METHODS FOR PREDICTING SPINAL ARTHRODESIS

Authors

  • Leandro Fabian Escobar UFPR
  • Deborah Ribeiro Carvalho PUC-PR
  • Egon Walter Wildauer UFPR

Keywords:

Data mining, predictign models, Decision trees, Linear regression, Spinal Arthrodesis

Abstract

This article presents the construction of models for predicting the need for Spinal Arthrodesis (fusion) based on Multiple Linear Regression and Decision Tree, and compares the outcomes and accuracy. The experiments involved 8,016 orthopedic patients, of which 34 underwent Spinal Arthrodesis (fusion) The Multiple Linear Regression Equation was based on Microsoft Excel and the Decision Tree on algorithm J48. The respective accuracies were compared and the understanding of the results was assessed by orthopedic specialists. Linear Regression and Decision Tree showed sensitivity of 77.8% and 60.0%, specificity of 99.4% and 99.7% and areas under the ROC curve of 73.5% and 74%, respectively. The Decision Tree, in turn, was said to be more understandable by the specialists.

Downloads

Published

2013-12-16

Issue

Section

Articles