Abstract #2222
Texture and Regression Tree Analysis in the Characterisation of Ovarian Lesions
Peter Gibbs 1 , Martine Dujardin 1 , and Lindsay Turnbull 1
1
MRI Centre, HYMS at University of Hull,
Hull, East Yorkshire, United Kingdom
MRI is the preferred technique for characterising
complex adnexal masses. However, the presence of solid
components in both benign and malignant lesions causes
diagnostic difficulties. In this work the utility of
co-occurrence matrix based textural analysis in the
diagnosis of ovarian malignancy is explored. Significant
differences between four groups (ovarian cancer,
borderline ovarian tumour, cystadenoma and
cystadenofibroma) were found for 8 of 16 calculated
texture parameters. Regression tree analysis yielded a
robust diagnostic model, based on 3 texture parameters,
with an overall accuracy of 70%.
This abstract and the presentation materials are available to members only;
a login is required.
Join Here
