| 摘要 |
A method for the automated segmentation of in vivo image data is disclosed. A region of carotid artery in a number of patients was imaged using MRI. Histological data for each imaged region was then obtained, identifying various atherosclerotic plaque components in the imaged region. A portion of the histological data, and the image data, was used to generate PDFs based on image intensity, and on morphological data (local wall thickness and distance from lumen). The remaining data was used to validate the method. A plurality of MRI images were taken at various weightings, and the images were registered and normalized. The lumen and outer wall boundary were identified. The PDFs were combined in a Bayesian analysis with the intensity and morphological data to calculate the likelihood that each pixel corresponded to each of four plaque components. A contour algorithm was applied to generate contours segmenting the images by composition.
|
