Effects of Sample Size in Classifier Design
IEEE Transactions on Pattern Analysis and Machine Intelligence
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 and texture profiles. A fuzzy operator for image enhancement is first used to increase the image's contrast, followed by fuzzy thresholding. Three images are generated per lesion from the enhanced and segmented regions, including: 1) region-, 2) lesion- and 3) border-based information. A single base-line system is designed to analyze all three images, based on texture, or wavelet coefficient complicated-ness. To localize texture, a shift-invariant wavelet transform (SIDWT) is employed, to ensure that it is robust to shifts. Graylevel co-occurrence matrices are found for a variety of directions in the wavelet domain, and homogeneity and entropy were extracted (resulting in a shift, scale and semi-rotational invariant feature set). Exhaustive feature selection was used with both a k-nn and LDA classifier, to find the best classification performance. The over all highest classification performance was found for LDA, 72.5% (for border-based analysis). As the highest results were achieved for the lesion boundary analysis, it proves that the shape and complexity of the border is the most discriminatory feature for the computer-aided diagnosis of mammographic lesions.