Algorithm for the Diagnosis of Melanocytic Lesions
Geoffrey J. Gottlieb
The Ackerman Academy of Dermatopathology
New York, NY
Problems in constructing algorithms for natural/pathologic conditions are twofold. First, it is an
artificial, man-made attempt to force that phenomenon into a discrete schema. Secondly, even if a
reasonable algorithm is attainable, there is often a wide variability in the perception and application
of the criteria. For example, in my consultation practice, submitted sections will come with a
description stating that there is no evidence of maturation of melanocytes with progressive descent into
the dermis whereas, to my eye, there clearly is maturation.
With these caveats aside, this is a first attempt to construct an algorithm for the diagnosis of
melanocytic lesions. Rather than including all types of melanocytic neoplasms, benign or malignant, this
first attempt centers on the fundamental question in melanocytic neoplasia, namely, melanoma or not.
Even when limited to the benign vs. malignant question, the construction of an algorithm is difficult
since, in the difficult cases, a melanoma can have more features in common with a nevus than a prototypic
melanoma. It is those lesions which are the real problem in diagnosis. With that in mind, instead of a
large, branching and complicated algorithm, I have attempted to construct a simple one which
progressively leads to the diagnosis of a melanoma but acknowledges the exceptions which occur along the
way. To state it differently, the diagnosis of an "obvious" melanoma is easy and the criteria quite
straightforward. Not all melanomas, however, fulfill all the criteria and, for that reason, this is an
algorithm of exceptions.
In this most simplified form, the algorithm will attempt to determine "melanoma or not." In the real
world, there are only two other diagnoses, namely, melanoma with nevus and "I don't know."
Algorithm: Melanoma or Not
The majority of melanomas are relatively easy to identify since what I consider to be the most
important criteria are usually met. Those criteria include large size, asymmetry, poor circumscription
and, at least focally, the absence of maturation of melanocytes with progressive descent into the dermis
(if there is a dermal component). In the simple algorithm below, as one progresses to the bottom of it
(along the thickened line/arrow), the likelihood of melanoma becomes more likely and the number of
The exceptions are usually the "problem" cases and the ones seen in consultation. It is recognizing
the exceptions and the criteria for evaluating them that is the most important aspect of the exercise.
At each starred, rectangular box, the additional features to be considered in order to decide between
nevus and melanoma are listed below.
Criteria for Exceptions:
|Any of the major criteria (Symmetry, Circumscription, Maturation) which are not addressed if the melanocytic lesion "branches" off the algorithm near the top|
|Type of Nevus|
|Evidence of Prior Trauma|
|"Too Many Cells"|
|Variation in the Sizes, Shapes, Spacing, and Tendency to Confluence of Nests of Melanocytes|
|Melanocytes Singly and/or in Collections above the Dermo-epidermal Junction|
|"Atypia" of Melanocytes|
|Mitoses in Dermal Melanocytes|
I struggled with a way to include these criteria into an algorithm, however, it is basically not
possible. All of the criteria must be considered as a whole because, depending upon the precise lesion,
one or more criteria can diminish or negate the importance of others. For example, if the lesion under
consideration is thought possibly to be a Spitz's nevus, then melanocytes above the dermo-epidermal
junction and the presence of dermal mitoses can be seen in Spitz's nevi and are, therefore, not helpful
in differentiating from melanoma. In some instances, the presence of only one of the criteria "trumps"
all of the others in confirming the diagnosis of melanoma. One such criterion is "too many cells."