A novel threshold score based multiclass segmentation technique for brain magnetic resonance images using adaptive opposition slime mold algorithm. Issue 4 (4th February 2022)

• Record Type: Journal Article
• Authors: Naik, Manoj Kumar; Panda, Rutuparna; Samantaray, Leena; Abraham, Ajith
• Is Part Of: International journal of imaging systems and technology. Volume 32:Issue 4(2022)
• Page Start: 1397
• Page End: 1413
• Publication Date: 2022-02-04
• Subjects: brain MR image -- evolutionary computing -- multilevel thresholding -- slime mold algorithm; Imaging systems -- Periodicals; Image processing -- Periodicals; 621.367
• DOI: 10.1002/ima.22713
• Languages: English
• ISSNs: 0899-9457

Abstract

Abstract: Clinicians can detect diseases early, thanks to the digital image processing methodologies, which improve health together with the treatment experience. The technology of magnetic resonance imaging (MRI) is frequently employed in the brain, research for any kind of related illness. The brain MR image requires precise automated thresholding for a meaningful representation to aid doctors, because of its different modalities and complexity. The majority of the threshold selection strategies are based on entropy. However, these strategies are limited by their reliance on the spatial distribution of gray values. There is also a pressing need to develop a thresholding technique that is independent of the spatial distribution, making it more suitable for a variety of modalities and complexity, such as the brain MR images. A novel non‐entropic maximizing objective function for the multilevel thresholding approach using a threshold score (TS) is presented in this paper, to address these concerns. An evolutionary TS‐AOSMA approach, using the optimizer called adaptive opposition slime mold algorithm (AOSMA), is suggested to lower the computational cost of TS‐based multiclass segmentation, which is a novel idea. The proposed approach is evaluated on T2‐weighted brain MR imaging slices from Harvard Medical School's whole brain atlas dataset. When compared to the state‐of‐the‐art Kapur's, Tsallis, and Masi entropy‐based technologies, the proposed scheme offered better quantitative and qualitative outcomes. The recommended strategies may be useful in medical image analysis.