes achados muitas vezes diferem entre si. A mudança de conduta ocorre em número expressivo de pacientes e geralmente nos casos em que ocorre sobre estadiamento. Embora o Adenocarcinoma apresente um número maior de casos de subestadiamento que o CEC, esta mudança de estadiamento não se reflete na mudança de conduta quando comparados. Torna-se importante, portanto, a avaliação multiprofissional em serviço de excelência no momento de decisão sobre a melhor terapêutica. Por fim, observamos a mesma curva de sobrevida entre aqueles pacientes em que há certeza da conduta a ser tomada e aqueles em que a conduta foi mudada após o uso do PET-CT, o que corrobora a necessidade da utilização em conjunto desses dois métodos / Introduction: Esophagus cancer is one of the most serious neoplasms of the digestive tract that presents great morbidity and mortality even in early diagnosis. Most patients are diagnosed in advanced stages. Treatment depends on tumor staging, which evaluates the depth of tumor invasion (T), lymph node spread (N) and the presence of distant metastases (M) and follow the guidelines of Union for International Cancer Control (UICC). In the last decades, staging was performed conventionally by computed tomography (CT) and currently with the use of equipments that evaluates tumor glycolytic metabolism such as 18F-FDG-PET and PET-CT. This study has as main Objective to evaluate the relationship between computed tomography and metabolic methods such as 18F-FDG-PET and PET-CT in the staging and treatment of Adenocarcinoma and Spinocellular Carcinoma (SCC) of the esophagus. Method: A total of 331 patients diagnosed with adenocarcinoma and esophageal SCC were evaluated between 2008 and 2014. 55 of these patients (16.6%) had adenocarcinoma and 276 (83.4%) had CPB, ranging from 38 to 92 years, mean age of 62.9 (+/- 9.8) years. Initially they underwent staging with CT and it was proposed a curative surgical management or palliative treatment. Lately they were evaluated with the inclusion of 18F-FDG-PET or PET-CT and then the final management was defined. Results: The proportion of positive lymph nodes (N +) on the CT scan was 71%, whereas in the metabolic methods it was 70.1% (p=0.834), however, with a fair agreement (Kappa=0.339). The proportion of metastases (M1) found in CT was 44.1% and in PET-CT, 47.1%. For metastases, the Kappa test showed that the two methods presented a moderate agreement (0.452), presenting staging changes in 36.5% of subjects, being 19.3% with upstaging and 17.2% with downstaging. However, only 63 patients (19%) showed changes in the final management and this was higher in upstaging patients (67,2%) (p < 0,005). In patients with Adenocarcinoma, a greater number of downstaging was observed (32.7%), compared to those with CPB (15.4%) (p < 0.0001), however, without any statistically significant difference when the change of management was evaluated. Evaluating individually the 140 patients staged with 18F-FDG-PET, 52.9% presented positive lymph nodes (N +), data similar to tomography (p = 0.053), however with fair agreement, about 32.9% of these had metastases (M1) (P=0.749) and with moderate agreement between the methods. After the staging, a conduct change of 23.6% occurred when evaluated by a multidisciplinary team. With the use of PET-CT, the proportion of T4 tumors was 27.2% (p = 0.071), but with good agreement with tomography (Kappa=0.616). The proportion of positive lymph nodes (N+) was 82.7%, with fair agreement with the tomography (Kappa=0.392). The proportion of metastases (M1) in PET-CT was 57.6%, with moderate agreement (Kappa=0.465). 34% of the individuals presented staging change, 19.3% with upstaging and 14.7% with downstaging. However, only 30 (15.7%) out of 191 presented a final change of behavior, 67.6% of which occurred in cases with upstaging when compared to cases with downstaging (17.9%) (p < 0.005). Patients with adenocarcinoma had a greater number of downstaging (30%) compared to those with CPB (11.8%), (p < 0.0001), however, with no statistically significant difference. Overall survival when staged with PET-CT was around 30% after 30 months, being similar both in the group of patients where there was change of management and in the group where this change of management