Brain abnormality detection from MRI of human head scans using the bilateral symmetry property and histogram similarity measures

The proposed method aims to detect brain abnormality using bilateral symmetry property about the interhemispheric fissure (IHF) of human head scans. MRI brain has structural symmetry between the right cerebral hemisphere (RCH) and left cerebral hemisphere (LCH) of brain cerebrum. Any brain abnormalities due to tumors, hemorrhage, etc disturbs the similarity between the two hemispheres. We split the right and left hemisphere of MRI image using central vertical line and compare the histograms using similarity measures. Fifteen histogram similarity measures (HSM) are used to generate the threshold vector (TV). The TV obtained from training images using fuzzy c means based clustering technique. TV values are helpful to classify the MR images with the help of artificial neural network (ANN) classifier in the testing process. Hundred of axial images of normal and abnormal subjects are taken for training process. We test our algorithm with 66 volumes collected from three popular repositories. The average of sensitivity, specificity and accuracy rates are 98%, 96% and 94%. The error rates are 2% false alarm and 4% missed alarm. The proposed method is able to classify the brain images to normal and abnormal classes with high accuracy and less time. The proposed method performance was compared with six existing methods and produced better results than other methods.