Beyond BOLD correlations: A more quantitative approach for investigating brain networks

Recently, an enormous amount of research has focused on detecting coherent spatial patterns of brain activity when a subject is simply lying quietly in an MRI scanner. These resting state networks are detected by analyzing correlations of the fluctuations of the bloodoxygenation-level dependent (BOLD) signal. A number of studies have shown that these spatial patterns are reproducible and consistent across subjects, and there is growing evidence that these patterns underlie task-driven responses as well as spontaneous fluctuations during the resting state. Although the significance of these patterns for brain function is still poorly understood, a number of studies have found evidence that resting state connectivity is predictive of task performance. Intriguingly, alterations of the magnitude of the correlations within the parts of a network have been reported in a number of disease states, suggesting that the strength (or coherence) of these patterns may provide an important window on brain function in health and disease. A primary challenge for interpreting these promising results, though, is a basic question: how should we quantify the ‘‘strength’’ of a network? The current problem is the complexity of the BOLD signal, which is driven by cerebral blood flow (CBF) fluctuations, but is strongly modulated by the baseline state and by the coupling of flow and oxygen metabolism fluctuations. While either CBF or metabolism alone could potentially reflect neural activity, the BOLD signal itself is unreliable as a quantitative metric for either one. For this reason, measurements of network strength have focused on correlations of BOLD signals, rather than the BOLD signal amplitude itself. However, this does not solve the problem; disease or medications can alter the BOLD response for the same underlying changes in neural activity and intrinsic connectivity, and this can translate into misleading changes in the correlation values. In short, BOLD correlations are a good way to spatially map networks, but it is critical to develop alternative quantitative physiological metrics for characterizing the strength of networks. In this issue, Weiying et al. show that arterial spin labeling (ASL) methods can be used to quantify network strength in terms of the magnitude of the associated CBF fluctuations. Despite the lower sensitivity, spatial resolution and temporal resolution of ASL compared to BOLD imaging, ASL provides something new. Along with other recent work, this is an important step forward, bringing the study of resting state networks into the realm of quantitative physiology. Shifting the approach to focus on measuring the strength of a network in terms of the magnitude of the associated CBF fluctuations, rather than BOLD correlations, not only avoids the potential pitfalls of relying on the BOLD signal, but also provides a physiological metric that can be compared across brain regions and across subjects. As ASL methods continue to improve, they are likely to become an essential tool for understanding the role of resting state networks in both healthy and impaired brain function.