フラクタル次元とノイズと分解能(一般講演-ニューラルネット・解析法・フラクタル-)

概要

論文の詳細を見る
Fractal dimension images were generated by box-counting method for MRI and PET glucose metabolic images of the brain. An effect of the noise and spatial resolution on the value of fractal dimension was investigated by smoothing the brain images in a normal volunteer and a patient with corticobasal degeneration. In MRIs with high resolution and low noise, mean values of fractal dimension in the region of degenerated cerebral cortex was shown to be lower than those in the normal cerebral cortex and the relation was not changed by ten times smoothing, which is blurring the image and is reducing the noise. Then the low value of the fractal dimension in the lesion shows character of the corticobasal degeneration. In PET images with low resolution and high noise, however, the difference of the fractal dimension, between the normal and the CBD patient, was disappeared by more than 4 times smoothing. This result shows the spatial resolution of the PET images is critical to evaluate the characteristics of the texture by fractal dimension.
バイオメディカル・ファジィ・システム学会の論文
1997-10-16