Alan Jeremiah Riordan MSC, Presenter: Nothing to Disclose

Mathias Prokop MD,PhD, Abstract Co-Author: Nothing to Disclose

Jan Willem Dankbaar MD, MSC, Abstract Co-Author: Nothing to Disclose

Hugo W. A .M. De Jong PhD, Abstract Co-Author: Nothing to Disclose

Recent research demonstrates a poor correlation between clinical CT brain perfusion (CTP) packages. Accurate measurement of performance of CTP software is necessary for validation and optimization. To this end, we developed a realistic dynamic brain phantom derived from CTP and MRI acquisitions which could serve as a gold-standard.

A series of digital phantoms were constructed by taking all noise features, effects of resolution and tube settings from a human skull phantom scanned using clinical CTP protocol at various mAs settings. This was combined with 7T MRI data to include healthy and diseased brain tissue, arteries and veins. Time attenuation curves (TAC) were added emulating normal and delayed bolus passage based on clinical CTP data and indicator-dilution theory. The combination of real CTP and MRI data with computed TACs makes this a hybrid phantom. Quantitative performance of 3 CTP software packages were assessed using phantoms at different mAs settings. Methods were: bcSVD and sSVD (as advised by the ASIST workgroup) and fit-based deconvolution (FBD) as implemented by Philips. To illustrate the clinical value of the phantom, the influence of software and settings on parameter maps as found in the phantom was compared to those as found in 4 patient studies.

In the quantitative performance measurements, all CTP software packages showed overestimation of low CBF values and underestimation of high CBF values in the phantoms. Lower mAs resulted in larger bias in all cases, although FBD was relatively insensitive to noise at higher CBF. In contrast to FBD and sSVD, bcSVD was insensitive to delayed bolus arrival time in diseased areas. Bland-Altman analysis of calculated CBF for specific software settings showed good correlation between phantom and clinical data in all tissue types. Change of parameter settings in the perfusion packages resulted in equal change in noise properties, bias and general image quality in the phantom as in the clinical data, illustrating the phantom’s ability to optimize protocols.

The hybrid head phantom provides a realistic gold standard for CTP and can be used to objectively assess and optimize CTP software packages and protocols.

Performance of CT perfusion analysis depends on protocol and software but lacks methods for objective evaluation. A realistic phantom provides a gold standard and enables assessment and optimization.

Riordan, A, Prokop, M, Dankbaar, J, De Jong, H, Validation of CT Brain Perfusion Analysis Methods and Protocols Using a Realistic Dynamic Hybrid Head Phantom.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9001715.html