Transactions on Machine Learning and Data Mining (ISSN: 1865-6781)

Volume 1 - Number 2 - October 2008 - Page 49-65

Experiences Using Clustering and Generalizations for Knowledge Discovery in Melanomas Domain

A. Fornells1, E. Armengol2, E. Golobardes1, S. Puig3, and J. Malvehy3

1Grup de Recerca en Sistemes Intel·ligents Enginyeria i Arquitectura La Salle, Universitat Ramon Llull, Spain
2IIIA - Artificial Intelligence Research Institute, CSIC - Spanish Council for Scientific Research, Campus UAB, Spain
3Melanoma Unit, Dermatology Department, IDIBAPS, U726 CIBERER, ISCIII, Hospital Clinic i Provincial de Barcelona, Spain


One of the main goals in prevention of cutaneous melanoma is early diagnosis and surgical excision. Dermatologists work in order to define the different skin lesion types based on dermatoscopic features to improve early detection. We propose a method called SOMEX with the aim of helping experts to improve the characterization of dermatoscopic melanoma types. SOMEX combines clustering and generalization to perform knowledge discovery. First, SOMEX uses Self-Organizing Maps to identify groups of similar melanoma. Second, SOMEX builds general descriptions of clusters applying the anti-unification concept. These descriptions can be interpreted as explanations of groups of melanomas. Experiments prove that explanations are very useful for experts to reconsider the characterization of melanoma classes.

PDFDownload Paper (3865 KB)

Back to Table of Contents