Program Information

Optimal Choice of Deconvolution Methods in CT Perfusion Imaging

E Harvey

Evan Harvey*, Guang-Hong Chen, Ke Li, University of Wisconsin-Madison, Madison, WI

Presentations

TH-CD-601-7 (Thursday, August 3, 2017) 10:00 AM - 12:00 PM Room: 601

Purpose: Recently, the so-called delay-insensitive deconvolution methods are gaining popularity in CT perfusion imaging with the claims of improved quantification accuracy over the conventional deconvolution method. The purpose of this work was to investigate under what conditions the delay-insensitive deconvolution method yields more quantitative perfusion measurements?

Methods: The major difference between delay-insensitive and conventional deconvolution methods is the construction of the convolution matrix, which is composed of elements from the arterial input function (AIF). For delay-insensitive deconvolution, it uses a block-circulant matrix that is different from the one used in conventional deconvolution. A cascaded systems model that is applicable to both conventional and delay-insensitive methods was developed to quantitatively study how the delay in contrast arrival time impacts the quantitative accuracy of both methods. Using this theoretical model, the quantification accuracy of cerebral blood flow (CBF) was estimated with different delay times ranging from 0 to 15 s. To validate the theory, simulated CT perfusion acquisitions of a digital phantom were performed, and CBF measured from the simulation was compared with theoretical results.

Results: For normal tissue, the bias in the CBF given by conventional deconvolution was 8 ml/s/100g less than that generated by delay-insensitive method, provided that there was no delay between AIF and tissue enhancement curve; the bias from the two methods became similar when there was delay. For hypoperfused but potentially salvageable tissue, the biases of the two methods were similar when there was no delay, but the conventional method led to 26.5 ml/s/100g higher bias when there was 15 s delay.

Conclusion: Delay-insensitive method does yield better quantification when there is severe delay in contrast arrival time between artery and tissue of interest. However, in the absence of delay, it is the conventional deconvolution method that yields better quantification accuracy.

Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by an NIH grant U01EB021183


Contact Email: