An updated algorithm integrated with patient data for the differentiation of atypical nevi from early melanomas: the idScore 2021

Abstract

Background: It is well known that multiple patient-related risk factors contribute to the development of cutaneous melanoma, including anagraphic, phenotypic and anamnestic factors.
 Objectives: We aimed to investigate which MM risk factors were relevant to be incorporated in a risk scoring-classifier based clinico-dermoscopic algorithm.
 Methods: This retrospective study was performed on a monocentric dataset of 374 atypical melanocytic skin lesions (aMSLs) sharing equivocal dermoscopic features, excised in the suspicion of malignancy. Dermoscopic standardized images of 258 atypical nevi and 116 early melanomas were collected along with objective lesional data (i.e., maximum diameter, specific body site and body area) and 7 dermoscopic data. All cases were combined with a series of 10 MM risk factors, including anagraphic (2), phenotypic (5) and anamnestic (3) ones.
 Results: According to the stepwise analysis logistic regression, the proposed iDScore 2021 algorithm is composed by 9 variables, including 3 patient data (age, skin phototype I/II, personal/familiar history of MM), 2 lesion characteristics (maximum diameter, location on the lower extremities (thighs/legs/ankles/back of the feet) and 4 dermoscopic features (irregular dots and globules, irregular streaks, blue gray peppering, blue white veil). The algorithm assigned to each lesion/case a progressive score from 0 to 18. The performance obtained in the ROC curve analysis reached 92% AUC in differentiating early melanomas from atypical nevi.
 Conclusions: An integrated checklist combining multiple anamnestic data with selected relevant dermoscopic features can be useful in the differential diagnosis and management of early melanomas and atypical nevi exhibiting equivocal features.

Dermoscopic predictors of melanoma in small diameter melanocytic lesions (mini‐melanoma): a retrospective multicentric study of 269 cases

“…This finding may be attributable to the idScore original dataset which comprised mainly lesions with a diameter exceeding 5 mm. 14 Concerning individual dermoscopic features, starburst pattern was documented in about a quarter of melanomas measuring ≤5 mm evaluated in this study, in line with previous reports. 16,17 However, it was neither the most prevalent global dermoscopic pattern in this group nor a significant predictor of malignancy when applied to the assessment of flat, non-facial melanocytic lesions measuring ≤5 mm.…”

“…Curiously, the idScore, an integrated clinical‐dermoscopic algorithm, performed similarly when applied to our cohort as compared with classic checklists employing only dermoscopic criteria. This finding may be attributable to the idScore original dataset which comprised mainly lesions with a diameter exceeding 5 mm 14 …”

“…The presence or absence of a set of predefined dermoscopic criteria was evaluated for each lesion. Selection of said criteria was the result of a consensus process among the authors and was based on the available literature 7–14 . Traditional melanoma‐specific as well as the recently published dermoscopic clues for melanoma in situ 15 were adopted.…”

A risk‐scoring model for the differential diagnosis of lentigo maligna and other atypical pigmented facial lesions of the face: The facial iDScore

“…These data were further processed to obtain a statistically significant dermoscopic signature for each lesion, a method that also helps avoid evaluator bias. [37][38][39][40] Fourthly, the model maintained satisfactory accuracy when tested over a new heterogeneous set of 86 difficult lesions with higher variability (i.e. 'validation set') confirming the robustness of the statistical method.…”

“…Thirdly, while other existing checklists were based on pattern analysis by very few selected experts, the present model relies on a pool of 34,244 dermoscopic data items obtained by 147 dermoscopists. These data were further processed to obtain a statistically significant dermoscopic signature for each lesion, a method that also helps avoid evaluator bias 37–40 . Fourthly, the model maintained satisfactory accuracy when tested over a new heterogeneous set of 86 difficult lesions with higher variability (i.e.…”