An Automated Map Process Based Improved Fuzzy C-Means Algorithm for Pathological Detection in MR Image

Abstract

Automated brain MR slices segmentation process is difficult, and further difficult is the process of detecting the tumor and tissue regions, with a constraint of delivering higher segmentation accuracy within reduced processing time. Automated algorithms were developed with an onus of reducing the intricacies involved during the manual inspection of the pathologies (radiologist/operator involvement). The shortages of an automated process are overthrown with the development of a novel combination of soft computing algorithms, and it employs automated map and clustering approaches. Self-Organizing map (SOM) and Improved Fuzzy C-Means clustering (IFCM) are the automated map and clustering approaches that are used to precisely provide the MRI slice analysis. The authors have utilized the quality metrics, such as Dice overlap Index (DOI), Jaccard index, Peak Signal to Nosie Ratio (PSNR) and Mean Squared Error (MSE) for verifying the performance of the SOM based IFCM, and the recommended algorithm tenders the corresponding values of the above as 84.83%, 91.69%, 0.0824 and 49.25dB. The novel SOM- IFCM algorithm delivers better demarcation outcomes when compared with other soft computing approaches. The exemplified outcomes of the proposed SOMIFCM algorithm provides superior segmentation quality of MR brain slices and offers versatile usage to the radiologists