Recent advancements in MRI hardware, including ultra-high magnetic field scanners (≥7T) and MR data acquisition methods, have enhanced functional imaging techniques, allowing for the detailed study of brain function, particularly at the mesoscopic level of cortical organization. This has enabled the measurement of blood oxygenation level-dependent (BOLD) functional magnetic resonance imaging (fMRI) signal changes across cortical depth in the human brain, facilitating the study of neuronal activity at laminar level. In order to better understand the generation of cortical depth-dependent BOLD signals, biophysical modeling and computational simulations permit the characterization of the impact of vascular architecture, as well as the biophysical and hemodynamic effects at the mesoscopic level. In this study, we employed four realistic 3D vascular models that mimic the human cortical vascular architecture and simulated various vessel-dependent oxygen saturation and hematocrit states, aiming to characterize the intravascular and extravascular contributions to gradient-echo (GE) and spin-echo (SE) BOLD signal changes across human cortical depth at 7T. We found that differences in the local vascular architecture between the four models, away from the pial surface, do not significantly influence the shape and amplitude of BOLD profiles. This implies that signal profiles within a cortical region of a given angioarchitecture can be averaged within a given layer without introducing substantial errors in the results. The findings futher reveal that in deeper laminae, relative relaxation rates for both GE and SE decrease linearly with increasing oxygen saturation levels, with GE showing a stronger effect. In contrast, the top lamina shows a non-linear behavior due to large vessel contributions, particularly venous, with GE displaying higher relaxation rates (4–8 times larger dependent on oxygen saturation levels) than SE. Relative BOLD signal changes also follow linear trends in deeper layers, with GE peaking at ~8% and SE at ~4%, reflecting the higher microvascular specificity of SE. However, SE does not fully eliminate large vessel contributions at the pial surface, where diffusion effects and vessel architecture play a role. Hematocrit levels linearly change the BOLD signal amplitude and significantly influence laminar contributions across cortical depth and imaging techniques. While GE signals are dominated by extravascular effects, SE retains notable intravascular venous contributions at high oxygen saturation levels, which is particularly relevant in experiments involving controlled vascular oxygenation, that is, gas challenges. These results underscore how vascular features, hematocrit, and biophysical interactions shape cortical depth-dependent BOLD signals and their specificity in ultra-high field imaging.

Assessment of neuronal activity using blood oxygenation level-dependent (BOLD) is confounded by how the cerebrovascular architecture modulates hemodynamic responses. To understand brain function at the laminar level, it is crucial to distinguish neuronal signal contributions from those determined by the cortical vascular organization. Therefore, our aim was to investigate the purely vascular contribution in the BOLD signal by using vasoactive stimuli and compare that with neuronal-induced BOLD responses from a visual task. To do so, we estimated the hemodynamic response function (HRF) across cortical depth following brief visual stimulations under different conditions using ultrahigh-field (7 Tesla) functional (f)MRI. We acquired gradient-echo (GE)-echo-planar-imaging (EPI) BOLD, containing contributions from all vessel sizes, and spin-echo (SE)-EPI BOLD for which signal changes predominately originate from microvessels, to distinguish signal weighting from different vascular compartments. Non-neuronal hemodynamic changes were induced by hypercapnia and hyperoxia to estimate cerebrovascular reactivity and venous cerebral blood volume ( C B V v O 2 ). Results show that increases in GE HRF amplitude from deeper to superficial layers coincided with increased macrovascular C B V v O 2 . C B V v O 2 -normalized GE-HRF amplitudes yielded similar cortical depth profiles as SE, thereby possibly improving specificity to neuronal activation. For GE BOLD, faster onset time and shorter time-to-peak were observed toward the deeper layers. Hypercapnia reduced the amplitude of visual stimulus-induced signal responses as denoted by lower GE-HRF amplitudes and longer time-to-peak. In contrast, the SE-HRF amplitude was unaffected by hypercapnia, suggesting that these responses reflect predominantly neurovascular processes that are less contaminated by macrovascular signal contributions.