Effects of Sample Size in Classifier Design
IEEE Transactions on Pattern Analysis and Machine Intelligence
DCC '00 Proceedings of the Conference on Data Compression
A robust approach to image enhancement based on fuzzy logic
IEEE Transactions on Image Processing
Image coding using translation invariant wavelet transforms with symmetric extensions
IEEE Transactions on Image Processing
| Hi-index | 0.00 |
In this paper, a new method for classification of mammogram lesions is presented based on the lesion's boundary profiles. A fuzzy operator for contrast enhancement is first used to increase the image's contrast, followed by a thresholding algorithm to segment the lesion from the background. Based on the binary mask of the segmented region, wavelet analysis is used for edge-detection. To ensure that features will be robust to shifts, a shift-invariant wavelet transform (SIDWT) is employed. The joint probability density function is estimated from the wavelet domain, and features which are semi-rotation invariant are extracted. The extracted features were homogeneity and entropy, which are used to describe the complexity of the lesion's boundary. After exhaustive feature selection and feature reduction by PCA, a k-nn classifier is used with the Leave-One-Out Method (LOOM) for small database scenarios. Based on this system, the sensitivity was found to be: 94%, while the specificity was found to be: 63% (which results in an average classification rate of 78.5%). The high classification rates of the malignant lesions is a highly desirable feature, as such a system should not miss any cancerous lesions. The experimental results show that this method can enhance, detect and classify lesions in mammography images for computer-aided diagnosis.