Past Research |
Automated Segmentation & Classification of Medical Images
Students: Jesse C. Ma, James L. Lee, Te-shen "Dickson" Liang
|
We have developed application-specific techniques for the automated
segmentation of magnetic resonance images of the head.
Segmentation involves the labeling of homogenous regions in the 3-D image data
set.
After segmentation, each voxel is classified according to tissue type: white
matter, gray matter, cerebral spinal fluid, etc.
Once the 3-D data set has been segmented and classified, we apply computer
graphics methods to generate a rendered, 3-D view of the surface of the brain.
We have investigated several approaches to the segmentation/classification
problem, including 3-D edge detection, fuzzy c-means, neural networks, and the
watershed transformation.
Applications of this work include visualization of data for surgical planning,
visual support for diagnosis of neurological disorders and tumors, and
metrological studies in neuroscience.
Publications:
-
Jesse C. Ma and Jeffrey J. Rodriguez, "MR Image Segmentation
Using a
Fuzzy-Based Neural Network," in Proc. 1995 IEEE Intl. Conf. on
Neural Networks, vol. 5, Perth, Australia, Nov. 27 - Dec. 1, 1995,
pp. 2190-2195. [ PDF ]
-
Jesse C. Ma and Jeffrey J. Rodriguez, "Segmentation of
Multidimensional
MR Images Using a Fuzzy Neural Network," in Applications of
Digital Image Processing XVII, A. G. Tescher, Ed., Proc. of SPIE, vol.
2298, 1994, pp. 636-43. Presented at the SPIE Intl. Symp. on
Optics, Imaging, and Instrumentation, San Diego, CA, July 26-29,
1994. [ PDF ]
|
|