Breast cancer is the most common and the second lethal cancer among women in the United States. Our group is constructing a dedicated cone beam breast CT (CBCT) system to provide true 3D image to improve the screening and diagnostic of breast cancer. Our result shows that dedicated CBCT out-perform a lot than conventional CT when detecting micro-calcification which is essential to the detection of early stage breast cancer. I also explored the possibility of using the super parallel computing power of GPU with CUDA environment to deal with data-immense and computationally-intensive image reconstruction process of CBCT. My results show that FDK algorithm image reconstruction with GPU is over 10 times faster than that with our PC cluster system. The faster and accurate image reconstruction implies potential new applications in diagnostic and therapy technology.