A multi-objective optimization-based clustering framework for the brain tumor segmentation followed by Optimized ABC for feature selection using MRI scans

Abstract

Medical image segmentation is a method for automatically or semi-automatically finding
boundaries within a 2D or 3D image. The wide variety of medical images poses a significant
difficulty for image segmentation. MRI images guide for the identification and presence of tumor
in the brain. However, this process requires precise demarcation of the tumor region within the
brain scan. To address this, k-means is one of the most effective techniques used to separate the
pixels of interest from the background, but their effectiveness is dependent on the centroids'
initialization values. The main objective of the paper is to segment the tumor regions within brain
MRI images. Multi-objective Genetic algorithm-based K-means is used to segment the tumor from
the brain after converting the gray MRI image to a color image. The RGB color formed in the
image is based on its intensity. The paper further addresses the issue of feature selection and
proposes an improved ABC algorithm. The selected key points have been passed to training and
classification using Neural Networks. The simulation analysis demonstrates that the proposed
multi-objective GA based k-means exhibits an average precision improvement of 1.85% to 5.33%,
sensitivity improvement of 4.33% to 5.09% and with accuracy improvement of 2.57% to 4.44%
against k-means and single objective GA based k-means.