Melanoma, the most lethal form of skin cancer, underscores the importance of early diagnosis for a positive prognosis. Detecting melanoma at its early stages is challenging, making timely medical care crucial. A study conducted by researchers at Dalhousie University, Canada, explored the viability of utilizing artificial intelligence (AI) to identify potential melanomas in self-referred patients expressing concerns about skin lesions.
Recruiting patients through hospital advertisements in Halifax, Nova Scotia, the study introduced a melanoma detection strategy based on computer-based learning, specifically deep learning. The FotoFinder Moleanalyzer Pro® Version 6.0, a sophisticated AI system, demonstrated capabilities equivalent to experienced dermatologists in assessing dermoscopic images for melanoma detection.
Initially, a medical student examined lesions meeting the study criteria, and those of concern were subsequently scanned using the FotoFinder System®. The proprietary computer software analyzed macroscopic and dermoscopic images. Three experienced dermatologists and a senior dermatology resident, unaware of AI results, evaluated the images. Suspicious lesions identified by the AI or any dermatologist underwent excision.
The study revealed 17 confirmed malignancies, including 10 melanomas. Notably, six melanomas not flagged by the AI exhibited ambiguous characteristics challenging even for experienced dermatologists and dermatopathologists. Interestingly, eight malignancies were observed in patients with a family history of melanoma. The AI system demonstrated comparable diagnostic ability to dermatologists in identifying malignancies through dermoscopic images.
While the AI system showcased technical and diagnostic limitations, its inclusion in melanoma screening programs, particularly for individuals concerned about specific lesions, holds promise. As a practical skin cancer screening aid, AI could contribute to providing timely access to skin cancer diagnoses, potentially enhancing early detection and improving patient outcomes. Further research and refinement of AI applications in dermatology may lead to more effective and widespread implementation in clinical settings.
