Diagnostic performance of the MelaFind device in a real-life clinical setting

Abstract

Background: MelaFind is a multispectral computer vision system intended to -provide additional information on melanocytic lesions suspected of being melanoma by -objectively assessing their three-dimensional morphology.

Objectives: Analysis of the diagnostic performance of MelaFind in a real-life clinical setting.

Patients and methods: In this observational study, 360 pigmented skin lesions (PSL) in 111 patients were assessed by office-based dermatologists using MelaFind. Scores ≥ 2 were considered to be suspicious of malignancy. The decision for surgical excision was left to the discretion of the examining dermatologists.

Results: MelaFind scores ≥ 2 were observed in 147 of 360 PSL (40.8 %). Of the 107 excised lesions with a MelaFind-score ≥ 2, the diagnosis of melanoma was made in three cases; 53 (49.5 %) lesions proved to be dysplastic nevi. Among all lesions biopsied (n = 113), the sensitivity and specificity of MelaFind was 100 % and 5.5 %, respectively. While a higher specificity of 68.5 % may be assumed with respect to the overall data set (n = 360), this assumption is limited by incomplete follow-up data required to confirm that all non-excised lesions with a score < 2 were actually benign.

Conclusion: The high sensitivity of MelaFind facilitated the detection of melanoma. The overall specificity and benign-to-malignant ratio of excised lesions were acceptable. These parameters may be improved by using higher cutoff scores for excisional biopsies, and by more vigorously selecting PSL for MelaFind examination.