MULTIDIMENSIONAL TEXTURE EXTRACTION METHOD BASED ON BRAIN NUCLEAR MAGNETIC RESONANCE IMAGES

摘要

A multidimensional texture extraction method based on brain nuclear magnetic resonance images of groups. Regions of interest within brain nuclear magnetic resonance images of groups are segmented using a region growing method, and curvelet transform and contourlet transform methods are used to extract texture feature parameters of the regions of interest. The groups comprise an Alzheimer's patient group, a mild cognitive impairment patient group and a normal old-aged persons group. The texture feature parameters of the regions of interest comprise entropy, grey average, correlation, energy, degree of homogeneity, variance, maximum probability, inverse gap, clustering trend, contrast, sum average, difference average, sum entropy and difference entropy. The regions of interest comprise an entorhinal cortex region and a hippocampus region.