Prediction of recurrence in thin melanoma using trees and random forests

UNCW Author/Contributor (non-UNCW co-authors, if there are any, appear on document)

Richard M. Reiter (Creator)

Institution

The University of North Carolina Wilmington (UNCW )

Web Site: http://library.uncw.edu/

Advisor

Dargan Frierson

Abstract: In this paper, we will try to predict the recurrence of melanoma in the subset of patients with a “thin melanoma,” defined by having a Breslow Thickness less than 1.00mm. The study used the 1610 patient data base from the Duke Melanoma Clinic. The methods used to try classifying these patients as to whether or not they will suffer a recurrence are based on the statistical processes of trees and random forests introduced by Dr. L. Breiman. Successful results were achieved only when all patients were studied including those with an advanced stage of disease, but who still qualified as patients with a thin melanoma by histological measurement. The variables collected on each patient and used in this study related to the Breslow Thickness, Clark level, initial stage, age, primary site of lesion, local pathological events, histological type, sex, race and previous immunotherapy. The factors other than the initial stage and extent of disease proved too weak to give any meaningful results when only patients with local disease were studied. The results using a single tree and a random forest were compared. A detailed discussion of trees and random forest with use in the R-2.01 CRAN software packages is included. Although the results of this study did not provide the answers for this specific set of melanoma patients, it is felt that the techniques and programs written would be applicable in a differently defined set of patients. The results could be an important step in determining which patients should receive therapy and then evaluating the results of the adjuvant therapy on a more properly defined group of patients.

Prediction of recurrence in thin melanoma using trees and random forests
PDF (Portable Document Format)
226 KB
Created on 1/1/2009
Views: 1600

Additional Information

Publication

Thesis

A Thesis Submitted to the University of North Carolina Wilmington in Partial Fulfillment Of the Requirements for the Degree of Master of Science

Language: English

Date: 2009

Keywords

Melanoma--Research, Prediction theory, Probabilities, Random graphs, Trees (Graph theory)

Subjects

Prediction theory

Random graphs

Trees (Graph theory)

Probabilities

Melanoma -- Research

Email this document to