Year: 2024
Type: article
Abstract: Artificial intelligence (AI) has demonstrated remarkable potential in the classification of skin lesions for melanoma screening, with algorithms achieving diagnostic accuracy comparable to that of expert dermatologists in controlled settings.1 However, there remains a considerable gap between the development of these AI systems and their integration into clinical practice.Whether patients and clinicians truly trust and accept these tools for real-world use has been largely unexplored.The study by Goessinger et al. 2 stands out as one of the few real-world evaluations of both patient and dermatologist perspectives on the use of AI in skin cancer screenings.The authors conducted a direct comparison of three skin cancer examination types: (1) traditional clinical examination performed by a dermatologist, (2) augmented intelligence (AuI) and (3) a solely AI-driven approach.Notably highlighted in this study is a strong patient preference for AuI-a hybrid method combining human expertise and AI-over the other two approaches.Similarly, dermatologists in this study favoured AuI; however, those with more years of experience were more critical of AI classifications.In some cases, dermatologists opted to override AI classifications when they conflicted with their own personal clinical judgement, despite high confidence in AI.These findings highlight the need to further explore human-AI configuration in healthcare, particularly in balancing AI's strengths with human judgement to address concerns such as overreliance, misdiagnosis and liability.Although dermatologists in the study were provided with AI-generated melanoma probability scores, 2 further research is needed to evaluate methods that enhance physician and patient trust.It is unclear whether additional interpretability tools, such as saliency maps, or training dataset transparency measures, such as data nutrition labels or model cards, could be beneficial.3 This area of research warrants
DOI: 10.1111/jdv.20376
Authors: Yingjoy Li, Veronica Rotemberg
Institution: Memorial Sloan Kettering Cancer Center
Language: English
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4404701631