Affiliations: [a] ECE Department, Noorul Islam University, Thuckalay, TamilNadu, India | [b] ECE Department, Ponjesly College of Engineering, Nagercoil, TamilNadu, India
Correspondence:
[*]
Corresponding author: M. Sucharitha, ECE Department, Noorul Islam University, Thuckalay, TamilNadu, India. Tel.: +91 9488283296; E-mail:sucharitha_m2002@yahoo.co.in
Abstract: BACKGROUND: Medical image segmentation is an
essential step for most consequent image analysis tasks. Medical images can
be segmented manually, but the accuracy of image segmentation using the
automated segmentation algorithms is more when compared with the manual
calculations. In this paper, an automated segmentation and classification of
tissues are proposed for MR brain images.
OBJECTIVE: To classify MR brain image into three
segments such as Grey Matter (GM), White Matter (WM) and Cerebro-Spinal
Fluid (CSF). Classification of brain into tissues helps to diagnose several
diseases such as tumors, Alzheimer's disease, stroke, multiple sclerosis.
METHODS: An unsupervised clustering technique
such as Fuzzy C-Means (FCM) algorithm has been widely used in segmenting the
images. The spatial information is not fully utilized by the conventional
clustering algorithm and hence it is not applicable for clustering a noisy
image. We incorporate a method for image clustering called out as
Reformulated Fuzzy Local information C-Means Clustering algorithm [RFLICM]
which is a variant of traditional Clustering algorithm by considering both
spatial and gray level information. In RFLICM, spatial distance is replaced
by local coefficient of variation in a fuzzy manner.
RESULTS: Experiments are conducted on brain
images to validate the performance of the proposed technique in segmenting
the medical images and the efficiency achieved in the presence of salt and
pepper noise is 99.86%.
CONCLUSION: Standard FCM, Fuzzy Local
information C-means clustering algorithm [FLICM], Reformulated Fuzzy Local
information C-means clustering algorithm [RFLICM] are compared to explore
the accuracy of our proposed approach. Clustering results show that RFLICM
segmentation method is appropriate for classifying tissues in brain MR
image.
