Strictly speaking, fMRI machines measure the blood's oxygen level through a technique called blood oxygenation level-dependent (BOLD) contrast. The BOLD fMRI signal is supposedly related to blood changes, which are supposedly related to local metabolic changes, which are supposedly related to changes at the neuronal population activation level. These activation changes are presumably associated with changes in neural information processing. "Almost all of those links are at least under-specified at the moment," says Michael Young, director of the Institute of Neuroscience, University of Newcastle upon Tyne, and a longtime investigator of the relationship between fMRI signals and neuronal information processing.
Young and other fMRI researchers have applauded the recent study as a technical triumph and an important first step toward a much better understanding of fMRI signals. Study leader Nikos Logothetis, director of the physiology of cognitive processes group at Max Planck, and his collaborators, spent years constructing a magnet that could experimentally demonstrate the nature of the fMRI-neural activity link. This magnet can simultaneously record fMRI measurements and individual neurons in macaque monkeys.
Logothetis cites three major conclusions of the study, the first being that "the fMRI in general reflects quite faithfully the underlying neural activity." However, things get more complicated, says Logothetis, when considering what actually causes the BOLD signal.
Logothetis' experiments showed, for the first time, that most BOLD activity actually corresponds to presynaptic events, which might not necessarily result in neural spike activity. Researchers had generally believed that the BOLD signal corresponded to spike rate. "So fMRI is probably seeing some things that might not actually manifest themselves in actual information processing," Young explains. As a result, researchers using fMRI might overestimate neural activity, Logothetis says.
The third conclusion relates to another potential interpretive pitfall: directly measuring neural activity with electrodes is more reliable than fMRI-measured activity by a couple of orders of magnitude, and with less variance. As a result, reported fMRI patterns can greatly understate the number of structures that are actually involved in information processing for a given task.
Logothetis emphasizes that his research is only the first step in what figures to be a wave of findings that better elucidate fMRI intricacies. Within a year, several institutions, including the National Institutes of Health, will have Logothetis's new fMRI device in hand, Logothetis says. "It's a sea of change that's come over this area," says Young, whose own lab will soon have the new device. "I hope we're really moving from ... a rather qualitative and informal approach to brain activation, to one which is much more grounded in experimental, computational, mathematical rigor."
1. E. Russo, "Debating the Meaning of fMRI," The Scientist, 14:20, Sep. 18, 2000.
2. N. K. Logothetis et al., "Neurophysiological investigation of the basis of the fMRI signal," Nature, 412:150-7, July 12, 2000.