Digital image analysis of pigmented skin lesions. Early diagnosis of melanoma

Resumen

Abstract:

Objetive: The purpose of this observation-based study was to differentiate between malignant melanoma (MM) and other skin pigmented lesions. This differentiation was done automatically and objectively through computer-generated images not depending on the viewer.

Methods: A videomicroscope with x20 and x50 magnification lenses, as well as polarized light, was employed, with a computer with an ad hoc software for image analysis.
This computerized viewing system allowed for the study of the quantity and distribution of colors on the images of pigmented lesions by means of RGB and HIS models.
Fifty five images of pigmented lesions were included in the study. They were divided into two groups: MM (n=16) and non-MM (n= 39).

Results: As regards to the quantity of colors, 92.3% of the non-MM lesions presented 4 or less colors, except for the blue nevi, which displayed 5 colors.
In the MM group, 6 or more shades were detected in 88.2% of the lesions and 5 colors in the remaining 11.8%.
In the distribution of colors, the value taken for cut-off point in the variety of hue was 20, where100% of the MM images were found to be above 20, unlike the non-MM group, with the exception of the basal cell carcinoma.

Conclusion: By means of a computer vision system with digital processing of images we can quantify the number and distribution of co-lors and diagnose the difference between MM and other skin pigmented lesions. This can be done automatically, objectively, not depending on an operator

(Dermatol Argent 2008;14(3):200-206).

Keywords: malignant melanoma, computer vision, digital processing of images