Wireless cyber-mammography is potentially a convenient screening method to be comfortable and effective in community and rural area early detection of breast cancer, but their interpretation is difficult due to the noise and low quality of images. In this paper, we study the accuracy of a Cyber-aided diagnosis system to help physicians to classify the detected regions in wireless mammogram images into malignant or benign categories. In this approach we investigate different sets of features and two classifier methods (SVM and GMM) and perform a comparative study to investigate the accuracy measurements in noisy condition. The results show that without any noise or errors, SVM classifier outperforms GMM; however GMM classifier is more robust and reliable in noisy circumstance especially in detecting malignant cases. The proposed study provides in-depth understanding of the accuracy and reliability of wireless mammography in early breast cancer detection.