did not occur. Conclusion: It was concluded that in the tomographic staging with 18F-FDG-PET and with PET-CT an expressive number of patients in advanced stages was identified, however these findings often differ from each other. The change in management occurs in an expressive number of patients, and usually in cases where upstaging occurs. Although Adenocarcinoma presents a greater number of cases of downstaging than CPB, this change in staging is not reflected in the change of management when both are compared. It is important, therefore, the multiprofessional evaluation in service of excellence when deciding on the best therapeutics. Finally, we observed the same survival curve between those patients in which there is certainty of the management to be taken and those in which the management was changed after the use of PET-CT, which corroborates with the need to use these two methods together

En oncologie ORL, l'imagerie multiparamétrique est utilisée par un nombre grandissant d'équipes. Parmi les bio-marqueurs, la captation normalisée du fluoro-désoxyglucose (SUV-FDG) en tomoscintigraphie par émission de positons (TEP) et la restriction de la diffusion en IRM (DWI-MRI) sont les plus utilisées.L'IRM couplée à la TEP (TEP/IRM) est une nouveauté qui permet une diminution très significative des doses d'irradiation délivrées par rapport à la TEP/TDM. Nous adressons notre première expérience concernant l'aspect en diffusion et en TEP/IRM dans la surveillance des patients après radio-chimiothérapie. A la question de l'interchangeabilité du FDG-PET et de la DWI-MRI, nous avons tenté d'identifier un lien en imagerie entre la cellularité tumorale et sa consommation glucidique. La cellularité tumorale est approchée en IRM par la mesure du coefficient apparent de diffusion (ADC) et son métabolisme glucidique est approché en TEP en utilisant le 18F-desoxyglucose (FDG) par la mesure de la valeur de fixation normalisée (SUV). Dans une série appariée de 33 patients, nous avons analysé la reproductibilité des mesures de l'ADC et de SUV. Puis nous avons évalué l'indépendance statistique de ces biomarqueurs. Nous avons ensuite voulu comparer les résultats de la TEP obtenus à partir de la TEP/TDM et de la TEP/IRM. Dans une série prospective appariée chez 32 patients explorés en FDG-TEP, nous avons évalué qualitativement les images obtenues par la fusion des images recalées en TEP/IRM et TEP/TDM. Nous avons ensuite comparé la pertinence clinique des deux techniques. Et enfin nous avons comparé les valeurs quantitatives de SUV obtenues du tissu sain et du tissu pathologique. / Multiparametric imaging interest and clinical use is rising for head and neck carcinoma (HNC). Among these modalities, FDG in PET and DWI-MRI are the most studied. PET/MRI is a new modality that allows in a single examination of combined various biologic biomarkers.After an optimization process of PET/MRI, we applied our first experience concerning the aspects of DWI-MRI and PET-MRI after radiation therapy. Thereafter we studied the correlation of SUV and ADC in HNC. In this study SUV and ADC values were independent parameters in HNSCC. Measurements of these two biomarkers were reproducible with almost perfect observer agreements for both methods. Neither SUV nor ADC values were able to predict the histologic grade, although a trend towards higher SUV and lower ADC values was observed in poorly differentiated tumours. Secondly, we we studied detection and quantification of focal uptake in head and neck tumours: 18F-FDG PET/MRI versus PET/CT in 32 consecutive HNSCC who underwent 18F-FDG PET/MRI and PET/CT. Attenuation correction sequence for PET/MRI and CT for PET/CT were used to caculate SUV. In results, PET/MRI coregistration and image fusion was feasible in all patients. There was no statistically significant difference between PET/MRI and PET/CT regarding rating scores for image quality, fusion quality, lesion conspicuity or anatomic location, number of detected lesions and number of patients with and without malignant lesions. A high correlation was observed for SUV measured on PET/MRI and PET/CT. SUV measured on PET/MRI were significantly lower than on PET/CT for malignant tumours, metastatic neck nodes, benign lesions, bone marrow, and liver (p <0.05).

