An Automated Two-Stage Brain Tumour Diagnosis System Using SVM and Geodesic Distance-Based Colour Segmentation

In the medical profession, brain tumour is a very crucial illness. A brain tumour is an unwanted mass growing in the brain cells; if it’s not prevented, it will eventually cause death. Therefore, tumour diagnosis is essential. Magnetic resonance imaging (MRI) is used to identify the brain tumour quickly. The approach of detecting a brain tumour from human eyesight is quite difficult. The proposed work automatically diagnoses the brain tumour. This proposed technique has two stages: classification and segmentation. The classification stage is used to classify the T2W-MRI images into a tumour and normal using 8 x 8 blocks with gray-level co-occurrence matrix (GLCM) features using a support vector machine (SVM). The second stage segments the FAIR and T1C type MRI images using colour-based segmentation technique. This proposed method uses the BraTS2013 dataset. Classification and segmentation result is calculated by sensitivity, specificity and accuracy. In the segmentation, it additionally uses the dice similarity coefficient (DSC) to find the accuracy. The outcomes denote the proposed method's accuracy of classification as 96.66% and the DSC of segmentation accuracy as 80%.