Brain Matters Emphasis in MRI by Kernel Independent Component Analysis
Tomoko Tateyama, Zensho Nakao, Yen‐Wei Chen · 2007
We propose a new method for brain matters emphasis in MR images based on kernel independent component analysis (KICA). First the method mappes MRI data into a higher-dimensional implicit feature space. Then we extract kernel independent components from 3-dimensional MR images; PD image, Tl image and T2 image by KICA. Since the KICA algorithm is based on minimization of a contrast function, it can perform image processing, considering a higher-dimensional non-linear model. We also give experimental results which are very helpful to emphasize tissue clusters included in images; not only giving contrast emphasis of the images but also image comparisons by with those ICA analysis.