Mammography is the most effective technique to detect breast abnormalities. In most cases, mammograms are evaluated by radiologists. However, diagnosis performed radiologist has a lot of limitations. Computer Aided Diagnosis (CAD) with various methods had been developed to help radiologist in evaluating mammograms. This research developed CAD for mammography based on image segmentation using Markov Random Field with Simulated Annealing optimization (MRF/SA). We combined MRF/SA method with various preprocessing algorithms, such as median filter, histogram equalization, and CLAHE (Contrast Limited Adaptive Histogram Equalization). MRF/SA without any filter and contrast enhancement was also performed. A total of 210 mammograms with normal and abnormal findings were used. Abnormal category means mammogram with abnormalities findings whether in a form of benign lesion, malignant lesion, benign microcalcification or malignant microcalcification. ROC (Receiver Operating Characteristic) analysis was used to measure methods’ performance. The values of area under the ROC curve (AUC) for MRF/SA only, median filter + MRF/SA, histogram equalization + MRF/SA and CLAHE + MRF/SA are 0.731, 0.840, 0.798, and 0.746 respectively. Combination of median filter + MRF/SA has the highest AUC value indicated that this method has the best performance in distinguishing normal and abnormal images. Histogram equalization + MRF/SA has inferior AUC value compare to median filter + MRF/SA, but this combination has the highest sensitivity, 90.4%. This result shows that histogram equalization + MRF/SA is the most successful method in detecting abnormal images correctly.