Related Experiment Video
Updated: Aug 18, 2026
Simultaneous PET/MRI Imaging During Mouse Cerebral Hypoxia-ischemia
Published on: September 20, 2015
Reassessing the clinical efficacy of two MR quantitative DSC PWI CBF algorithms following cross-calibration with PET
Jean J Chen1, Richard Frayne, Michael R Smith
1Department of Electrical and Computer Engineering, University of Calgary, 2500 University Drive, NW, Calgary, Alberta T2N1N4, Canada.
Abstract:
Clinical cerebral blood flow (CBF) maps generated through dynamic- susceptibility contrast (DSC) magnetic resonance (MR) perfusion imaging are currently cross-calibrated with PET studies. The cross-calibration is achieved by rescaling the MR CBF values so that normal white matter CBF corresponds to 22 ml/100 g/min. Examples are provided in this paper to show how this rescaling procedure changes both the clinical interpretation of CBF maps and the manner by which the performance of a given deconvolution algorithm should be assessed. (i) Singular-value decomposition-based (SVD) algorithms produce absolute CBF estimates that are inherently under-estimated for all tissue mean transit times (MTT) but, after rescaling, will generate CBF maps that are over-estimated for MTT >4.8 s. (ii) In principle, frequency-domain modelling techniques are expected to be inherently less sensitive to contrast recirculation biases than the time-domain SVD algorithms. However, it is shown that both CBF algorithms become greatly less sensitive to distortions from recirculation after clinical cross-calibration through rescaling has been performed. It is concluded that, when rescaling procedures are employed, it is relatively more important to develop deconvolution algorithms that produce CBF estimates with accuracies that vary little with MTT than to produce algorithms that provide inherently more accurate CBF estimates, but whose relative accuracy varies significantly with MTT.
Insights
Clinical cerebral blood flow (CBF) maps from DSC MR perfusion imaging require rescaling for accurate interpretation. This process impacts how deconvolution algorithm performance is assessed, emphasizing consistency over absolute accuracy with varying mean transit times (MTT).
Area of Science:
- Neuroimaging
- Medical Physics
- Radiology
Background:
- Clinical cerebral blood flow (CBF) maps are derived from dynamic-susceptibility contrast (DSC) magnetic resonance (MR) perfusion imaging.
- Current cross-calibration with PET studies involves rescaling MR CBF values to a standard normal white matter value (22 ml/100 g/min).
Purpose of the Study:
- To evaluate the impact of CBF map rescaling on clinical interpretation.
- To assess how rescaling affects the evaluation of deconvolution algorithm performance.
Main Methods:
- Analysis of singular-value decomposition (SVD) and frequency-domain modeling algorithms for CBF estimation.
- Examination of the effects of rescaling on CBF estimates across varying mean transit times (MTT).
Main Results:
- Rescaling can lead to over-estimated CBF maps for MTT >4.8 s when using SVD algorithms.
- Both SVD and frequency-domain algorithms show reduced sensitivity to recirculation artifacts after rescaling.
- The relative accuracy of CBF estimates becomes more critical than inherent accuracy after rescaling.
Conclusions:
- Rescaling procedures significantly alter CBF map interpretation and algorithm assessment.
- Developing deconvolution algorithms with consistent accuracy across different MTTs is crucial when rescaling is applied.
