期刊名称：International Journal of Image Processing (IJIP)
电子版ISSN：1985-2304
出版年度：2013
卷号：7
期号：3
页码：227-236
出版社：Computer Science Journals
摘要：This paper develops a new computerized vision of skin cancer prognosis based on symmetry and color matching for lesion pigments. Initially, the lesion/tumor edge is detected and segmented. Then, the symmetrization is computed for all images to isolate benign (mole) tumor. The even symmetry parameter is introduced here to improve the symmetrization computations. The suspicious images would be nominated into one of three categories: melanoma, Basal Cell Carcinoma (BCC), or Squamous Cell Carcinoma (SCC) tumor depending on the symmetrization and pigment-color matching score table. Two matching procedures have been developed for nominating the suspicious images. First procedure matches pigment values with artificial spectrums of Reddish, Yellowish, Brownish, and Blackish. The second procedure matches pigment values with true malignancy/benign pigment database. The results of two procedures are compared over 40 pre-classified images. With Mean Squared Error (MSE) value equals to 0.003, procedure#1 satisfied 80% true classification while 92.5% for procedure#2. These results could be improved if lesion segmentation and/or spectrums/pigment-database are increased.
关键词：Melanoma; SCC; BCC; Skin Cancer; Color Matching.
