A Skin lesion classification from dermoscopic images using deep learning techniques (T-0002) (Old 8652)

Abstract

The recent emergence of deep learning methods in medical image analysis field has enabled the development of intelligent medical imaging based diagnosis systems that can assist physicians in making bener decisions about a patient's health. In particular. skin imaging is a field where these new methods can be applied with a high rate of success. Skin cancer diseases bring silent death due to ability of spreading to other parts of the body. Small variations in lesion pattern can cause mixing of non-melanoma with melanoma lesions. Early detection of these diseases can reduce the level of disease severity. The framework consist on consecutive steps : pre-processing . segmentation and classification. We have evaluated classification model on segmented and original skin lesion classification datasets. We have trained CNN via transfer learn ing in three differem manners: I) Fine tuned convolmional neural network. 2) Train as a classifier and 3) Train as a fixed feature extractor. We found the best results 0.95 accuracy, 0.43 loss, 0.94 sensitivity and 0.98 precision using ResNer 50 on segmented skin lesion dataset on I 00 epochs