Vascular and neural basis of the BOLD signal

Abstract

Neural activity in the brain is usually coupled to increases in local cerebral blood flow, leading to the increase in oxygenation that generates the BOLD fMRI signal. Recent work has begun to elucidate the vascular and neural mechanisms underlying the BOLD signal. The dilatory response is distributed throughout the vascular network. Arteries actively dilate within a second following neural activity increases, while venous distensions are passive and have a time course that last tens of seconds. Vasodilation, and thus local blood flow, is controlled by the activity of both neurons and astrocytes via multiple different pathways. The relationship between sensory-driven neural activity and the vascular dynamics in sensory areas are well-captured with a linear convolution model. However, depending on the behavioral state or brain region, the coupling between neural activity and hemodynamic signals can be weak or even inverted.

Figure 1:

Schematic showing vasodilatory and vasoconstrictory signaling pathways used by neurons and astrocytes. Interneurons are blue, pyramidal neurons are grey. The diameter increase happens first in the penetrating arterioles and precapillary arterioles, and propagates to other vascular regions either electrically (pial arterioles) or passively (veins and capillaries). The conducted hyperpolarization initiates in the capillary bed.

Figure 2:

Time courses of neural and vascular changes underlying BOLD signal. A) A stimulus increases gamma-band power in the LFP and spiking activity. B) Arteries dilate rapidly, while the venous response is much slower. The BOLD response is delayed relative to the arterial dilation due to the time taken for the oxygenated blood to transit into the capillaries and veins. C) Hemodynamic response functions (HRFs) for the arterial blood volume, venous blood volume and BOLD signal. The units are arbitrary. Note that the CBV and BOLD HRFs have different scales.

Figure 3:

Schematic showing how non-neuronally driven hemodynamic fluctuations can additively combine with neuronally-generated hemodynamic signals. Top, neural activity showing sensory stimulation and self-generated ‘fidget-related’ neural activity in the sensory cortex. Below is the HRF convolved with the neural activity, giving the component of the hemodynamic signal driven by neural activity. The neurally-driven component (due to vessel-autonomous arterial diameter oscillations that can be detected with optical imaging and two-photon microscopy [68**,93,94]) interact additively with the non-neuronal component, giving the resulting hemodynamic signal, plotted at the bottom. In this schematic, the non-neuronal component is approximately the same magnitude as a the neurally evoked component, so only about half the variance in the measured signal (bottom) can be attributed to neural activity. Note that the amplitude of the spontaneous neurally-evoked component and the non-neuronal component are of comparable amplitude, therefore any isolated fluctuations in the hemodynamic signal should not be taken to be definite indicators of changes in neural activity. Adapted from [68**]. The changes in reflectance (which are inversely related to CBV) plotted in the original figure have been inverted to mimic positive BOLD signals