Seborrheic keratoses mimicking melanoma unveiled by in vivo reflectance confocal microscopy

Seborrheic keratoses (SebK) with atypical dermoscopy presentation are increasingly reported. These lesions do not exhibit typical dermoscopy features of SebK and sometimes mimic melanoma, thus complicating the differential diagnosis. Reflectance confocal microscopy (RCM) is a non‐invasive tool, which allows an in vivo imaging of the skin. The study objectives were to evaluate the agreement between RCM classification and histological diagnoses, and the reliability of well‐known RCM criteria for SebK in the identification of SebK with atypical dermoscopy presentation.

management of these atypical SebK remains undefined, often requiring a final excision of the lesion to reach a correct diagnosis. 7 To complicate the picture, melanomas mimicking SebK have also been reported. 9 Reflectance confocal microscopy (RCM) is a non-invasive technology, which allows in vivo imaging of the skin at a nearly histological resolution. RCM criteria for classic SebK were defined in 2012. 10 The aim of our study was to evaluate the reliability of the wellknown RCM criteria for SebK identification in a group of lesions with atypical dermoscopy presentation, mimicking melanoma.

| Study design and population
We retrospectively analysed all lesions with an initial clinical diagno- Lesions with at least one of the features of atypical melanocytic lesions at dermoscopy, as defined by the revisited seven-point checklist, 11 were selected. These features included: atypical pigment network, blue-whitish veil, atypical vascularization, peripheral streaks, regressive structures, irregular blotches and globules. Lesions of the face and lesions without a full set of investigations were excluded. Of the 1142 cases selected, those with known features for melanocytic lesions at RCM were excluded. [12][13][14][15][16][17][18][19][20][21][22] A total of 117 non-facial lesions without features for melanocytic proliferation were included ( Figure 1). All investigations were conducted according to the Declaration of Helsinki principles, with respect to human subjects in biomedical research.

| Instruments
Clinical, dermoscopic and RCM images stored in a dedicated database were previously acquired through a Canfield Nikon D90 Digital SLR ® , a Canfield Close-up Scale ® (Canfield Imaging Systems, Fairfield, NJ, USA) and a RCM laser scanning microscope (Vivascope 1500 ® ; MAVIG GmbH, Munich, Germany), respectively. Instruments and acquisition procedures have been described elsewhere. 12,13 Lesions were first evaluated according to dermoscopic inclusion criteria. Selected RCM images were evaluated by an expert dermatologist in dermoscopy and RCM analyses (F.F.) and lesion features were recorded. RCM evaluations were performed blinded to the histopathological diagnoses.

| Dermoscopic evaluation
All lesions were first investigated for the presence in dermoscopy of at least one of the well-known criteria for atypical melanocytic lesions proposed by the revisited seven-point checklist. 11 The following criteria were then evaluated: atypical pigment network, bluewhitish veil, atypical vascularization, peripheral streaks, regressive structures, irregular blotches or atypical globules. Each criterion was scored for its presence or absence. Cases without any criterion for atypical melanocytic proliferation were not included in the study.

| RCM evaluation
For each case, a complete set of at least three VivaBlock ® mosaic images, taken at three different standardized levels (epidermal layer, dermoepidermal junction and upper dermis), was available for evaluation.

| Histopathology evaluation
All cases underwent histopathology diagnoses. Lesions were grouped according to histological diagnoses as SebK (hSebK) or histologically proven non-SebK (hnSebK) and matching with RCM classifications and observed RCM features was performed.

| Statistical analysis
Descriptive statistics and complete case analysis was used for all comparisons between groups. Pearson's chi-squared test and Fisher's exact test were used to reveal associations between variables and groups.
T A B L E 1 Reflectance confocal microscopy (RCM) criteria for non-melanocytic lesions evaluated in the study. Lesions with histological diagnosis of SebK are classified as "hSebK" and those with other non-melanocytic histological diagnoses (i.e. dermatofibromas, basal cell carcinomas, squamous cell carcinomas and other diagnosis), are listed as "hnSebK." Data concerning the frequency of each RCM pattern in the two groups are reported in percentages. A brief description of distinct RCM features is also present, with references to the relevant literature (Ref Cohen's kappa was calculated to evaluate the agreement between RCM classification and histological diagnosis.
The interpretation of agreement adopted is: less than chance agree-

| Study population
The study consisted of 117 cases, with RCM features considered non-

| RCM criteria for SebK identification
All RCM features considered are listed in Table 1 The current study aimed to evaluate whether the use of RCM could improve the differential diagnosis of SebK with atypical dermoscopy presentation, mimicking melanoma, from melanocytic lesions. All cases included in the study showed at least one of the 7-point checklist criteria at dermoscopy for atypical melanocytic proliferation, thus leading to a possible dermoscopic misclassification. In this clinic however, patients with uncertain diagnosis are evaluated with RCM prior to histology, enabling an analysis of RCM features with microscopic outcomes, and the potential identification of specific features favourable to a differential diagnosis.
The current study confirmed that RCM features previously described for classic SebK, 10 were also present in SebK with atypical  Table 1, these patterns were significantly more frequently observed in hSebK, compared to hnSebK.
The resulting agreement between the RCM diagnosis and hSebK was high (97%), and suggestive of a good diagnostic accuracy.
Interestingly, the presentation of atypical dermoscopy SebK com- The main limit of the present study was the sample size. Only lesions with atypical dermoscopy presentation and RCM imaging, excised and sent for histopathological analysis, were selected and included.
Furthermore, as RCM was installed as a non-invasive analysis tool at our centre, many SebK with atypical dermoscopy presentation were assessed at RCM only and are not always sent for histopathological analysis.

| CONCLUSION
Reflectance confocal microscopy classification has a high agreement with histopathological diagnoses for SebK with atypical dermoscopy presentations. RCM features for the selected lesions were similar to those observed in typical SebK. Data suggest that RCM is an optimal non-invasive examination for early differential diagnosis of SebK with atypical dermoscopy presentations. Therefore, RCM may be able to assist in differential diagnosis and avoid unnecessary excisions.