Please use this identifier to cite or link to this item:
|Title:||A multiresolution segmentation technique for spine MRI images|
|Keywords:||Fuzzy c-means clustering|
|Citation:||Li, H., Yan, C.H., Ong, S.H., Chui, C.K., Teoh, S.H. (2002). A multiresolution segmentation technique for spine MRI images. Proceedings of SPIE - The International Society for Optical Engineering 4684 III : 1709-1717. ScholarBank@NUS Repository. https://doi.org/10.1117/12.467141|
|Abstract:||In this paper, we describe a hybrid method for segmentation of spinal magnetic resonance imaging that has been developed based on the natural phenomenon of "stones appearing as water recedes". The candidate segmentation region corresponds to the "stones" with characteristics similar to that of intensity extrema, edges, intensity ridge and grey-level blobs. The segmentation method is implemented based on a combination of wavelet multiresolution decomposition and fuzzy clustering. First thresholding is performed dynamically according to local characteristic to detect possible target areas, We then use fuzzy c-means clustering in concert with wavelet multiscale edge detection to identify the maximum likelihood anatomical and functional target areas. Fuzzy C-Means uses iterative optimization of an objective function based on a weighted similarity measure between the pixels in the image and each of c cluster centers. Local extrema of this objective function are indicative of an optimal clustering of the input data. The multiscale edges can be detected and characterized from local maxima of the modulus of the wavelet transform while the noise can be reduced to some extent by enacting thresholds. The method provides an efficient and robust algorithm for spinal image segmentation. Examples are presented to demonstrate the efficiency of the technique on some spinal MRI images.|
|Source Title:||Proceedings of SPIE - The International Society for Optical Engineering|
|Appears in Collections:||Staff Publications|
Show full item record
Files in This Item:
There are no files associated with this item.
checked on Aug 21, 2018
checked on Jul 27, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.