Pema Choejey. Segmentation of medical images using statistical methods with evaluation on a public Database . Master's Degree(Information Technology). King Mongkut's University of Technology North Bangkok. Central Library. : King Mongkut's University of Technology North Bangkok, 2007.
Segmentation of medical images using statistical methods with evaluation on a public Database
Abstract:
Segmentation and interpretation of medical images which are characterized by
complex and variable structures is apparently a difficult problem. Active Appearance
Model (AAM) method has proved to be a successful tool for effectively modeling and
interpreting variable images in a variety of applications. AAM contains a statistical
model of both shape and gray-level appearance of the object of interest, which can
subsequently be used in matching new instances of shape and texture in unseen
images. Since iterative matching during the search involves prediction of model
parameters, which optimize the best fit between the model and the target image,
image matching can be treated as an optimization problem. In this study, we propose
to adapt a continuous value coded genetic algorithm (GA) as an optimization method
to predict the optimal parameters during image matching. In order to compare the
performance and accuracy of standard AAM and AAM optimized by GA, we also
propose to use fast search algorithm based on canonical correlation analysis (CCA).
The proposed methods were evaluated using publicly available database containing
247 chest radiograph images. Our experimental results show that the performance of
AAM based on GA optimization is comparable to the standard approach. CCA
approach performs consistently better than the two approaches.