Reduced task durations in functional PET imaging with [¹⁸F]FDG approaching that of functional MRI

The brain's energy budget can be non-invasively assessed with different imaging modalities such as functional MRI (fMRI) and PET (fPET), which are sensitive to oxygen and glucose demands, respectively. The introduction of hybrid PET/MRI systems further enables the simultaneous acquisition of these parameters. Although a recently developed method offers the quantification of task-specific changes in glucose metabolism (CMRGlu) in a single measurement, direct comparison of the two imaging modalities is still difficult because of the different temporal resolutions. Thus, we optimized the protocol and systematically assessed shortened task durations of fPET to approach that of fMRI.

Twenty healthy subjects (9 male) underwent one measurement on a hybrid PET/MRI scanner. During the scan, tasks were completed in four blocks for fMRI (4 × 30 s blocks) and fPET: participants tapped the fingers of their right hand repeatedly to the thumb while watching videos of landscapes. For fPET, subjects were randomly assigned to groups of n = 5 with varying task durations of 10, 5, 2 and 1 min, where task durations were kept constant within a measurement. The radiolabeled glucose analogue [¹⁸F]FDG was administered as 20% bolus plus constant infusion. The bolus increases the signal-to-noise ratio and leaves sufficient activity to detect task-related effects but poses additional challenges due to a discontinuity in the tracer uptake. First, three approaches to remove task effects from the baseline term were evaluated: (1) multimodal, based on the individual fMRI analysis, (2) atlas-based by removing presumably activated regions and (3) model-based by fitting the baseline with exponential functions. Second, we investigated the need to capture the arterial input function peak with automatic blood sampling for the quantification of CMRGlu. We finally compared the task-specific activation obtained from fPET and fMRI qualitatively and statistically.

CMRGlu quantified only with manual arterial samples showed a strong correlation to that obtained with automatic sampling (r = 0.9996). The multimodal baseline definition was superior to the other tested approaches in terms of residuals (p < 0.001). Significant task-specific changes in CMRGlu were found in the primary visual and motor cortices (t_M1 = 18.7 and t_V1 = 18.3). Significant changes of fMRI activation were found in the same areas (t_M1 = 16.0 and t_V1 = 17.6) but additionally in the supplementary motor area, ipsilateral motor cortex and secondary visual cortex. Post-hoc t-tests showed strongest effects for task durations of 5 and 2 min (all p < 0.05 FWE corrected), whereas 1 min exhibited pronounced unspecific activation. Percent signal change (PSC) was higher for CMRGlu (∼18%-27%) compared to fMRI (∼2%). No significant association between PSC of task-specific CMRGlu and fMRI was found (r = 0.26).

Using a bolus plus constant infusion protocol, the necessary task duration for reliable quantification of task-specific CMRGlu could be reduced to 5 and 2 min, therefore, approaching that of fMRI. Important for valid quantification is a correct baseline definition, which was ideal when task-relevant voxels were determined with fMRI. The absence of a correlation and the different activation pattern between fPET and fMRI suggest that glucose metabolism and oxygen demand capture complementary aspects of energy demands.