Classification of Trained Input Images using Neural Networks

Breast Cancer is considered as the most frequent cancer among women nearly 2.1 million women are affected in each year. In order to recover the scenario, various image processing algorithms are used to detect breast cancer in its initial stage. Breast cancer can be usually recognized by following methodologies such as Mammograms, MRI, Ultrasound and Biopsy. Mammogram is a preliminary diagnosis methodology in breast cancer. In the proposed method three main stages are used. In the first stage input image is preprocessed by Discrete Wavelet Transform, Gray Level Co-occurrence Matrix is used as second stage to extract features in an image and In third stage Probabilistic Neural Network is used for classification of trained input images. Finally, the percentage of affected cells by tumor is calculated by Fuzzy C Means algorithm. By using proposed method the accuracy of tumor cells detection in its preliminary stage has been improved.