Artificial Intelligence in Dermatology: A Review of Literature and Application to Pediatric Dermatology

Abstract

Background: Artificial intelligence (AI) is increasingly investigated for use in dermatologic conditions. We review recent literature on AI, its potential application for pediatric dermatology, and its impact on the underserved community. Objective: To evaluate the current state of AI in dermatology and its application to pediatric patients. Methods: Literature search was performed in PubMed and Google Scholar using the following key terms in combination with "pediatric", and "dermatology": "artificial intelligence," "AI," "machine learning," "augmented intelligence," "neural network," and "deep learning". Results: Current research is based on images from adult databases, with minimal delineation of patient age. Most literature on AI and dermatologic conditions pertains to melanoma and non-melanoma skin cancers, reporting accuracy from 67-99%. Other commonly studied diseases include psoriasis, acne vulgaris, onychomycosis, and atopic dermatitis, having varying accuracy, sensitivity, and specificity. A recently developed AI algorithm for diagnosis of infantile hemangioma found 91.7% accuracy. AI may be a means to increase access to pediatric dermatologic care, yet challenges remain for its use in underserved communities. Conclusion: Literature on AI systems for dermatologic diseases continues to grow. Further research may tailor AI algorithms for pediatric patients and those of diverse skin color to decrease algorithm bias and increase diagnostic accuracy.