The inherent resolution of high-quality, commercially available PET scanners is around 4 to 7 mm (full-width half-maximum; DeGrado et al., 1994; Spinks et al., 2000). This is high enough to measure activity in most cortical and subcortical regions, but limits the ability to look at subnuclei and often leads to difficulties in determining the exact origin of foci that occur near the boundaries between regions. By comparison, fMRI is capable of higher spatial resolution. However, many fMRI studies are performed with parameters that provide no higher spatial resolution than that produced by high-quality PET cameras. Moreover, draining vein effects often lead to mislocalization of the source of fMRI signals (Lai et al., 1993), thus lowering the effective resolution for localizing responses. Both fMRI and PET images are usually filtered to a lower spatial resolution after the data is collected. This filtering serves several purposes. First, it reduces noise and hence improves signal to noise characteristics. Second, it lowers the number of resolution elements and hence reduces correction factors for multiple comparisons. Third, it improves the detectability of large-volume activations (Poline & Mazoyer, 1994). However, this comes at the cost of restricting the ability to detect more discrete focal activations and therefore biases the methods toward detections of large-volume activations. This bias is at its most extreme in older PET studies, but remains a bias in the fMRI literature as well.
Was this article helpful?