AI in dermatology

Over the last years, artificial Intelligence (AI) is steadily transforming the healthcare landscape and dermatology has been at the forefront of this evolution. During a plenary lecture at EADV 2024, Prof Dr H. Peter Soyer, dermatologist at the University of Queensland in Brisbane (Australia) provided an overview of the current state of affairs of AI in dermatology and assessed how this technology can further reshape dermatology care in the years to come.

A first important application of AI in dermatology concerns cognitive AI tools that can help in diagnosing skin cancer. Over the last years several AI models have been developed that can identify melanoma, squamous cell carcinoma, basal cell carcinoma and benign skin lesions with a sensitivity and specificity that often exceeds that of traditional dermatological evaluations.1 For the moment, however, the use of these models is largely restricted to a research setting. An important challenge that needs to be tackled when these models enter the daily clinical practice relates to faulty AI. In fact, we as humans have a tendency to think that AI models are always right. However, this is not the case. These faulty AI interpretations can cloud the interpretation of dermatologists leading to a misdiagnosis. As such, AI has the inherent potential to destroy clinical expertise. To mitigate this risk, the dermatology society needs to continue investing in a solid education of dermatologists in training while in the meantime embracing AI.

In addition to cognitive AI, several tools have been developed that can help to improve the efficiency of dermatology care. A prime example of this consists of AI scribe, which automatically takes notes during a consultation. With such a tool, the dermatologist can give his/her full attention to the patient, without having to think about taking notes. In a second step, large language models can be used to transform these notes into a structured medical report. A third and final example that was addressed by Prof Soyer consists of medical chatbots. In these models, humanoids can answer simple medical questions from patients and provide them with suitable advice.

As such, it is clear that AI will become part of the routine dermatology care in the years to come. In this respect, however, it will be important to involve dermatologists in the development of the different AI models. In fact, when the development of these models is left to computer specialists and mathematicians, they will lack the dermatological expertise and nuance that will be essential for their applicability in daily practice.