Attention deficit hyperactivity disorder (ADHD) is a neurobehavior developmental disorder that affects around 7.5% of Taiwan children. With the use of magnetic resonance imaging , many results have been reported that ADHD patients have volume atrophy in gray matter and dysfunction in couples of cortical regions. In recent years, diffusion MR imaging with diffusion-sensitizing gradients has been used to investigate the abnormality of neural fibers in disease involved with central nervous system. In this study, the anisotropy of white matter in both ADHD patients and age-matched healthy subjects was estimated using diffusion tensor imaging to undergo inter-subject comparison. In this work, a significant decrease (FWE-corrected p-value <0.05) of FA values has been found in white matter of adolescents diagnosed as ADHD patients, compared with normal controls group. The areas that confirmed by two different algorithms of inter-subject comparison are mainly diffused on white matter region, including middle cerebellar peduncle, left inferior longitudinal fasciculus, internal capsule, left optic radiation, external capsule, splenium of the corpus callosum, superior longitudinal fasciculus, superior frontal and parietal-occipital nerve fibers.

No description available.

Therapeutic response assessment is a key component in adaptive image-guided radiotherapy. Conventional anatomic measures of response offer little information about the spatial distribution of tumor change. Recently developed voxel-wise response assessment methods operating on functional and biological imaging are better capable of evaluating the heterogeneity of response within the tumor, and thus may yield greater sensitivity than conventional approaches. However, voxel-wise analyses are limited by local registration uncertainties inherent to longitudinal imaging of tumors with changing morphology. A multi-resolution local histogram (LH) moment-based measure of voxel similarity was developed for the purpose of assessing the strength of correspondence between voxels of serial tumor images. This measure was first benchmarked through a series of experiments designed to establish robustness to image intensity variation and sensitivity to alterations in tissue structure through application of simulated deformations. The LH similarity method was subsequently developed as a means of mapping the spatial extent of structural change in tumors through the incorporation of an estimate of image complexity. The change maps were applied to a voxel-wise analysis of diffusion-weighted magnetic resonance imaging of patients with glioblastoma, acquired pre- and post-chemoradiotherapy. The sensitivity of the voxel-wise analysis in differentiating responding/stable patients from non-responding/progressing patients was improved by stratifying the analysis voxels according to regions of interest (ROI) based on the LH similarity-based estimate of tumor change. Meaningful correspondence relationships between evaluated voxels are essential for accurate image-based quantification of tumor structure and function with voxel-wise analysis techniques. The LH similarity methods developed here can robustly evaluate the quality of spatial and temporal voxel correspondence relationships and provide an automated tool for ROI selection and voxel change stratification. It is readily extendable to the analysis of the wide array of anatomic, functional and biological imaging currently used to characterize tumors, guide therapy and assess response.

Therapeutic response assessment is a key component in adaptive image-guided radiotherapy. Conventional anatomic measures of response offer little information about the spatial distribution of tumor change. Recently developed voxel-wise response assessment methods operating on functional and biological imaging are better capable of evaluating the heterogeneity of response within the tumor, and thus may yield greater sensitivity than conventional approaches. However, voxel-wise analyses are limited by local registration uncertainties inherent to longitudinal imaging of tumors with changing morphology. A multi-resolution local histogram (LH) moment-based measure of voxel similarity was developed for the purpose of assessing the strength of correspondence between voxels of serial tumor images. This measure was first benchmarked through a series of experiments designed to establish robustness to image intensity variation and sensitivity to alterations in tissue structure through application of simulated deformations. The LH similarity method was subsequently developed as a means of mapping the spatial extent of structural change in tumors through the incorporation of an estimate of image complexity. The change maps were applied to a voxel-wise analysis of diffusion-weighted magnetic resonance imaging of patients with glioblastoma, acquired pre- and post-chemoradiotherapy. The sensitivity of the voxel-wise analysis in differentiating responding/stable patients from non-responding/progressing patients was improved by stratifying the analysis voxels according to regions of interest (ROI) based on the LH similarity-based estimate of tumor change. Meaningful correspondence relationships between evaluated voxels are essential for accurate image-based quantification of tumor structure and function with voxel-wise analysis techniques. The LH similarity methods developed here can robustly evaluate the quality of spatial and temporal voxel correspondence relationships and provide an automated tool for ROI selection and voxel change stratification. It is readily extendable to the analysis of the wide array of anatomic, functional and biological imaging currently used to characterize tumors, guide therapy and assess response.

