Computer Vision and Fractals: a Possible Aid for the Dermatologist in Recognizing Skin Tumors?

Fabrizio Guarneri


Early identification of skin tumors decisively influences patient survival. Symmetry, size, and regularity of borders and color patterns clinically suggest the benign or malignant nature of pigmented skin lesions. Recently, videomicroscopy of cutaneous lesions has provided high resolution digital images that can be electronically processed, but results of image analysis have been variable, probably also because of lack of standardization. In this study, Intel’s Open Computer Vision function library was used to standardize shape recognition, and fractal dimension to assess the regularity of contours; diagnostic performance of human operators on videomicroscopic images and images showing only the contours of the lesions was also tested. Results of this pilot study show that contour regularity alone is probably not decisive for diagnosis, and suggest the need for larger casuistics and numeric assessment of the other relevant diagnostic parameters. Improvement of computer vision algorithms appears also necessary for future use of image analysis in ordinary clinical conditions.

[DOI: 10.1685/CSC09268] About DOI


Computer vision; fractals; skin tumors.

Full Text:



Creative Commons License   Except where otherwise noted, content on this site is
  licensed under a Creative Commons 2.5 Italy License