Agglutination Multivariate Optimal Spanning SOLOv3 Algorithm: Novel Augment-Based Approach for Early-Stage Brain Tumor Image Segmentation

Abstract

In recent days instance image segmentation made great work with the real-time rapid development of video analytics using neural networks. Still, there are two main challenges faced during the segmentation of image objects and creating the relationship between each pixel for identification of infected cells in brain image. Commonly the T1 and T2 image was used, but in the proposed work the FLAIR three-dimensional based brain skull image was referred and identified. In the proposed work, many mining processes of frameworks for the instant detection of the image object. The proposed work introduced the new algorithm called Agglutination Multivariate Optimum Spanning SOLOv3 algorithm (AMOSv3 Algorithm). In this proposed mechanism, preliminary semantics are instance classification and prediction of location in indistinguishable image objects surrounded with original instance-mined images, to ensure fully excavated. For the proposed work two different datasets were used, BraTS9 and Fluid Attenuated Inversion Recovery (FLAIR). The matrix calculation was done with the proposed AMOSSV3 Algorithm. This mechanism is calculating the regional area based on the matrix value, which is calculated by the watershed algorithms. This algorithm gives a better accuracy score compared to others.