Meeting Banner
Abstract #3451

K-Means Clustering of DCE-MRI Pharmacokinetic Parameters for Prediction of Chemotherapeutic Response of Bladder Cancer

Huyen Thanh Nguyen1, Guang Jia1, Zarine K. Shah, Kamal S. Pohar2, Amir Mortazavi3, Debra L. Zynger4, Xiangyu Yang1, Michael V. Knopp1

1Wright Center of Innovation in Biomedical Imaging, Department of Radiology, The Ohio State University, Columbus, OH, United States; 2Department of Urology, The Ohio State University, Columbus, OH, United States; 3Department of Internal Medicine, The Ohio State University, Columbus, OH, United States; 4Department of Pathology, The Ohio State University, Columbus, OH, United States

Purpose: to evaluate the capability of k-means clustering of DCE-MRI pharmacokinetic parameters in predicting chemotherapeutic response of bladder cancer at mid-cycle of chemotherapy. Methods: K-means clustering of voxel-based pharmacokinetic parameters was performed to determine the three cluster centroids. For each tumor, volume fraction of each cluster was calculated; and their changes from baseline to mid-cycle were determined and correlated with the tumors chemotherapeutic response. Results: The changes in the volume fraction of the three clusters from baseline to mid-cycle MRIs trended in the opposite directions and were significantly different between responders and non-responders (P<0.01). Conclusion: K-means clustering of DCE-MRI pharmacokinetic parameters shows robustness in revealing the complex change of microcirculation to enable prediction of chemotherapeutic response of bladder cancer at mid-cycle MRI.

Keywords

administered agent aimed amplitude annual applied applying array assessment biomedical bladder body burden cancer capability cause centroids channel characteristics characterized chemotherapeutic chemotherapy classified cluster clustering clusters coil compartment complex confirmation confirmed considered constant contrast correlated curves cycle cycles decreased defined definitive delay determine determined display dose early enable enhancement environment evaluated evaluating examination exams exchange females findings flow fraction gradient greater illustrated included innovation internal invasive kilogram last linear male males malignancy materials measure medicine medium meeting model modified ongoing opposite parametric pathological pathologically pathology patients plasma plot post predicting prediction previous proceedings processed progression promising properties prospective pure quantitative radiology reduce reduction reliable resistant resolution responder responders response responsive responsiveness reveal revealing robustness scanned sent shah significantly slope software space specimens spin spoiled stable stage statistical statistically steep subjects subsequently surface surgical system transmit treatment trended tumor tumors turbo unnecessary upper urology volume whole wright yang year