Development of computer-aided detection of breast lesion using gabor-wavelet BASED features in mammographic images

Here the problem of breast lesions detection using spatiotemporal based features is addressed. As the breast cancer is one of the most crucial reasons for death of women, earlier finding such lesions can increase the life of patients and more efficient treatment. Previously, Gabor wavelet has been introduced for spatiotemporal feature methods. In this approach, a multi-channel Gabor wavelet filter bank applied to the mammography image along with wavelet fusion made feature set. A Bayesian classifier classifies the final features into two different classes, normal and lesion class, with considering sparseness in features. Morphological operation with various structure elements of the mammography images has a crucial role in pre-processing stage of classification. The benchmark has been done utilizing 40 cases of Digital Database for Screening Mammography (DDSM) dataset. Training stage of classification performs by training map comprising 5 cases, normal and abnormal lesion cases. Proposed approach achieves accuracy of 98.75 percent which is relatively comparable with state-of-the-art methods(i.e.[10]).