LAUSR

ABSTRACT Segmentation of brain MR images for the detection of various healthy brain tissues such as white matter, gray matter, and cerebrospinal fluid is of immense interest to detect and to diagnose different brain-related disorders at the primitive level. MR image segmentation becomes a difficult task owing to the presence of intensity inhomogeneity (IIH), noise, partial volume effects, and intrinsic nature of the MR images. This paper proposes an efficient, region-based, energy minimization technique named as anisotropic multiplicative intrinsic component optimization (AMICO) to segment the brain image in the presence of IIH and noise and to detect different healthy brain tissues. The proposed algorithm utilizes a powerful anisotropic diffusion filter to denoise the image. The MICO algorithm segment the denoised image after correcting IIH. In the proposed technique, MR brain image is decomposed into two multiplicative and intrinsic components, such as the true image component and the bias field component. Brain tissue physical properties are represented by the component of true image and the IIH is characterized by the bias field component. Optimization of these two multiplicative and intrinsic components by employing the proposed effective energy minimization process, result in IIH correction and tissue segmentation simultaneously. The pursuance of the proposed technique is compared with some other existing techniques using the parameters, dice similarity coefficient, sensitivity, specificity, and segmentation accuracy. The results validated the excellent performance of the AMICO in detecting various brain tissues consisting of various levels of IIH and noise.