Medical Image Classification using Neural Networks Techniques

Detection of tumors at an early stage is one important step in diagnosing abnormalities in a mammogram. In many cases, preprocessing process of raw images will involve image enhancement, filtering and determination of its textural feature. Similarly, this study also use the same affirmed preprocesses towards its images. One of the Gabor wavelet based algorithm namely Gabor filter was used to extract the feature of mammogram images. The output of this filter was then being used for classification purpose, so as to determine the characteristics of the normal tissue and abnormal tissue. Since Radial Basis Function (RBF) network is the best classifier for classifying problems particularly in term of its high accuracy rate, this classifier was used in this research. The performances of these classifiers were compared with Back Propagation (BP) network. 10-fold cross validation techniques was applied in order to measure the percentage of accuracy of RBF classifier. The work conducted proves that Radial Basis Function (RBF) network has outperformed BP network in classifying the normal and abnormal tissue of cell with 92.27% and 88.28% accuracy respectively.