One of the most effective techniques for early breast cancer detection is Digital mammography. In this technique, the electronic images of breast are taken and stored in the computer. Abnormalities in these images can be found by Gray-Level Co-occurrence Matrix (GLCM) and its textural features. The calculation of GLCM and it's textural features is a time consuming process. The affordable solution is Graphical Processing Unit (GPU) based Compute Unified Device Architecture (CUDA) framework where the calculations can be made parallel. The sequential program on CPU as well as parallel program on GPU of GLCM and textural feature calculation have been implemented. The results of optimized parallel implementation on GPU show up to 12 times faster calculation over sequential implementation on CPU. © 2017 IEEE.