Open AccessDigital videomicroscopy with image analysis and automatic classification as an aid for diagnosis of Spitz nevus
Author(s) -
Pellacani Giovanni,
Martini Mia,
Seidenari Stefania
Publication year - 1999
Publication title -
skin research and technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.521
H-Index - 69
eISSN - 1600-0846
pISSN - 0909-752X
DOI - 10.1111/j.1600-0846.1999.tb00140.x
Subject(s) - spitz nevus , magnification , nevus , computer aided diagnosis , medicine , classifier (uml) , artificial intelligence , linear discriminant analysis , digital image , dermatology , computer science , pattern recognition (psychology) , melanoma , image processing , image (mathematics) , cancer research
Aims: The aims of the study were: 1) to evaluate features that refer to digital images of ESC (epithelioid and/or spindle cell) nevi recorded by means of a videomicroscope and to compare them to those that refer to nevi and melanomas; 2) to assess the efficacy of an automatic classifier for the differentiation between ESC nevi and melanomas. Methods: Prior to biopsy, 29 ESC nevus images were recorded by means of a digital videomicroscope. All digital images were analyzed with software that calculates different parameters that relate to the geometry, color distribution and internal pattern of the lesion. Digital parameters that refer to ESC nevi and other PSLs (pigmented skin lesions: nevi and melanomas) were compared. Values that refer to parameters calculated on ESC nevi and melanomas acquired with the same magnification underwent elaboration by means of multivariate discriminant analysis, enabling a distinction among the groups. Finally, the diagnosis obtained by the automatic classifier was compared to the ones obtained by the naked eye and by videomicroscopic observation. Results: Significant differences between values that refer to ESC nevi and other PSL (nevi and melanomas) were noticed for most digital parameters. Automatic classification enabled the distinction among ESC nevi and melanomas with a 100% sensitivity and a 100% and 86% specificity employing the 50‐ and 20‐fold magnification, respectively. Conclusion: Digital analysis of videomicroscopic images by means of dedicated software enables an objective characterization of morphological features. Values obtained by image analysis may be immediately examined by the automatic classifier, in order to provide an aid to clinical diagnosis.