Breast cancer is one of the major causes of death among women. Early detection of breast cancer is possible by the detection of clustered microcalcifications on X-ray mammograms. Texture is an important characteristic used in identifying objects or region of interest in a digitised mammogram. This work focuses on a statistical texture analysis method called Surrounding Region Dependence Method (Kim and Park, 1999) - based on second order histogram in two surrounding regions. Six textural features are extracted and are used to classify region of interests into positive ROIs, containing clustered microcalcifications and negative ROIs composed of normal breast tissues. A 3-layer backpropagation neural network is used as a classifier. Results are evaluated using Receiver Operating Characteristics analysis.