Abstract #0102

The Use of k-Means Clustering and Bayesian Inference Framework for the Processing of Vessel-Encoded P-CASL Images as Compared with Super-Selective P-CASL MRI

Nolan S. Hartkamp¹, Michael Helle², Michael A. Chappell³,⁴, Thomas W. Okell⁴, Reinoud P H Bokkers¹, Jeroen Hendrikse¹, Matthias J.P. van Osch⁵

¹Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands; ²Philips Research Laboratories, Hamburg, Germany; ³Institute of Biomedical Engineering, Department of Engineering Science, University of Oxford, Oxford, United Kingdom; ⁴FMRIB Centre, University of Oxford, Oxford, United Kingdom; ⁵C.J. Gorter Center, Department of Radiology, Leiden University Medical Center, Leiden, Netherlands

We show that the territorial perfusion maps produced by VE p-CASL agree reasonably well with the perfusion maps acquired with super-selective p-CASL. Special consideration should be taken when using k-means clustering since it tends to fail in regions with high mixed perfusion, such as the deep gray matter. VE p-CASL with k-means clustering appears suitable as a general purpose T-ASL strategy, but the Bayesian framework is preferable since it can determine mixed perfusion. This is however only reliable where the VE p-CASL images contain sufficient vessel selectivity. To accurately determine the perfusion territories of a vessel, super-selective p-CASL is still recommended.