用於核磁共振影像之影像分割與上色演法

Abstract

核磁共振成像是一個知名的醫療成像技術。核磁共振成像可以產生詳細的人 體內部結構的影像，而不需做任何侵入性治療。核磁共振成像也屬非放射性治 療，對人體無任何副作用。本論文提供了一種方法，針對核磁共振成像的灰階影 像分群上色，其中的顏色是根據不同的結構決定。利用影像分割方法─影像譜方 法編織結果，並且考慮腦部結構的對稱特性去做結構的分群；再依照分割出來的 結構間彼此的相關性而決定結構顏色的對比度。由測試結果顯示，上色後的核磁 共振影像相較於灰階的核磁共振影像較能夠突顯一些細微的結構，亦可讓我們由 核磁共振影像得到更多的資訊，這可能是在不久的將來應用於醫療診斷。

Magnetic resonance imaging (MRI) is a well-known medical imaging technique that can produce images of inside structures from a human body in detail without any mechanical invasion and implantation. This study provides a method for coloring the grayscale MRI images, in which the colors are decided according to the structures that can be clustered. The experimental results show that tiny structures in a human brain will be displayed clearer in the colored structures than that in the gray-scale MRI images. Meanwhile, the structures in the brain are assumed to be symmetric in this study. The clustering algorithm is developed in the scheme of brain symmetry by adopting the image segmentation method and spectral matting. The color contrast between any two structures is decided based on the correlation of two clusters. The process can reveal some subtle structures in the MRI image and more information can be obtained from the MRI image, which could be applied on medical diagnosis in the near future. ii