Meeting Banner
Abstract #4089

Improving Bladder Cancer Staging by using quantitative DCE-MRI with k-means clustering

Huyen T Nguyen 1 , Guang Jia 2 , Kamal S Pohar 3 , Amir Mortazavi 4 , Zarine K Shah 5 , Debra Zynger 6 , Lai Wei 7 , Xiangyu Yang 1 , Daniel Clark 1 , and Michael V Knopp 1

1 Wright Center of Innovation in Biomedical Imaging, Department of Radiology, The Ohio State University, Columbus, Ohio, United States, 2 Department of Physics and Astronomy, Louisiana State University, Baton Rouge, Louisiana, United States, 3 Deparment of Urology, The Ohio State University, Columbus, Ohio, United States, 4 Deparment of Internal Medicine, The Ohio State University, Columbus, Ohio, United States, 5 Deparment of Radiology, The Ohio State University, Columbus, Ohio, United States, 6 Deparment of Pathology, The Ohio State University, Columbus, Ohio, United States, 7 Center for Biostatistics, The Ohio State University, Columbus, Ohio, United States

This study is to evaluate the value of k-means clustering of DCE-MRI pharmacokinetic parameters in T staging of bladder tumors. k-means clustering was performed on the non-dimensionalized Amp and kep values of all twenty-four patients in the study to determine three cluster centers. The volume fractions (VFs) of three clusters were correlated with the tumor stage. Significant difference in the VF of cluster 2 was found between T1/lower vs. T2, T1/lower vs. T3, and T3 vs. T4. The differences in all three cluster VFs were also statistically significant. Fat-invasive tumors had significantly higher VFs of cluster 1 and 3 and a significantly lower VF of cluster 2 than did non-fat-invasive tumors. The VF of cluster 2 had area-under-the-curve (AUC) value of 0.83 in the differentiation of fat-invasive from non-fat-invasive tumors. k-means clustering of DCE-MRI pharmacokinetic parameters can be a useful tool for the quantitative assessment of T stages to improve the accuracy of the T staging of bladder cancer.

This abstract and the presentation materials are available to members only; a login is required.

Join Here