Background: Melanoma can mimic other cutaneous lesions, but the full spectrum and prevalence of these morphologic variants remain largely unknown.
Objective: To classify nonacral cutaneous melanomas into distinct morphologic clusters and characterize clusters' clinicopathologic features.
Methods: All pathologic melanoma diagnoses (occurring during 2011-2016) were reviewed for routine prebiopsy digital photographs (n = 400). Six dermatologists independently assigned lesions into 1 of 14 diagnostic classes on the basis of morphology. Image consensus clusters were generated by K-means; clinicopathologic features were compared with analysis of variance and χ2.
Results: Five morphologic clusters were identified: typical (n = 136), nevus-like (n = 81), amelanotic/nonmelanoma skin cancer (NMSC)-like (n = 70), seborrheic keratosis (SK)-like (n = 68), and lentigo/lentigo maligna (LM)-like (n = 45) melanomas. Nevus-like melanomas were found in younger patients. Nevus-like and lentigo/LM-like melanomas tended to be thinner and more likely identified on routine dermatologic examinations. NMSC-like melanomas were tender, thicker, more mitotically active, and associated with prior NMSC. Typical and SK-like melanomas had similar clinicopathologic features.
Limitations: Cluster subdivision yielded diminished sample sizes. Visual assignment was performed without clinical context.
Conclusion: When primary cutaneous melanomas were assigned into diagnostic groups and subjected to novel consensus clustering, recurrent morphologic patterns emerged. The spectrum of these morphologies was unexpectedly diverse, which might have implications for visual training and possibly clinical diagnosis.
Keywords: diagnostic accuracy; early diagnosis; education; melanoma; melanoma appearance; melanoma mimics; morphology; nevi; pigmented lesions; unusual-appearing melanoma.
Copyright © 2018 American Academy of Dermatology, Inc. Published by Elsevier Inc. All rights reserved.