Deep learning architecture using transfer learning for classification of skin lesions

Skin cancer is one of the most dangerous health problems in many countries the development in automatic medical image analysis technique leads to accurate classification of diseases. Deep learning is a recent technology which solves the complexity in diagnosing the skin cancer. Due to high dissimilarity of skin lesion, it is challenging to classify the Image automatically. By using Convolutional Neural Network the classification of skin lesions can be done with high accuracy. The proposed models was developed by transfer learning concept with three popular architectures: inception V3, VGG16 and VGG19. Our models were trained by ISIC dataset that contain 2487 train images and 604 test images of seven skin lesion classes. Our models give the best performance of test data with 74%, 77% and 76% accuracy for Inception V3, VGG16 and VGG19.

Automated summary

The study utilizes deep learning technology and Convolutional Neural Network to achieve high-accuracy automatic classification of skin lesions, employing the concept of transfer learning and three popular architectures, obtaining good performance in the testing data.