(Sunday, 7/14/2019)
Room: ePoster Forums
Purpose: Comparison of the accuracy of different image segmentation methods in evaluating the percentage of region of interest in CT scan images.
Methods: The segmentation of CT scan images is a challenging problem that has received an enormous amount of attention lately. In this research, two threshold and clustering approaches (implemented in MATLAB software) are used to segment the images. In the threshold-based approach, the OTSU multi-threshold algorithm and in the clustering approach, Fuzzy c-means algorithm (FCM), Probabilistic c-means algorithm (PCM) and Fuzzy possibility C-Means Clustering Algorithm (FPCM) techniques are used. The best method for segmentation is a method that minimizes the effect of CT measurement parameters, such as the size of the voxel, on the volume percentage changes of the region of interest. In order to evaluate the effect of the magnification factor on the volume percentage evaluated by different methods of segmentation, a cylindrical phantom with three cavities of the specified dimensions in 4 different magnifications was investigated by CT scan. The data was reconstructed by filter back projection algorithm (with Hann filter) and SIRT (classical Land Weber) method coded in MATLAB software. In the next step using MATLAB's Find circle technique, the evaluated level by different segmentation methods was calculated and the results were compared.
Results: Changing the distance from the source to the phantom, the volume evaluated by the ΟΤSU method has a rather large variation. While the volume evaluated by the fuzzy algorithm has less variation.
Conclusion: The level changes in fuzzy algorithms at different magnifications are less than the OTSU algorithm, so fuzzy algorithms are better methods.
Keywords
CT, Segmentation, Image Processing
Taxonomy
Contact Email
