•  
  •  
 

Lynchburg Journal of Medical Science

Lynchburg Journal of Medical Science

Specialty

Urgent Care

Advisor

Thomas Colletti

Abstract

Purpose: The purpose is to review whether artificial intelligence can improve disparities in diagnosis of melanoma among people of color.

Method: PubMed, Google Scholar, and Cochrane literature searches were conducted with the search terms “artificial intelligence”, “machine learning”, “deep learning”, “smartphone”, “skin cancer”, “melanoma”, “diagnosis”, “skin tone”, “racial disparities”, and “people of color”.

Results: Multiple high-quality studies show that artificial intelligence consistently performs better than primary care providers and dermatologists in identifying cancer skin lesions. There were no studies specifically comparing black and non-black patients. Three low-quality studies found that artificial intelligence algorithms had equivalent or better accuracy in diagnosing melanoma than dermatologists in both light and dark skin tones. One study revealed 100% accuracy when combing an algorithm with a panel of dermatologists. Three systemic reviews revealed there is insufficient evidence at this time to recommend artificial intelligence for routine skin cancer screening in primary care.

Conclusion: Artificial intelligence shows promise that it can be used by primary care providers to assist triaging high-risk skin lesions on all skin tones if pitfalls such as algorithmic bias are avoided. However, more high-quality research is needed to assess outcomes of melanoma in people of color when using artificial intelligence as a diagnostic tool.

Keywords: artificial intelligence, skin cancer, melanoma, people of color, racial disparities.

Restricted

Available when accessing via a campus IP address or logged in with a University of Lynchburg email address.

Off-campus users can also use 'Off-campus Download' button above for access.

Share

COinS