'DETECTION OF MALIGNANT MELANOMA 
    USING NEURAL CLASSIFIERS'

Authored by:

  Mads Hintz-Madsen, Lars Kai Hansen, and Jan Larsen
  Dept. of Mathematical Modelling
  Technical University of Denmark
  DK-2800 Lyngby

  Eric Olesen, and K.T. Drzewiecki
  Dept. of Reconstructive Surgery
  The National Hospital of Denmark
  DK-2100 Copenhagen

To appear in the EANN'96 conference proceedings.


                ABSTRACT

In this paper we propose a method for design 
of feed-forward neural classifiers based on 
regularization and adaptive architectures, 
and we apply the scheme to the problem of 
detecting malignant melanoma. Using features 
acquired from color photographs describing 
color and texture properties of skin tumors,
we are able to detect 76.0% +/- 7.8% of 
melanoma cases in a test set.