A new skin cancer diagnosis method based on deep learning

Nowadays, the desire for research in the field of using intelligent algorithms in the diagnosis and classification of diseases, especially cancer, has increased greatly. Software calculation methods are very important in the diagnosis of medical diseases due to their classification performance. Diagnosis and classification of medical images is a challenging task. Deep networks are a type of learning methods that have the ability to model high-level relationships in data.One of the most widely used types of deep networks are convolutional or convolutional networks, which have increased efficiency. In this article, the performance of a unified model using convolutional and rosenet networks has been investigated, which was chosen due to the importance of diagnosing skin lesions in medicine, the complexity of the images, their large number, and the unbalanced categories. In order to better extract the diversity in skin lesions, changes were made in the initial layers of the network, and due to the imbalance in the mentioned dataset, changes were made in the cost function of the network, and the effect of using different activation functions in the network was also considered. Was investigated. The obtained results show that in this idea, by making appropriate adjustments, it can also be used on complex data sets.

Keywords:

CNN , Resnet , skin cancer , deep learning.