Background: Hibernomas are rare benign soft-tissue tumours arising from brown fat tissue. Although imaging characteristics are not specific certain imaging features, common locations and patient demographics may suggest hibernoma as a differential diagnosis. Case presentation: We report on two 48-year-old male patients with hibernoma. The tumour presented with local swelling of the inguinal region in the first patient and was an incidental imaging finding in the second patient. Imaging included magnetic resonance imaging in both patients and computed tomography as well as 18 F-fluorodeoxyglucose positron emission tomography-computed tomography in the second patient. In both cases histological diagnosis was initially based on excisional and needle core biopsy, respectively. Complete surgical resection confirmed the diagnosis of hibernoma thereafter. Conclusion: In soft tissue tumours with fatty components hibernoma may be included into the differential diagnosis. Because of the risk of sampling errors in hibernoma-like tissue components of myxoid and well-differentiated liposarcoma, complete resection is mandatory. This article also reviews the current imaging literature of hibernomas.

Dieser Beitrag ist mit Zustimmung des Rechteinhabers aufgrund einer (DFG-geförderten) Allianz- bzw. Nationallizenz frei zugänglich.

Nuklearmedicin är en specialitet där Flour18-fluorodeoxyglukos positronemissionstomografi/datortomografi (18F-FDG PET/DT) är en vanligt förekommande undersökning. Den vanligaste indikationen för att genomgå undersökningen är misstänkt eller känd cancer. Syfte: Syftet var att studera patienters upplevelse av information i samband med en 18F-FDG PET/DT-undersökning. Metod: En kvantitativ tvärsnittsstudie genomfördes på nuklearmedicinska avdelningen där 100 patienter som uppfyllde inklusionskriterierna tillfrågades om medverkan i studien. Efter undersökningen ombads studiedeltagarna att besvara en enkät innehållande nio studiespecifika frågor gällande informationen med fem tillhörande svarsalternativ. Resultat: Studiens resultat visar att 93,4% av studiedeltagarna upplevde den skriftliga informationen som tydlig och lättförståelig. Samtliga enkätfrågor erhöll en median på 4 vilket överensstämmer med svarsalternativet 5 “Instämmer helt”. Resultatet visar att 86 studiedeltagare (94,5%) var nöjda med den muntliga informationen samt att 78 studiedeltagare (88,7%) var nöjda med den skriftliga informationen. Resultatet påvisade att informationen gällande den radioaktiva isotopen erhöll 7 "Instämmer i låg grad" och informationen efter undersökningen erhöll 3 "Instämmer inte alls". Slutsats: Majoriteten av patienter som genomgår en 18F-FDG PET/DT-undersökning är nöjda med informationsinnehållet i samband med undersökningen. Informationen gällande den radioaktiva isotopen samt informationen efter undersökningen behöver dock förbättras. En mer omfattande studie som inkluderar fler studiedeltagare under en längre tidsperiod bör genomföras för ett mer generaliserbart resultat hos den aktuella patientgruppen. / Nuclear Medicine is a speciality in which Fluorine-18-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) is a common examination. The most common indication for undergoing the examination is known or suspected cancer. Aim: Our aim was to study patients’ experiences regarding the information during an 18F-FDG PET/CT examination. Method: A quantitative cross-sectional study was carried out at the Nuclear Medicine department in which 100 patients who met the inclusion criteria were asked to participate. The participants were asked to answer a questionnaire regarding the information containing of nine study-specific questions with five options. Results: The results of our study showed that 93,4% of the participants perceived the written information as direct and easily comprehensible. All questions received a median score of 4, which corresponds with alternative 5 “Strongly agree”. The result shows that 86 participants (94,5%) were satisfied with the verbal information and that 78 participants (88,7%) were satisfied with the written information. The results showed that the information regarding the radioactive isotope received 7 “Disagree” and the information after the examination received 3 “Strongly disagree”. Conclusion: The majority of patients who undergo an 18F-FDG PET/CT examination are satisfied with the information during the examination. However, the information regarding the radioactive tracer and the information after the examination needs to be improved. A larger study that includes more participants over a longer period of time is suggested to be conducted for more generalized results.

