Tumor Detection from Brain Magnetic Resonance Images Using MRDWTA-RBFNNC

The automatic classification and detection of MR images (brain) for abnormality play very important role in the analysis and diagnosis of brain disorders. This manuscript proposed an abnormality detection method from brain MR images using the RBFNNC. MRDWT is utilized for the brain image preprocessing and also for feature extraction where preprocessing step comprises of grayscale MR image conversion and removal of noise from MR images. The recognition of abnormalities reveals the detection of benign types of tumors, malignant types of tumors and common brain conditions. Thirteen types of MRDWT-based features of the MR (brain) images were extracted by applying the DWT method which is mean, median, variance, power spectral density (PSD), standard deviation (SD), root mean square (RMS), correlation, entropy, energy, contrast, smoothness, skewness, homogeneity. Ninety-seven MR images were used for testing of the brain tumor of benign, malignant and normal brain condition. The accuracy percentage attained using proposed RBFNNC is 100% as compared with the FFNNC (97.87%) and BPNNC (98.94%).