A novel fuzzy approach for segmenting medical images
- PDF / 2,595,169 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 96 Downloads / 186 Views
METHODOLOGIES AND APPLICATION
A novel fuzzy approach for segmenting medical images Prabhjot Kaur1 · Tamalika Chaira2
© Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract In various medical imaging studies, the segmentation of the medical image is a crucial step. Due to the inherent behavior of the images, the computerized tomography (CT) scan/magnetic resonance imaging (MRI) images are not clear. As the medical images are captured using electronic devices, these images are ambiguous and unclear. This ambiguity in the images is due to factors related to noise and the environment while capturing images. A fuzzy set is a valuable method to deal with data uncertainty. In this paper, a novel clustering approach based upon a fuzzy concept is proposed, which enhances the vague MRI/CT scan image before segmentation. Initially, the image noise is minimized by processing eight immediate neighborhood pixels around each pixel in an image. After noise processing, upper and lower membership levels of the image are computed. Hamacher T-conorm is used as an aggregation operator to construct a new membership function using upper and lower membership levels. The enhanced image, created using the new membership function, is then segmented using Gaussian kernel-based fuzzy c means clustering. The experiment is performed on MRI/CT scan brain tumor images. Real data experiments demonstrate that the proposed algorithm has better performance when compared with existing methods both quantitatively and qualitatively. It is observed that even in the presence of noise, the proposed method exhibits better efficiency. Keywords Gaussian kernel · Interval Type 2 fuzzy set · Hamacher T-conorm · Fuzzy clustering · Biomedical imaging
1 Introduction Due to the intrinsic existence of the images, the segmentation of medical images is a crucial phase in many medical imaging studies. Many computer-aided diagnoses help physicians to detect the exact size of tumor/abnormal lesions in the brain, but the accurate size is still a challenging task. The accurate size and quality of the segmentation result in a mask image that is the basis for further detection of the brain tumors and may have a direct impact on the diagnosis. If the image is
Communicated by V. Loia.
B
Prabhjot Kaur [email protected] Tamalika Chaira [email protected]
1
Maharaja Surajmal Institute of Technology, GGSIP University, New Delhi, India
2
Aravali Pharma and Lifesciences, Dwarka, New Delhi, India
noisy or not clear, this process becomes more difficult. Clustering is the most commonly used segmentation technique for brain image segmentation, where different regions are clustered or grouped depending on the similarity of pixels. It is an unsupervised classification of data to groups called clusters. Brain images obtained from the CT scan or MRI process have several problems, such as nonuniform intensity, unsharp/unclear or vague boundaries. Apart from these, noise is also present in these images. The effective segmentation of brain images is still a dauntin
Data Loading...
