Paper Title
A Method to Enhance Skin Lesions Segmentation Based on a Combination of Deep Learning and Wavelet Scattering on Clinical Images

Abstract
Melanoma is a high-risk skin disease, of which many people die every year worldwide. Early detection of melanoma can help patients to recover more efficiently. Computer-aided diagnostic systems can aid doctors in diagnosing melanoma cases in the early stages. Detecting skin lesions is one of the most critical steps in the early diagnosis of melanoma. More accuracy in the detection of skin lesions leads to more accuracy in the final melanoma diagnosis. In this study, an ensemble method of a convolutional neural network and wavelet scattering is proposed to detect skin lesions. The proposed method first uses data augmentation techniques to boost the number of input data and then learns a convolutional neural network for image segmentation. The network output is combined with the wavelet scattering result. The proposed method has an accuracy of 98.9%, a sensitivity of 92.1% and a specificity of 99.5%, and is efficiently able to segment skin lesions. Keywords - Melanoma, Segmentation, Deep Learning, Wavelet Scattering