La radiothérapie est une des méthodes principales utilisée dans le traitement thérapeutique des tumeurs malignes. Pour améliorer son efficacité, deux problèmes essentiels doivent être soigneusement traités : la prédication fiable des résultats thérapeutiques et la segmentation précise des volumes tumoraux. La tomographie d’émission de positrons au traceur Fluoro- 18-déoxy-glucose (FDG-TEP) peut fournir de manière non invasive des informations significatives sur les activités fonctionnelles des cellules tumorales. Les objectifs de cette thèse sont de proposer: 1) des systèmes fiables pour prédire les résultats du traitement contre le cancer en utilisant principalement des caractéristiques extraites des images FDG-TEP; 2) des algorithmes automatiques pour la segmentation de tumeurs de manière précise en TEP et TEP-TDM. La théorie des fonctions de croyance est choisie dans notre étude pour modéliser et raisonner des connaissances incertaines et imprécises pour des images TEP qui sont bruitées et floues. Dans le cadre des fonctions de croyance, nous proposons une méthode de sélection de caractéristiques de manière parcimonieuse et une méthode d’apprentissage de métriques permettant de rendre les classes bien séparées dans l’espace caractéristique afin d’améliorer la précision de classification du classificateur EK-NN. Basées sur ces deux études théoriques, un système robuste de prédiction est proposé, dans lequel le problème d’apprentissage pour des données de petite taille et déséquilibrées est traité de manière efficace. Pour segmenter automatiquement les tumeurs en TEP, une méthode 3-D non supervisée basée sur le regroupement évidentiel (evidential clustering) et l’information spatiale est proposée. Cette méthode de segmentation mono-modalité est ensuite étendue à la co-segmentation dans des images TEP-TDM, en considérant que ces deux modalités distinctes contiennent des informations complémentaires pour améliorer la précision. Toutes les méthodes proposées ont été testées sur des données cliniques, montrant leurs meilleures performances par rapport aux méthodes de l’état de l’art. / Radiation therapy is one of the most principal options used in the treatment of malignant tumors. To enhance its effectiveness, two critical issues should be carefully dealt with, i.e., reliably predicting therapy outcomes to adapt undergoing treatment planning for individual patients, and accurately segmenting tumor volumes to maximize radiation delivery in tumor tissues while minimize side effects in adjacent organs at risk. Positron emission tomography with radioactive tracer fluorine-18 fluorodeoxyglucose (FDG-PET) can noninvasively provide significant information of the functional activities of tumor cells. In this thesis, the goal of our study consists of two parts: 1) to propose reliable therapy outcome prediction system using primarily features extracted from FDG-PET images; 2) to propose automatic and accurate algorithms for tumor segmentation in PET and PET-CT images. The theory of belief functions is adopted in our study to model and reason with uncertain and imprecise knowledge quantified from noisy and blurring PET images. In the framework of belief functions, a sparse feature selection method and a low-rank metric learning method are proposed to improve the classification accuracy of the evidential K-nearest neighbor classifier learnt by high-dimensional data that contain unreliable features. Based on the above two theoretical studies, a robust prediction system is then proposed, in which the small-sized and imbalanced nature of clinical data is effectively tackled. To automatically delineate tumors in PET images, an unsupervised 3-D segmentation based on evidential clustering using the theory of belief functions and spatial information is proposed. This mono-modality segmentation method is then extended to co-segment tumor in PET-CT images, considering that these two distinct modalities contain complementary information to further improve the accuracy. All proposed methods have been performed on clinical data, giving better results comparing to the state of the art ones.

