En médecine, les analyses de population à grande échelle ont pour but d’obtenir des informations statistiques pour mieux comprendre des maladies, identifier leurs facteurs de risque, développer des traitements préventifs et curatifs et améliorer la qualité de vie des patients.Dans cette thèse, nous présentons d’abord le contexte médical de la maladie d’Alzheimer, rappelons certains concepts d’apprentissage statistique et difficultés rencontrées lors de l’application en imagerie médicale. Dans la deuxième partie,nous nous intéressons aux analyses transversales, c-a-d ayant un seul point temporel.Nous présentons une méthode efficace basée sur les séparateurs à vaste marge (SVM)permettant de classifier des lésions dans la matière blanche. Ensuite, nous étudions les techniques d’apprentissage de variétés pour l’analyse de formes et d’images, et présentons deux extensions des Laplacian eigenmaps améliorant la représentation de patients en faible dimension grâce à la combinaison de données d’imagerie et cliniques. Dans la troisième partie, nous nous intéressons aux analyses longitudinales, c-a-d entre plusieurs points temporels. Nous quantifions les déformations des hippocampus de patients via le modèle des larges déformations par difféomorphismes pour classifier les évolutions de la maladie. Nous introduisons de nouvelles stratégies et des régularisations spatiales pour la classification et l’identification de marqueurs biologiques. / In medicine, large scale population analysis aim to obtain statistical information in order to understand better diseases, identify their risk factors, develop preventive and curative treatments and improve the quality of life of the patients.In this thesis, we first introduce the medical context of Alzheimer’s disease, recall some concepts of statistical learning and the challenges that typically occurwhen applied in medical imaging. The second part focus on cross-sectional studies,i.e. at a single time point. We present an efficient method to classify white matter lesions based on support vector machines. Then we discuss the use of manifoldlearning techniques for image and shape analysis. Finally, we present extensions ofLaplacian eigenmaps to improve the low-dimension representations of patients usingthe combination of imaging and clinical data. The third part focus on longitudinalstudies, i.e. between several time points. We quantify the hippocampus deformations of patients via the large deformation diffeomorphic metric mapping frameworkto build disease progression classifiers. We introduce novel strategies and spatialregularizations for the classification and identification of biomarkers.

The study investigated the application of the crime threat analysis process at station level within the Nelson Mandela Metro City area with the objective of determining inhibiting factors (constraints) and best practices. Qualitative research methodology was applied and interviews were conducted with crime analysts and specialised investigators/intelligence analysts. The research design can be best described as descriptive and explorative in nature. The crime threat analysis process embroils the application of various crime analysis techniques and the outcomes thereof intends to have a dual purpose of generating operational crime management information in assisting crime prevention initiatives and crime detection efforts, mainly focussing on the criminal activities of group offenders (organised crime related), repeat offenders and serial offenders. During the study it became evident that crime analysts understand and thus apply the crime threat analysis process indifferently, which impeded on the relevancy and the utilisation thereof as an effective crime management tool. / Criminology / M.Tech. (Policing)

The study investigated the application of the crime threat analysis process at station level within the Nelson Mandela Metro City area with the objective of determining inhibiting factors (constraints) and best practices. Qualitative research methodology was applied and interviews were conducted with crime analysts and specialised investigators/intelligence analysts. The research design can be best described as descriptive and explorative in nature. The crime threat analysis process embroils the application of various crime analysis techniques and the outcomes thereof intends to have a dual purpose of generating operational crime management information in assisting crime prevention initiatives and crime detection efforts, mainly focussing on the criminal activities of group offenders (organised crime related), repeat offenders and serial offenders. During the study it became evident that crime analysts understand and thus apply the crime threat analysis process indifferently, which impeded on the relevancy and the utilisation thereof as an effective crime management tool. / Criminology and Security Science / M.Tech. (Policing)