Ces travaux s'inscrivent dans la thématique de l'inférence géométrique dont le but est de répondre au problème suivant : étant donné un objet géométrique dont on ne connaît qu'une approximation, peut-on estimer de manière robuste ses propriétés? On se place dans cette thèse dans le cas où l'approximation est un nuage de points ou un ensemble digital dans un espace euclidien de dimension finie. On montre tout d'abord un résultat de stabilité d'un estimateur de normale basé sur l'analyse en composante principale, ainsi qu'un résultat de convergence multigrille d'un estimateur du Voronoi Covariance Measure qui utilise des matrices de covariance de cellules de Voronoi. Ces deux résultats, comme la plupart des résultats en inférence géométrique, utilisent la stabilité de la fonction distance à un compact. Cependant, la présence d'un seul point aberrant suffit pour que les hypothèses des résultats de stabilité ne soient pas satisfaites. La distance à une mesure est une fonction distance généralisée introduite récemment qui est robuste aux points aberrants. Dans ce travail, on généralise le Voronoi Covariance Measure à des fonctions distances généralisées et on montre que cet estimateur appliqué à la distance à une mesure est robuste aux points aberrants. On en déduit en particulier un estimateur de normale très robuste. On présente également des résultats expérimentaux qui montrent une forte robustesse des estimations de normales, courbures, directions de courbure et arêtes vives. Ces résultats sont comparés favorablement à l'état de l'art. / The purpose of geometric inference is to answer the following problem : Given a geometric object that is only known through an approximation, can we get a robust estimation of its properties? We consider in this thesis the case where the approximation is a point cloud or a digital set in a finite dimensional Euclidean space. We first show a stability result for a normal estimator based on the principal component analysis, as well as a result of multigrid convergence of an estimator of the Voronoi covariance measure, which uses covariance matrices of Voronoi cells. As most of geometric inference results, these two last results use the robustness of the distance function to a compact set. However, the presence of a single outlier is sufficient to make the assumptions of these results not satisfied. The distance to a measure is a generalized distance function introduced recently, that is robust to outliers. In this work, we generalize the Voronoi Covariance Measure to generalized distance functions and we show that this estimator applied to the distance to a measure is robust to outliers. We deduce a very robust normal estimator. We present experiments showing the robustness of our approach for normals, curvatures, curvature directions and sharp features estimation. These results are favorably compared to the state of the art.

No description available.

Charles Bonnet syndrome (CBS) is defined as complex persistent visual hallucinations in the absence of mental disorder. It is associated with advanced age and poor vision. It is common, with prevalence estimates of up to 63% among older people with significant visual impairment. CBS would not be diagnosed in the presence of dementia, but its relationship to milder cognitive impairment is unclear. The few studies that have examined this are underpowered and provide contradictory results. There are 16 case reports of dementia emerging in people with a diagnosis of CBS. These cases raise the possibility of an association between impaired insight at diagnosis of CBS and the subsequent development of dementia. This thesis reports the findings of a prospective cohort study which describes changes in cognitive functioning over one year in patients with CBS and age-matched controls. Participants were recruited from low vision and glaucoma assessment clinics. A clinical assessment was carried out by an old age psychiatrist, and participants had a detailed assessment of visual functioning. This thesis also describes the findings of the first study to use voxel-based morphometry (VBM) to investigate changes in volume of grey and white matter in CBS. Participants were recruited from the same clinics as the cohort study, and underwent MRI scanning on a 1.5T scanner, to a protocol designed to produce 1mm3 voxels. Twelve participants with CBS and ten controls were followed up. Two people in the CBS group developed dementia, while none did in the control group. The CBS group showed a mean change in the score on the Addenbrooke’s cognitive examination (ACE-R) of -3.7 points, compared to a change of +1.4 in the control group. This difference was not statistically significant. The CBS participants performed worse on the verbal fluency item of the ACE-R, a difference which was statistically significant. The VBM analysis was conducted on 11 CBS participants and 11 controls. The CBS group showed an increase in grey matter volume in the right cerebellar hemisphere. This difference retained significance after family-wise error correction, non-stationary correction, and ANCOVA to control for the effects of possible confounders. As far as the author is aware, these are the most methodologically robust studies to date to have investigated cognition and morphological brain changes in CBS. The findings of the cohort study were inconclusive. However, the two cases of dementia in CBS patients add weight to the suspicion that this is a clinically important outcome in the condition, and the finding of abnormalities in frontal lobe testing in participants with CBS fits with a theoretical model of visual hallucination generation. Moreover, this type of research appears to be acceptable to a frail and visually disabled population, and studies powered to investigate this issue more fully would be feasible. The VBM findings report the presence of underlying structural brain abnormalities in CBS, in a region not usually associated with visual hallucinations. Possible links with Lewy body dementia, and implications for theories of visual hallucinations, are discussed.

Conversion of discrete volumetric data to boundary representation is quite common operation. Standard approach to resolve this problem is to use well-known Marching cubes algorithm, which although simple and robust, generates low-quality output that requires subsequent post-processing. This master's thesis deals with uncommon algorithms used for isosurface extraction from volumes. The reader will be acquainted with fundamental principles of Hierarchical Iso-Surface Extraction method, that was independently implemented and tested in this work.

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第19616号 / 医博第4123号 / 新制||医||1015(附属図書館) / 32652 / 京都大学大学院医学研究科医学専攻 / (主査)教授 佐藤 俊哉, 教授 富樫 かおり, 教授 藤渕 航 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

Different signal processing techniques for synthetic aperture acoustic (SAA) and highresolution voxel radar (HRVR) sensing modalities for side-attack explosive ballistic (SAEB) detection are proposed in this thesis. The sensing modalities were vehicle mounted and the data used was collected at an army test site. More specifically, the use of a frequency azimuthal (fraz) feature for SAA and the fusion of a matched filter (MF) and size contrast filter (SCF) for HRVR was explored. For SAA, the focus was to find a signature in the target’s response that would vary as the vehicle’s view on the target changed. For the HRVR, the focus was put on finding objects that were both anomalous (SCF) and target-like (MF). The results in both cases are obtained using receiver operating characteristic (ROC) curves and both are very encouraging.