<\/p>\n
This thesis deals with the detection of new multiple sclerosis (MS) lesions in longitudinal brain magnetic resonance (MR) imaging. This disease is characterized by the presence of lesions in the brain, predominantly in the white matter (WM) tissue of the brain. The detection and quantification of new lesions are crucial to follow-up MS patients. Moreover, the manual detection of these new lesions is not only time-consuming, but is also prone to intra- and inter-observer variability. Therefore, the development of automated techniques for the detection MS lesions is a major challenge. After a thorough analysis of the state-of-the art in MS lesion detection approaches, we present a new classification of techniques pointing out their main strengths and weaknesses. A complementary quantitative evaluation of some of the most remarkable methods in the literature is also provided. Subsequently, we present a new proposal based on a change detection approach, which combines various characteristics of different MR image modalities. Firstly, several preprocessing methods are included in the pipeline to improve the quality of MR images. We analyze these processes as well as several rigid and non-rigid image registration methods in detail.<\/p>\n
The subtraction of the baseline and follow-up images is used to determine changes between the images. Moreover, we apply a WM masking step in order to reduce the search space for lesions only within WM. Afterwards, we apply a threshold to the subtraction images. Although determining the threshold can be done by experts, we propose an automated thresholding process which provides a satisfactory trade-off between sensitivity and specificity. Finally, we refine the candidate lesions detected using lesion features, particularly in order to reduce false positive lesions. For this purpose, including the baseline and follow-up images, we join both results obtained from PD-w and T2-w images in a supervised and an unsupervised manner. Experimental results are evaluated on a database of 20 MS patients with a variable lesion load, where manual segmentation provided by experts was available. The evaluation, carried out in a quantitative and qualitative manner, includes a comparison and uses several metrics for detection and segmentation.<\/p>\n
___________________________________<\/p>\n
Aquesta tesi es centra en la detecci\u00f3 autom\u00e0tica de lesions noves d’esclerosi m\u00faltiple (EM) en estudis longitudinals del cervell mitjan\u00e7ant l’\u00fas d’imatges de resson\u00e0ncia magn\u00e8tica (RM). Aquesta malaltia es caracteritza per la pres\u00e8ncia de lesions al cervell, predominantment en el teixit de la mat\u00e8ria blanca, i la detecci\u00f3 i la quantificaci\u00f3 de les noves lesions s\u00f3n elements crucials per al seguiment dels pacients. No obstant aix\u00f2, la detecci\u00f3 manual d’aquestes noves lesions no nom\u00e9s requereix de molt temps, sin\u00f3 que tamb\u00e9 \u00e9s propensa a la variabilitat intra- i inter-observador. Cal tenir en compte que les lesions d’EM s\u00f3n molt petites en comparaci\u00f3 amb tot el cervell. Per tant, el desenvolupament de t\u00e8cniques autom\u00e0tiques per a la detecci\u00f3 de lesions d’EM \u00e9s un gran repte.<\/p>\n
Despr\u00e9s d’un an\u00e0lisi exhaustiu de l’estat de l’art de les diferents t\u00e8cniques de detecci\u00f3 de lesions d’EM, en aquesta tesi es presenta una nova classificaci\u00f3 assenyalant-ne les principals fortaleses i debilitats. Tamb\u00e9 es proporciona una avaluaci\u00f3 quantitativa complement\u00e0ria d’alguns dels m\u00e8todes m\u00e9s rellevants en la literatura. Posteriorment, es presenta una nova proposta, que combina diverses caracter\u00edstiques de les diferents modalitats d’imatges de RM, basada en la subtracci\u00f3 d’imatges per tal de determinar els canvis entre una imatge basal i una de seguiment. En primer lloc, s’inclouen en la proposta m\u00e8todes de preprocessament, per tal de millorar la qualitat de les imatges de RM, aix\u00ed com m\u00e8todes de registre d’imatges r\u00edgids i no r\u00edgids. S’analitza en molt deteniment l’efecte d’aquests preprocessats en el resultat final de la proposta. Tamb\u00e9 s’aplica un pas d’emmascarament de la mat\u00e8ria blanca amb la finalitat de reduir l’espai de cerca de les lesions nom\u00e9s dins de la m\u00e0scara.<\/p>\n
Posteriorment, s’aplica un llindar a les imatges de resta. Tot i que la determinaci\u00f3 del llindar pot ser realitzada pels experts, en aquesta tesi es proposa un proc\u00e9s automatitzat de detecci\u00f3 del llindar \u00f2ptim que proporciona un comprom\u00eds satisfactori entre sensibilitat i especificitat. Finalment, es refinen les lesions candidates detectades utilitzant les caracter\u00edstiques de la lesi\u00f3, sobretot amb la finalitat de reduir la detecci\u00f3 de falsos positius. Per a aquest prop\u00f2sit, s’inclouen les imatges basals i de seguiment, i es fusionen els resultats obtinguts a partir d’imatges PD-w i T2-w d’una manera supervisada i tamb\u00e9 no supervisada. Els resultats experimentals s’avaluen en una base de dades de 20 pacients amb EM amb una c\u00e0rrega variable de la lesi\u00f3, on es disposa tamb\u00e9 de la segmentaci\u00f3 manual proporcionada pels experts. L’avaluaci\u00f3 s’ha realitzat de forma qualitativa i quantitativa, incloent una comparaci\u00f3 dels diferents processos i usant diverses m\u00e8triques per a la detecci\u00f3 i segmentaci\u00f3.<\/p>\n","protected":false},"excerpt":{"rendered":"
This thesis deals with the detection of new multiple sclerosis (MS) lesions in longitudinal brain magnetic resonance (MR) imaging. This disease is characterized by the presence of lesions in the brain, predominantly in the white matter (WM) tissue of the brain. The detection and quantification of new lesions are crucial to follow-up MS patients.… <\/p>\n","protected":false},"author":1,"featured_media":3036,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"neve_meta_sidebar":"","neve_meta_container":"","neve_meta_enable_content_width":"","neve_meta_content_width":0,"neve_meta_title_alignment":"","neve_meta_author_avatar":"","neve_post_elements_order":"","neve_meta_disable_header":"","neve_meta_disable_footer":"","neve_meta_disable_title":"","footnotes":""},"categories":[75,12,58],"tags":[87,88,89],"class_list":["post-3035","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-medical-imaging-lab","category-news","category-scientific-results","tag-detection","tag-lesion","tag-multiple-sclerosi"],"acf":[],"yoast_head":"\n