Segmentation of white matter, grey matter and cerebrospinal fluid from brain MR images using a modified FCM based on double estimation. (July 2021)

• Record Type: Journal Article
• Authors: Tavakoli-Zaniani, Maryam; Sedighi-Maman, Zahra; Fazel Zarandi, Mohammad Hossein
• Is Part Of: Biomedical signal processing and control. Volume 68(2021)
• Publication Date: 2021-07
• Subjects: Magnetic resonance image -- Image segmentation -- Fuzzy C-means -- Double estimation -- Noisy image; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102615
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Developed a new fuzzy clustering method for segmenting noisy brain MRIs. Incorporated both double estimation and the kernel distance metric. Demonstrated that the new method offers comparable results to KWFLICM. Showed the computational efficiency of the new clustering algorithm. Abstract: This paper presents a new fuzzy-based method for the segmentation of brain structures from noisy magnetic resonance (MR) images, in the presence of noise. Our algorithm is a new extension of the fuzzy C-means (FCM) algorithm. The proposed algorithm is developed by modifying the objective function in the FCM using double estimation by incorporating both the original and denoised images in place of using solely the denoised image. To the best of our knowledge, the proposed algorithm is the first extension of the FCM method that is capable of segmenting images (per pixel) based on both noisy and denoised image estimates. In this algorithm we: (a) introduce a novel formulation that assigns weights for each estimation using spatial image information and (b) apply a kernel distance metric for image segmentation. This formulation is highly applicable in segmenting images corrupted by high levels of noise. Experimental results on both simulated and original MR images are presented to demonstrate the robustness and effectiveness of our proposed algorithm in the presence of noise. These results are compared to the nonlocal fuzzy C-means method (LNLFCM), discrete cosine transform-LNLFCM (DCT-LNLFCM), kernel weighted fuzzy local information C-means (KWFLICM), and bias correction embedded fuzzy C-means with spatial constraint (BCEFCM-S) algorithm.