The way we live and work today is being significantly changed by digital technologies, especially by modern AI-based services in practice. In particular, the current COVID-19 pandemic has proven the importance of digital technologies and computational services to positively support people in difficult situations. This is revolutionizing healthcare towards autonomous and intelligent diagnosis and patient care for various diseases. The field of dermatology, and in particular skin cancer detection, is one area that can benefit significantly from the use of AI, especially computer vision and image recognition for skin cancer detection.
According to the experiments of the European Society of Oncologists, the goodness of AI in detecting skin cancer is 93% compared to dermatological oncologists. Despite this fact, people still do not trust AI systems and would rather choose human experts than computerized systems for their medical treatments. The main reason for this is the inability of AI systems to explain their decisions. Almost all decisions in AI-based services are based on probability methods. The main problems to be solved in this subproject are the following (1) Incorrect diagnoses and treatments by human experts, (2) The lack of autonomous and intelligent skin cancer diagnosis using AI, (3) The lack of a decision support system for medical experts that provides recommendations for medical treatments of skin cancer using AI, (4) The lack of explanations for AI-based decisions in skin cancer detection.
The main innovations of this subproject related to the problems mentioned above are the efficient use of AI for the intelligent early diagnosis of skin cancer, the support of urgent cases by providing a real-time communication channel with dermatologists (teledermatology), the provision of explanations of the decisions made by AI systems and autonomous patient care, and the tracking of conspicuous skin areas with regard to potential future skin cancer. Through the innovations envisaged in this subproject, medical experts and especially newcomers will be able to benefit from such an AI-supported system in their decision-making, and patients will also be able to care for their skin more attentively and diagnose their potential skin disease early enough to avoid massive side effects of too late diagnosis.
The main scientific methods focused on in this subproject are the use of (1) generative algorithms to enrich the training datasets by generating more skin cancer images, (2) deep learning algorithms (CNN) for melanoma detection, (3) a cloud-based solution for real-time data exchange of urgent and high-risk cases with dermatologists, and secure communication channels between dermatologists in the form of international tumor boards and between dermatologists and patients in the form of teledermatology. In general, this subproject is an essential part of the overall project concept, which is based on the use of AI for decision support and patient care in skin cancers. The results can be used by dermatologists for their daily decision support as well as by patients for their suspicious skin sites. Furthermore, the algorithms developed in this subproject will be used by scientific communities for further diagnostic solutions in the future.