نوع مقاله : علمی-پژوهشی
نویسندگان
1 گروه علوم مهندسی، دانشکده فناوری‌های نوین، دانشگاه محقق اردبیلی، نمین، ایران
2 عضو هیات علمی / دانشکده مهندسی پزشکی، دانشگاه صنعتی سهند
3 گروه بیوالکتریک، دانشکده مهندسی پزشکی، دانشگاه صنعتی سهند، تبریز، ایران
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
The segmentation of brain tissues is a crucial step in evaluating morphological changes in various brain regions for the identification of different diseases. This process is influenced by factors such as noise and intensity non-uniformity. Fuzzy c-means clustering (FCM) is a widely used method for image segmentation but is sensitive to noise, and its convergence rate is affected by data distribution. FCM-based approaches typically use Euclidean distance for clustering, assuming a spherical distribution of data, which overlooks distance variations in similar and compact clusters. Additionally, varying levels of intensity non-uniformity can impact clustering performance. To address these limitations, this study presents the Conditional Spatial Gustafson-Kessel (CSGK) algorithm, which performs well in segmenting compact clusters such as cerebrospinal fluid (CSF) by considering data distribution in an elliptical space. The robustness of the conventional Gustafson-Kessel algorithm is enhanced by incorporating both local and global spatial information into a weighted membership function. Furthermore, a Wiener filter combined with wavelet transform (WFWT) is applied during preprocessing to reduce the sensitivity of input data to intensity non-uniformity. Experimental results demonstrate that CSGK is an accurate algorithm for segmenting multiple brain tissues across different levels of noise and intensity non-uniformity.
کلیدواژه‌ها [English]
)