Automatic tumor segmentation in contrast-enhanced dynamic MR images by using artificial neural network technique

The aim of this study is to develop a method to learn the benign and malign tumors which are shown in the breast MR by giving contrast enhanced. Neural network is used for the learning ability of this application. Backprogation is used for the learning algorithm. Study consists of three main parts and additional parts. In the first part, Neural Network's features and backprogation, which is used in practice, are mentioned. In the second part, technique of MR and how to determine the tumors with contrast enhanced MR displays are explained. Developing the application and its results are explained in the third part. Backprogation that is used in application is explained more detailed and how the application learns is told. Developed application consists of four parts. First part is creating the Neural Network. Second part is educate the application with MR images which include benign and malign tumors. Third part is testing the application. Fourth part is the part which application makes analysis on the new MR. Application analyses the MR's, which it learned with the education, pixel by pixel and finds the tumor. Consequently, developed application is a computer aided diagnoses system, is aimed to help the doctors to diagnose. Keywords: Neural Networks, pattern recognition, image segmentation