A Study on Mining of Tumor Pixel Group in Flair Modality Brain MRI

Abstract

Recently, a number of examination procedures are proposed to extract the tumor section of brain Magnetic Resonance Images (MRI) recorded with varied modalities. The major aim of the proposed study is to verify the need of the soft computing approaches in brain MRI examination. In this work, the brain MRI recorded with the Flair modality is adopted for the investigation and all the test images considered are obtained from Brats 2015 database. Initially, a threshold masking procedure is implemented to extract the Tumor Pixel Group (TPG) from the test pictures and the demerit of this threshold masking technique is discussed with appropriate results. Later, an automated technique based on Social Group Optimization (SGO) is implemented to extract the TPG with a hybrid imaging procedure based on the combination of the Otsu’s thresholding and watershed segmentation. The proposed approach is tested on the chosen test pictures from the Brats 2015 database and the results are validated by computing the well-recognized Picture Similarity Parameters (PSP). The experimental results of this study confirm that when a hybrid technique is implemented, it is possible to increase the accuracy in brain MRI assessment.