The rCBF (Watabe) model was intended for the quantitative assessment of the regional cerebral blood flow. The only required measurement is a dynamic PET acquisition after the injection of a H215O bolus, obviating the need for blood sampling. What is required instead are the time-activity curves of two cerebral regions - a low-flow and a high-flow region. By comparing the TACs of these regions their flows (f1, f2) and distribution volumes (DV1, DV2) can be estimated. Then, the TAC from each pixel can be compared to that of the low-flow region. Its flow and distribution volume are hereby estimated, whereby the previously obtained parameter values of the low-flow region (f1, DV1) and a fixed DV are contained in the calculation prescription.
CAUTION: With practical data it has turned out that the fitting is too dependent on the initial parameter values and does not provide sufficiently stable results.
Acquisition and Data Requirements
Image Data |
A dynamic PET data set representing the measurements of brain activity after injection of a H215O bolus. |
Model Preprocessing
A methodology has been implemented to automatically extract low-flow and high-flow TACs as follows:
1.The signals are integrated over the acquisition duration in all pixels.
2.A lower threshold (40% in the example below) is applied to restrict the volume-of-interest to brain.
3.A histogram within the threshold volume is calculated.
4.The C1 (ex. 5000) pixels at the lower end of the histogram are assumed to represent low-flow pixels; their average curve TAC1 is calculated. Only this GENERATED approach is supported, no manual TAC specification.
5.The C2 (ex. 2000) pixels at the upper end of the histogram are assumed to represent high-flow pixels; their average curve TAC2 is calculated. Only this GENERATED approach is supported, no manual TAC specification.
6.Then an iterative fit of Watabe's eq. (7) is performed to calculate the flows (f1,f2) and distribution volumes (DV1,DV2) of the two TACs. The resulting f1 and DV1 together with an assumed distribution volume DV (which must be specified by the user) are then used for the pixelwise flow calculations by Watabe's eq. (8).
f1 |
Flow in low-flow region estimated during preprocessing and subsequently used in pixelwise calculations. |
DV1 |
Distribution volume in low-flow region estimated during preprocessing and subsequently used in pixelwise calculations. |
f2 |
Flow in high-flow region estimated during preprocessing but NOT further used. |
DV2 |
Distribution volume in high-flow region estimated during preprocessing but NOT further used. |
DV |
Fixed distribution volume which is assumed for each pixelwise TAC. |
Weighting |
Different schemes for residual weighting in the iterative preprocessing fit. |
C1 points |
Number of points used for the generation of TAC1. |
C2 points |
Number of points used for the generation of TAC2. |
Percent masked pixels |
Exclude the specified percentage of pixels based on histogram analysis of integrated signal energy. Not applied in the presence of a defined mask. |
Important Note: Experience has shown that the results of this reference method highly depend on the iterative fit with 4 parameters. As illustrated in this example, the identifiability of the parameters is often poor, and the results may heavily depend on the starting values. In an attempt to make the method more stable, one can fix VD2 to a reasonable value such as 0.9. To this end, just deactivate the box next to VD2 .
Model Configuration
f |
Regional perfusion in [ml/min/100ml]. It is calculated for each pixel by a closed-form calculation (Watabe's eq. 8). Reasonable values are about 30 (white matter) and 60 (gray matter) ml/min/100ml [11]. |
Reference
1.Watabe H, Itoh M, Cunningham V, Lammertsma AA, Bloomfield P, Mejia M, Fujiwara T, Jones AK, Jones T, Nakamura T: Noninvasive quantification of rCBF using positron emission tomography. J Cereb Blood Flow Metab 1996, 16(2):311-319.