Citation
Klebanov N, Gunasekera NS, Lin WM, Hawryluk EB, Miller DM, Reddy BY, Christman MP, Beaulieu D, Rajadurai S, Duncan LM, Sober AJ, Tsao H. Clinical spectrum of cutaneous melanoma morphology. J Am Acad Dermatol. 2019;80(1):178–188.e3. doi:10.1016/j.jaad.2018.08.028.
Why this study matters
Melanoma is often taught through a relatively narrow set of classic images, yet real-world tumors can closely resemble benign nevi, seborrheic keratoses, lentigines, actinic keratoses, basal cell carcinomas, or squamous cell carcinomas. That visual diversity can delay recognition by both patients and clinicians.
This study used routine prebiopsy photographs rather than a curated teaching collection, providing a less selected view of how melanoma actually presents in clinical practice.
Cutaneous melanoma does not have a single recognizable appearance. Five recurrent morphologic patterns emerged, and the tumors that looked least like conventional melanoma often carried clinically important differences in thickness, mitotic activity, symptoms, and mode of detection.
The study at a glance
400 melanomas
Consecutive pathologically confirmed, nonacral cutaneous melanomas diagnosed from 2011 through 2016 were included when routine prebiopsy digital photographs were available.
Six independent dermatologists
Each lesion was assigned to one of 14 diagnostic classes based on morphology alone, without access to the surrounding clinical context.
Consensus clustering
K-means clustering was applied to the dermatologists’ image classifications to identify recurrent patterns of visual agreement.
Clinicopathologic comparison
The resulting clusters were compared across patient characteristics, detection patterns, symptoms, tumor thickness, subtype, and mitotic activity.
Five recurrent morphologic clusters
Typical melanoma
Conventional pigmented lesions with clinicopathologic features broadly similar to the SK-like group.
Nevus-like
More common in younger patients and generally thinner at diagnosis.
Amelanotic/NMSC-like
Often pink, raised, tender, thicker, and more mitotically active.
SK-like
Keratotic or seborrheic-keratosis-like tumors resembling common benign lesions.
Lentigo/LM-like
Typically flatter, thinner, and more often identified during routine examination.
How morphology related to clinical behavior
- Nevus-like melanomas occurred in younger patients than the other morphologic groups.
- Nevus-like and lentigo/lentigo maligna–like melanomas tended to be thinner and more superficially invasive.
- These two groups were also more likely to be detected during routine dermatologic examinations.
- Amelanotic/NMSC-like melanomas were more often pink or erythematous, raised, tender, thick, and mitotically active.
- Amelanotic/NMSC-like tumors were associated with a prior history of nonmelanoma skin cancer.
- Typical and seborrheic keratosis–like melanomas had broadly similar clinicopathologic profiles.
The most clinically consequential melanomas may not advertise themselves through classic melanoma morphology. A pink, keratotic, scaly, nevus-like, or lentigo-like lesion can still represent melanoma, and the amelanotic/NMSC-like pattern may be associated with particularly aggressive pathologic features.
Why the image source matters
Many melanoma image collections are assembled after the diagnosis is known and may overrepresent lesions that already look suspicious. Here, photographs were captured during routine care before biopsy, reducing dependence on the clinician’s ultimate diagnostic impression and revealing morphologies that may be underrepresented in textbooks and teaching sets.
The unexpectedly common seborrheic keratosis–like and ill-defined erythematous or scaly patterns underscore how curated image libraries can narrow clinicians’ expectations of what melanoma looks like.
Interpretive considerations
- Dividing the cohort into five clusters reduced the sample size available for individual comparisons.
- Dermatologists classified photographs without clinical history, palpation, dermoscopy, or longitudinal change.
- The clusters describe recurrent visual patterns rather than mutually exclusive biologic melanoma subtypes.
- The study was designed to characterize morphology, not to establish the diagnostic accuracy of a specific classifier.
- External validation would be needed before using these clusters as a formal clinical prediction system.
Visual training in melanoma should deliberately include benign-appearing and nonclassic tumors—not only textbook examples. Expanding the recognized visual spectrum may improve diagnostic vigilance and reduce delays for melanomas that resemble more familiar benign or keratinocytic lesions.
Authors
Nikolai Klebanov, Nicole S. Gunasekera, William M. Lin, Elena B. Hawryluk, David M. Miller, Bobby Y. Reddy, Mitalee P. Christman, Derek Beaulieu, Suvithan Rajadurai, Lyn M. Duncan, Arthur J. Sober, and Hensin Tsao.