Computer-automated ABCD versus dermatologists with different degrees of experience in dermoscopy

Abstract

Dermoscopy is a very useful and non-invasive technique for in vivo observation and preoperative diagnosis of pigmented skin lesions (PSLs) inasmuch as it enables analysis of surface and subsurface structures that are not discernible to the naked eye. The authors used the ABCD rule of dermoscopy to test the accuracy of melanoma diagnosis with respect to a panel of 165 PSLs and the intra- and inter-observer diagnostic agreement obtained between three dermatologists with different degrees of experience, one General Practitioner and a DDA for computer-assisted diagnosis (Nevuscreen(®), Arkè s.a.s., Avezzano, Italy). 165 Pigmented Skin Lesions from 165 patients were selected. Histopathological examination revealed 132 benign melanocytic skin lesions and 33 melanomas. The kappa statistic, sensitivity, specificity and predictive positive and negative values were calculated to measure agreement between all the human observers and in comparison with the automated DDA. Our results revealed poor reproducibility of the semi-quantitative algorithm devised by Stolz et al. independently of observers' experience in dermoscopy. Nevuscreen(®) (Arkè s.a.s., Avezzano, Italy) proved to be 'user friendly' to all observers, thus enabling a more critical evaluation of each lesion and representing a helpful tool for clinicians without significant experience in dermoscopy in improving and achieving more accurate diagnosis of PSLs.