Investigation of the FFANN Model for Mammogram Classification Using an Improved Gray Level Co-occurrences Matrix

Abstract

In the current years the usage of computer application during the identification of breast cancer has increased significantly. Mammography is being used as an effective technique for early identification of breast masses. The computerized technique has shown an effective mechanism which allows radiologist with the second judgment of diagnosis of cancer in the mammogram. In this manuscript, we have used a serial of image preprocessing and segmentation approaches, including 2D adaptive median filtering, thresholding algorithm, to remove noise from a digitalized mammogram. We have also developed an entire image statistical feature based classification approach, with the combination of a GLCM technique to extract six fundamental texture attributes from an image. Finally, Feed Forward Artificial Neural Network (FFANN) is being used to classify a mammogram as normal, benign and malignant. The usefulness of this manuscript is examined on MIAS databank along with classification sensitivity, specificity, and accuracy. The performance of Computer aided diagnosis (CAD) has been evaluated through confusion matrix. The final outcome shows that the presented technique has achieved 98% accuracy, 97% sensitivity, and 97% specificity.