Brain tumour segmentation and detection using modified region growing and genetic algorithm in MRI images

In modern medical research several approaches of image segmentation are used for the detection of brain tumour. Several pieces of medical equipment such as magnetic resonance imaging (MRI) scans, X-ray and computed tomography (CT) are used for diagnosis of brain tumour. This paper proposes a new segmentation method which combines modified region growing and genetic algorithm for detecting brain tumour. This consists of four steps-pre-processing, segmentation, classification and fitness calculation. Pre-processing uses Gaussian filter for removal of noise present in the image. The pre-processed image is segmented using modified region growing (MRG) method which includes the orientation constraint in addition to the intensity constraint used in region growing (RG) method. Back propagation neural network (BPNN) classifier classifies the tumour as normal or abnormal. Then a genetic approach of initial population and fitness calculation is done to find the optimum value for the best segmented tumour portion of the MRI image. The proposed approach overcomes dark abnormalities as well as over segmentation problem. Implementing the proposed method on MRI image helps in creating awareness to patients and it also serves as a perquisite for Radiologists, doctors in rural areas for providing effective treatment to the brain tumour patients.