An efficient biomedical imaging technique for automatic detection of abnormalities in digital mammograms

Many biomedical imaging systems provide significant contribution in aiding radiologists in interpreting digital mammograms and thus enhancing early detection of breast cancer. The motivation of this paper is to study the efficiency of existing systems and propose an efficient biomedical imaging technique which can detect abnormalities in digital mammograms. This proposed technique involves several image-processing stages combined with image preprocessing, optimum thresholding, extracting haralick features and classifying the extracted features. Preprocessing images involve 2D median filtering with Contrast-Limited Adaptive Histogram Equalization (CLAHE) techniques and optimum thresholding using Otsu's method. Feature extraction involves extracting 6 textural features using Gray Level Co-occurrence Matrices (GLCM) for four spatial orientations: 0°, 45°, 90° and 135°. Radial basis function neural (RBFN) networks were used to classify the mammograms in to two types: benign or malignant. The experimental results were tested using MIAS database of digital mammograms. The overall accuracy of the proposed technique is 99.70%. The results proved that the proposed technique outperformed other existing techniques in the aspects of sensitivity, specificity and accuracy.