Image segmentation with deep learning

In recent years, the occurence of artificial intelligence and deep learning methods for the analysis of medical images has led to the development of intelligent diagnostic systems that help doctors make better decisions about the patient's health. By means these new methods, high success rates are achieved in many field. One of them is the field of dermatology which deals with skin diseases and treatment. One of the most important diseases in dermatology is skin cancer. Skin cancer is a type of cancer that can be seen as fatal and early diagnosis is vital for this disease. In this thesis, a deep learning based approach is used to classify dermoscopic images with semantic segmentation and skin lesion detection. In the segmentation step, the lesion region is estimated on the skin images, and in the classification step, the images are classified as benign and malignant, and then classified as triple. The methods developed for classification and segmentation are novel with respect to hierarchical studies in the literature and have shown more successful results than some.