کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Melanoma Detection Using Multi-Color LBP-FPl and Optimized VGG16

نویسندگان :

Vida Esmaeili¹ Mahmood Mohassel Feghhi²

1- دانشکده مهندسی برق و کامپیوتر، دانشگاه تبریز، تبریز، ایران 2- دانشکده مهندسی برق و کامپیوتر، دانشگاه تبریز، تبریز، ایران

کلمات کلیدی :

Dermoscopy images،Melanoma،Multi-color LBP-FPl،Optimized VGG16

چکیده :

Melanoma is the most serious and deadly kind of skin cancer, which arises from a gene mutation in melanocytes. It can be overgrown and spread throughout the body. Thus, early and prompt detection is required to avoid adverse effects from this disease. However, the challenges of melanoma early diagnosis are widely known such as the similarity of nevus and melanoma, variation in skin lesion sizes, and color variety. In this paper, we propose combining the multi-color Local Binary Pattern on Five Planes (multi-color LBP-FPl) and optimized VGG16 for automatic melanoma detection. First, the lesion region in the dermoscopic image is detected using an Active Contour (AC) model. Second, the multi-color method splits the lesion area into three color channels. Then, five planes are applied to lesion blocks in various color channels to compute the multi-color LBP-FPl feature. Finally, the extracted features are fed to optimized VGG16. Simulation results show that melanoma has been automatically detected in 99% on the used datasets using our proposed method.