Computer-aided systems for melanoma - The objectives of this project are to develop automated computer systems to help interpret complex dermoscopic images with the ultimate aim of distributing the technique to non-dermatologist healthcare providers and, maybe, even patients.
Dermoscopy is a well-accepted technique for screening melanomas by dermatologists. The technique visualizes skin internal structures using a magnifying lens with special light or a drop of oil. Malignant and benign skin lesions can be differentiated by detecting signs of malignancy within the skin internal structures. However, there are a large number of signs, which are often subtle; thus, the technique needs special training even for experienced dermatologists. The steep learning curve prohibits the spread of the technique outside trained experts. My research team has been developing computer-aided diagnostic algorithms based on deep learning, machine learning and medical imaging analysis methods in order to build automatic systems for melanoma detection.