| Summary: | 碩士 === 國立成功大學 === 電機工程研究所 === 82 === A method using neural networks and the concept of fractal
geometry to automatically locate a sequence of CT liver images
has been proposed. In this approach, we have found, an image is
first segmented using a Kohonen neural network based on the
gray level distribution. The next step is to use the texture
feature of liver images to remove the regions of other organs.
After this, a liver boundary can be allocated, then map this
area to next or previous images by a cost function to obtain
all of the CT liver boundaries of a patient. From the
experimental results, we have found that the proposed method
can not only properly segment out the liver boundary but also
show the locations of vessels and tumors clearly.
|
|---|
