Using AI to Identify Challenging Ambiguous Spitzoid Lesions

The use of artificial intelligence (AI) in dermatology is becoming increasingly prevalent and advancing rapidly. Current research has demonstrated its great potential in identifying early signs of skin cancer and detecting melanoma. Atypical Spitz nevus can appear similar to melanoma, making differentiation between the two lesions challenging. Because of the potential to be malignant, these skin lesions may be excised in ambiguous cases as a precautionary measure. In children, this often requires removal in the operating room under anesthesia. Because of the associated risks with such a procedure, there is a significant need for improved diagnostic tools to accurately differentiate atypical Spitz nevi from melanoma.

The objective of our project is to create a comprehensive database that can be used to train an AI model capable of distinguishing between the features of atypical Spitz nevi and melanoma. We will collect images from a variety of sources including clinical, dermoscopic, and histopathologic that will be used as the foundation for training the model. The images will cover a spectrum of Spitz nevi, primarily focusing on atypical, as well as representation in patients with varying skin pigmentation. The images will subsequently be annotated by research team members to emphasize important distinguishing features. Photos will also be processed to ensure that there is consistent quality that will meet the training standards. This model can then be validated and compared to dermatologist evaluation to assess its accuracy.

If the model is successful, this tool may help reduce the number of unnecessary excision and anesthesia exposure in children, improving the safety and quality of care for these patients. Overall, this project would advance pediatric dermatology research by exploring the utility of AI as a non-invasive diagnostic tool which could have long-term benefits in minimizing healthcare burdens and improving patient outcomes.