Texture Analysis in the Characterisation of Ovarian Lesions: Use of Synthetic Minority Oversampling
Peter Gibbs 1 , Martine Dujardin 1 , and Lindsay Turnbull 1
Centre for MR Investigations, University of
Hull, Hull, East Yorkshire, United Kingdom
T2 weighted images of ovarian lesions have been assessed
using texture analysis. Uneven group sizes, which can
cause bias in statistical models leading to over-fitting
of the majority class, were avoided by employing
synthetic minority oversampling methods. Significant
differences between groups were noted for 12/16 texture
parameters and a multinomial logistic regression model
utilised 4 parameters (f2, f14, f15, f16) in achieving a
diagnostic accuracy of 71% for a training dataset and
61% for a test dataset.
This abstract and the presentation materials are available to members only;
a login is required.