Soft computing techniques based brain tumor detection and segmentation analysis

Abstract

The abnormal growth of the cells in brain region leads to the formation of tumor. At present, the treatments for brain tumours are radiation and surgery which are suggested by physician. Radiation treatment slows down the spreading capability of tumours in other brain regions and slowly kills the affected brain cells. Surgery removes the affected brain cells in brain and through this the spreading to other regions in the brain is prevented. For the case of proper surgery, the location identification of the abnormal cells in brain region is important. If it is not properly and completely removed from the brain, then few affected cells skipped from the surgery affects the other cells in the brain. The brain tumours are classified into either benign or malignant based on their capability of spreading. The benign are abnormal tissues which are not spreading to nearby tissues and it can be cured by proper medication suggested by physician. The malignant are an also abnormal tissue which spreads or affects the nearby tissues and it can be cured by medication. The only solution is to remove these affected tumor regions through proper surgery by physician. The affected malignant tumor cells region is manually detected by physician or radiologist in conventional method. This is the time consuming and error probe methodology due to manual intervention. This limitation is overcome by proposing the computer aided automatic approach for brain tumor image classification. Detection and diagnosis of brain tumor is complicated due to its similar characteristics between tumor pixels and non-tumor pixels in brain image. This research work proposes an efficient methodology for the detection and segmentation of tumor region in brain. The proposed methodology has the following stages as image registration, noise reduction, transformation, feature extraction and classification. The linear image registration technique is used to align the reference image with respect to source brain image to obtain higher classification rate.