Premium
Dermoscopic diagnosis by a trained clinician vs. a clinician with minimal dermoscopy training vs. computer‐aided diagnosis of 341 pigmented skin lesions: a comparative study
Author(s) -
Piccolo D.,
Ferrari A.,
Peris K.,
Daidone R.,
Ruggeri B.,
Chimenti S.
Publication year - 2002
Publication title -
british journal of dermatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.304
H-Index - 179
eISSN - 1365-2133
pISSN - 0007-0963
DOI - 10.1046/j.1365-2133.2002.04978.x
Subject(s) - medicine , medical diagnosis , melanoma diagnosis , cohen's kappa , false positive paradox , dermatology , melanoma , nevus , diagnostic accuracy , radiology , artificial intelligence , machine learning , computer science , cancer research
Summary Background In the last few years digital dermoscopy has been introduced as an additional tool to improve the clinical diagnosis of pigmented skin lesions. Objectives To evaluate the validity of digital dermoscopy by comparing the diagnoses of a dermatologist experienced in dermoscopy (5 years of experience) with those of a clinician with minimal training in this field, and then comparing these results with those obtained using computer‐aided diagnoses. Methods Three hundred and forty‐one pigmented melanocytic and non‐melanocytic skin lesions were included. All lesions were surgically excised and histopathologically examined. Digital dermoscopic images of all lesions were framed and analysed using software based on a trained artificial neural network. Cohen's κ statistic was calculated to assess the validity with regard to the correct diagnoses of melanoma and non‐melanoma. Results Sensitivity was high for the experienced dermatologist and the computer (92%) and lower for the inexperienced clinician (69%). Specificity of the diagnosis by the experienced dermatologist was higher (99%) than that of the inexperienced clinician (94%) and the computer assessment (74%). Notably, computer analysis gave a higher number of false positives (26%) compared with the experienced dermatologist (0·6%) and the inexperienced clinician (5·5%). Conclusions Our results indicate that analysis either by a trained dermatologist or an artificial neural network‐trained computer can improve the diagnostic accuracy of melanoma compared with that of an inexperienced clinician and that the computer diagnosis might represent a useful tool for the screening of melanoma, particularly at centres not experienced in dermoscopy.