Efficient local asymmetric region detection in medical images
一种在医学图像中挖掘非对称区域的方法

A new method for efficient local asymmetric regions detection in medical images was proposed by removing symmetric regions. First, the 3D image was aligned to its reflex using symmetry based rigid image registration algorithm. Then based on the registered image, a modified adaptive erosion operator was used to remove the approximately symmetric region. These two steps were iteratively repeated using a multi-resolution scheme to extract the final asymmetric region. The experiment to detect cholesteatoma in human head CT shows that the proposed algorithm detects asymmetric regions correctly in 80% cases.