All fMRI techniques in use today measure brain function only indirectly, by tracking the changes in blood flow, volume and oxygenation that accompany neuronal activity, and this has often been viewed as the fundamental limitation of the technique. However, recent evidence from invasive in-vivo microscopy studies has shown that the brain's smallest blood vessels respond far more precisely, in space and in time, to neuronal activity than previously believed. This insight suggests that the “biological resolution” of fMRI is intrinsically high, and, with sufficiently high imaging resolution, it should be possible to extract more meaningful neuronally specific information from fMRI—if we can understand how brain vascular anatomy and physiology shape the hemodynamics that generate the fMRI signals.

  In this presentation, I will describe ongoing efforts to improve the neuronal specificity of fMRI and pose the question: How far can we go with fMRI? The limits of fMRI spatial and temporal resolution are actively being investigated using advanced imaging technologies. While high-resolution human fMRI studies are increasingly operating at the boundaries of what is achievable, a key challenge is that the vascular architecture of the brain reflects its structure and function across spatial scales. Both classic and modern vascular anatomy studies have shown how the macro-vascular geometry is coupled with the tissue geometry, including the gray matter folds and the white matter tracts, while the micro-vascular density closely follows borders of subcortical nuclei, cortical areas and cortical layers. I will present evidence that both the large- and small-scale vascular anatomy strongly influence patterns of fMRI activation and describe strategies for how to account for this.

  As examples of the intrinsically high biological resolution of fMRI, I will present results showing cortical columnar and laminar imaging, and new directions in the emerging field of ”fast fMRI” that show how the BOLD response can track surprisingly fast neural dynamics. Lastly, I will share our recent progress towards building bottom-up biophysical models of the fMRI signals based on realistic vascular anatomy and dynamics that provide insights into the interrelationship between hemodynamics and neural activity. Overall, many lessons can be learned through a deeper understanding of brain vascular anatomy and physiology, which can both shed light on the brain's functional organization and help neuroscientists more accurately interpret the fMRI signals in terms of the underlying neural activity.

     7T fMRI enables the resolution of brain activity across cortical depths to understand feedforward and feedback dynamics. The relationship between these hemodynamic signals and neural activity is less well explored. In this talk, we present results correlating 7T fMRI in healthy individuals with invasive electrophysiological recordings from epilepsy patients to examine layer-dependent coupling between neuronal activity and fMRI during passive music listening. Specifically, Layer-specific fMRI responses were modeled using neuronal oscillation envelopes elicited by the same naturalistic stimuli. From deep toward superficial layers, the relationship between oscillatory power and fMRI responses systematically changed: alpha/beta activity (8-30 Hz) was increasingly associated with negative fMRI responses, while gamma band (>30 Hz) oscillations showed increasingly positive associations. The envelope of broadband high-frequency activity (>70 Hz) showed the strongest link with fMRI signals in the intermediate layers. This "feedforward type" dominance of intermediate layers was also clearly present in the fMRI analysis using the acoustical envelope itself. Our findings reveal a spectrolaminar organization of neurovascular coupling in the human auditory cortex.