Estimación de información faltante en imágenes de resonancia magnética

The sources of noise in magnetic resonance are multiple and varied, the artifacts by movement and the spatial limitation of the captor are the best known. The best documented problem is the so-called partial volume, which consists of the fact that due to the spatial limitation of the captor, the limit between two different tissues is stored as a weighted signal between the two tissues.[3]. Medical diagnostic techniques have changed in recent years, one of the most widely used are radiological images due to its easy acquisition and sufficient information contained in the tissue studied, however the acquisition protocols are limited and the images obtained lose information important for diagnosis [12]. These deficiencies in the acquisition of information limits the radiologist and impairs the diagnosis and study of some diseases [46]. Magnetic resonance imaging (MRI) is widely used in medicine nowadays, yet a significant disadvantage is the amount of artifacts that affect the image during the acquisition process. As an example, Cardiac Magnetic Resonance (CMR) requires synchronization with the ECG to correct many types of noise. However, the complex heart motion frequently produces displaced slices that have to be either ignored or manually corrected since the ECG correction is useless in this case. This work presents a novel methodology that detects the motion artifacts in CMR using a saliency method that highlights the region where the heart chambers are located. Once the Region of Interest (RoI) is set, its center of gravity is determined for the set of slices composing the volume. The deviation of the gravity center is an estimation of the coherence between the slices and is used to find out slices with certain displacement. Another type of acquisition technique that is affected by the missing information is the Diffusion imaging (dMRI) is a magnetic resonance technique widely used to study the withe matter architecture and to understand their changes. The spatial resolution of brain diffusion weighted imaging (DWI) is limited due to high frequency of the image and brain structures like edges or bifurcations when the data are captured. In this approach the main idea is improved the spatial resolution using a dictionary learning strategy and for this way use the statistical dependence between different gradients of the same image for create a prior that improve the resolution. In the present work it has been different validation methods in order to made automatic detection and the increase resolution of abnormalities using probabilistic information estimation of the medical images.

• Autores: Salguero López, Jennifer
• Fecha de publicación: 2019
• Institución: Universidad Nacional de Colombia
• Repositorio: Universidad Nacional de Colombia
• Idioma: spa
• OAI Identifier: oai:repositorio.unal.edu.co:unal/76244
• Acceso en línea: https://repositorio.unal.edu.co/handle/unal/76244; http://bdigital.unal.edu.co/72361/
• Palabra clave: Missing information; Magnetic Resonance Images; Motion artifacts; SuperResolution algorithm Singular Value Decomposition; Pérdida de información; Resonancia magnética; Artefactos por movimiento; Super-Resolución SVD
• Rights: openAccess
• License: Atribución-NoComercial 4.0 Internacional