Abstract

Selecting User-defined Parameters
Commercially available deconvolution-based CTP software requires the user to select a region-of-interest (ROI) from an artery and vein to serve as the "arterial input" and "venous outflow" parameters for perfusion calculations (Fig. 3). The arterial input function (AIF) is required mathematically to perform the CTP calculations for the deconvolution method. An attempt should be made to select arterial and venous locations that yield the TDC with maximal peak contrast intensity. The location of the AIF has no restrictions, and does not affect the final quantitative results (15). However, the selection of the venous function (VF) does significantly influence the final quantitative values of CBF and CBV (15), as it serves to correct the AIF for partial volume averaging effects. The peak enhancement values of the venous ROI's also correlate significantly with the signal-to-noise ratio of the CBF and CBV maps. The torcular is an ideal location for the venous ROI, as the confluence of draining venous sinuses are large and perpendicular to the imaging plane, avoiding potential partial volume averaging.

The observer also needs to define where along the time-axis of the TDC to start and end the deconvolution calculation - the so-called "pre" and "post" enhancement cutoff values (Fig. 4). The perfusion analysis program will ignore data from slices outside the range of the cut-off points. The post-enhancement cut-off value should be set after the initial arterial, venous, and tissue downslopes - prior to any second, smaller upslope, reflecting recirculation (15). Variations in the post-enhancement time points can significantly influence the resulting quantitative CTP values, due to the scaling function nature of the venous TDC. Additionally, deconvolution software requires a sufficient baseline for accurate calculation of CTP values; inappropriate inclusion of pre-enhancement portions of the arterial and venous curves may result in inaccuracies due to increased noise (15).

Post Processing Pearls: Construction of CTP Maps with Deconvolution Software

Avoid partial volume averaging:
1. Select the largest vessel perpendicular to the imaging plane for the AIF and VF
2. Choose an ROI pixel size that fits within the lumen of the arterial and venous structures
3. Avoid small, deep veins for the venous ROI
Limit truncation of the time-density-curves (TDC):
1. Select the post-enhancement cut-off at a timepoint after the venous and tissue TDC's have completed their downslopes and reached a plateau, but before recirculation occurs
2. Select the pre-enhancement cut-off at a timepoint immediately preceding the upslope of the arterial TDC

Reproducibility
Few studies have evaluated the variability in quantitative CTP values constructed by different observers from the same datasets. In a pilot investigation involving experienced radiologists from different institutions (Fig. 5), quantitative CTP maps were calculated using the same software program (GE Functool) (16). The resulting mean CBF, CBV and MTT values were not statistically different between the two observers (p<0.001). In no case did disagreements result in change of classification of tissue at risk (16). Moreover, a high level of correlation was also shown amongst technologists post-processing CTP datasets (17). Despite these findings, the level of agreement was insufficient to endorse the use of quantitative CT perfusion data for routine clinical decision-making (17). This may be partly explained by the selection of the user-defined inputs, as described above. Therefore, a standardized approach in selection of user-defined inputs may improve the reproducibility of CTP maps within and between institutions.

III. Post-processing