A Deep Multi-Task Learning Approach to Skin Lesion Classification

However, instead of treating the skin lesion classification Visual aspects of skin diseases, especially skin lesions, play as a standalone problem and training a CNN model a key role in dermatological diagnosis. A successful identification using skin lesion labels only, we further propose to jointly of the skin lesion allows skin disorders to be placed in optimize the skin lesion classification with a related auxiliary certain diagnostic categories where specific diagnosis can be task, body location classification. The motivation behind established (Cecil, Goldman, and Schafer 2012). However, this design is to make use of the body site predilection categorization of skin lesions is a challenging process. It of skin diseases (Cox and Coulson 2004) as it has long usually involves identifying the specific morphology, distribution, been recognized by dermatologists that many skin diseases color, shape and arrangement of lesions. When these and their corresponding skin lesions are correlated with their components are analyzed separately, the differentiation of body site manifestation. For example, a skin lesion caused skin lesions can be quite complex and requires a great deal by sun exposure is only present in sun-exposed areas of the of experience and expertise (Lawrence and Cox 2002).