Premium
Validation of an integrated dermoscopic scoring method in an European teledermoscopy web platform: the iDScore project for early detection of melanoma
Author(s) -
Tognetti L.,
Cevenini G.,
Moscarella E.,
Cinotti E.,
Farnetani F.,
Lallas A.,
Tiodorovic D.,
Carrera C.,
Puig S.,
Perrot J.L.,
Longo C.,
Argenziano G.,
Pellacani G.,
Smargiassi E.,
Cataldo G.,
Cartocci A.,
Balistreri A.,
Rubegni P.
Publication year - 2020
Publication title -
journal of the european academy of dermatology and venereology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.655
H-Index - 107
eISSN - 1468-3083
pISSN - 0926-9959
DOI - 10.1111/jdv.15923
Subject(s) - teledermatology , medicine , concordance , checklist , classifier (uml) , artificial intelligence , diagnostic accuracy , medical physics , machine learning , melanoma diagnosis , melanoma , computer science , telemedicine , radiology , psychology , health care , cancer research , economics , cognitive psychology , economic growth
Background Although live and teledermoscopic examination has been successfully used to achieve non‐invasive diagnosis of melanocytic skin lesions ( MSL s), early melanoma ( EM ) and atypical nevi ( AN ) continue to be a challenge, and none of the various algorithms proposed have been sufficiently accurate. We designed a scoring classifier diagnostic method, the iDS core that combines clinical data of the patient with dermoscopic features of the MSL . Objective To test the accuracy of the iDS core in differentiating EM from AN in a teledermoscopy setting and to compare it with intuitive diagnosis, the ABCD rule and the seven‐point checklist. Materials and methods A dedicated teledermoscopy web platform was designed. This involved the following: (i) collecting a large integrated clinical–historical–dermoscopic data set of difficult MSL s from eight European dermatology centres; (ii) online testing, education and training in using the iDS core . A total of 904 images were combined with age, sex, lesion diameter and body site data and evaluated on the platform by 111 participants with four levels of skill in dermoscopy. Each testing session consisted of 30 blind cases to examine consecutively by the above four methods. ‘Management decisions’ and personal participant data were also recorded. Results iDS core‐ aided diagnosis achieved satisfactory diagnostic accuracy for all lesions, irrespective of centre of affiliation, showing an average AUC of 0.776 in all participant testing sessions. All skill groups improved their accuracy by 10–16% with respect to intuitive diagnosis and the other methods, showing high concordance and avoiding wrong management decisions. Conclusion We demonstrated the validity of the iDS core method for managing suspicious MSL s in a large multicentric data set and a teledermoscopic setting. The platform designed for the iDS core project provides ready support for physicians of any dermoscopy skill level and is useful for education and training.