Cerebral blood flow quantification is a vital diagnostic tool for disease monitoring and used for diagnosing a variation of pathological conditions. The human brain requires roughly about 20 % of the total cardiac output to sustain normal functioning, hence the perfusion of blood is an important factor to deliver oxygenated blood. The golden standard for quantifying the cerebral blood flow follows by measurement with dynamic positron emission tomography of 15O-labelled water modelled by tracer kinetic compartments. For implementation, knowledge of an input function must exist which is in general being sampled through arterial cannulation of the radial artery with a continuous sampling instrument. The core of this thesis is to establish if two sampling instruments contradicts in comparison to each other when sampling the data to the input function.  In total 22 subjects underwent a 10-minute dynamic  15O-labeled water brain PET scan on two imaging modalities PET/CT and PET/MR. Continuous arterial sampling was performed either by a Veenstra on PET/CT or a Swisstrace on PET/MR during a baseline scan. In two subjects the two sampling instruments were coupled in series and imaged solely on the PET/CT.  Cerebral blood flow analysis was done comparing varying dispersion times, the two imaging modalities compared each other and comparing the calculated and measured blood flows obtained through this study with the values obtained prior. To be able to compare the values showing inconsistency to the values obtained through this thesis, a comparison between two different iterative reconstruction methods was done. Here the method of ordered subsets expectation maximum was compared to a Bayesian penalized-likelihood method. To further compare the two sampling instruments an image derived input function was constructed and compared with the blood sampled input function. The results showed that there was no significant difference between measured cerebral blood flow between the two imaging modalities with the currently used reconstruction method based on Bayesian penalized likelihood but presented in the earlier data there was an inconsistency. A dispersion analysis with variation on the external dispersion time shows that if the time was chosen to low or to high compared to the standard time used it introduced distorted fitted models of the activity curves. This distortion creates further errors in the calculation of the cerebral blood flow, however with the analysis the standard dispersion time could be confirmed as an accurate fit. Subjects imaged with the two sampling instruments in series showed no significant difference except for the measured values on Veenstra to be slightly higher. Lastly the correlation between the image derived input function and the blood sampled input function showed poorly performance. Only a R2 value of 0.42 was achieved on the PET/CT while a meagre R2 value of 0.18 was achieved on the PET/MR. Although the correlation was poorly, the plotted activity curves from the two functions showed a representable appearance between each other. / Kvantifiering av det cerebrala blodflödet är ett nödvändigt diagnostiskt verktyg som används för att kontrollera och diagnostisera en variation av patologiska sjukdomstillstånd. Den mänskliga hjärnan kräver kring 20 %av den totala produktionen från hjärtat för att upprätthålla normal funktion, följaktligen är perfusion av blod en viktig faktor för att distribuera syrerikt blod runt om i kroppen. Den gyllene standarden för kvantifiering av det cerebrala blodflödet följer som undersökning med dynamisk positron emission tomografi av 15O-märkt vatten, modellerat  med kinetisk kompartment teori. För att kunna implementera detta måste information om en input-funktion erhållas, generellt erhålls detta genom att blod tags genom arteriell kanylering av antingen den radiella artären med ett kontinuerligt samplings instrument. Målet med detta arbete är att fastställa om två samplings instrument motsäger varandra vid mätning av data till input-funktionen. Totalt deltagande är 22 patienter som genomgick en 10-minuters dynamisk 15O-märkt vatten PET undersökning av hjärnan på två bildtagningsmodaliteter PET/CT och PET/MR. Kontinuerlig blodtagning genomfördes antingen med en Veenstra sampler instrument på PET/CT eller en Swisstrace sampler instrument på PET/MR tillsammans med en baseline undersökning. Vid två undersökningar seriekopplades de två instrumenten och patienterna blev endast undersökta vid PET/CT. För ytterligare kunna utvärdera de två instrumenten, konstruerades en bild framtagen input-funktion som sedan kunde jämföras med den blod samplade input-funktionen. Cerebrala blodflödes analyser gjordes med olika dispersions tider, även för att kunna jämföra de två bildtagningsmodaliteterna mot varandra och jämföra erhållna värden framtagna under denna studie med en tidigare studie. För att kunna jämföra avvikelserna i de uppmätta värdena har även två olika rekonstruerings metoder studerats. Resultaten visar ingen signifikant skillnad mellan de uppmätta cerebrala blodflödena mellan de två bildtagningsmodaliteterna rekonstruerade med den nuvarande standarden. Dispersions analysen med varierande extern dispersions tid visar att om tiden är för kort eller för lång jämfört med standardtiden, introduceras en osann anpassning av aktivitets kurvorna. Denna förvrängning av datat resulterar till fler avvikelser i beräkningarna av blodflödet, likväl var det möjligt att bekräfta standardtiderna som används. Patienter som undersöktes med instrumenten i seriekoppling visade ingen signifikant skillnad förutom att det uppmättes en aningens högre värden hos patienter med Veenstra som blod sampler. Slutligen, korrelationen mellan den bild framtagna input-funktionen och den blod samplade input-funktionen visade ett dåligt resultat. Endast ett R2 värde på 0.42 erhölls för PET/CT medan endast ett R2 värde på 0.18 på PET/MR erhölls. Trotts att korrelationen var dålig, visade de plottade aktivitets kurvorna ett representativt utseende mellan de två typerna av input funktion.