| Summary: | 碩士 === 國立成功大學 === 工程科學系碩博士班 === 91 === Low back pain is one of the most common health
problems treated by orthopedic surgeons. According
to the experience, about eighty percent of adults
have had low back pain and the age is mostly
located between forty-five and sixty. For those low
back pain patients, Computer Tomography (CT) and
Magnetic Resonance Imaging (MRI) are two helpful
instruments for clinical diagnosis. In general, CT
and MRI images are inspected by turns for finding
the causes. However, it is inconvenient to change
images frequently back and forth. Thus, the purpose
of this study is to use image registration
technique for integrating both of the information
of CT and MRI images to provide more complete
information for diagnosis.
For CT images, the lumbar vertebra is firstly
segmented out. Then the Laplacian edge detector is
used to extract the contours of the segmented
images. These extracted contours from CT images are
used as the reference models. For those MRI images,
their gradient images are obtained. Then they are
converted into its corresponding distance maps by
using distance transformation. Finally, the
corresponding relationship of spatial position
between the reference models and the distance maps
is found by using chamfer matching techniques for
registration. External landmarks are also used to
estimate the accuracy and robustness of the
registered result of the proposed algorithm for
confirming the application of clinic. For helping
the observation, a visualization system, including
2D image, 3D reconstruction and registered image
display. It shows the doctor the corresponding
relationship between nerve roots and the edge of
lumbar vertebra more easily.
|
|---|
