Does BOLD functional MRI have the capability of mapping submillimeter-scale functional columns?

[Background and aims] The ability of blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI) in mapping sub-millimeter-scale functional columns still remains debatable mainly due to the spatially non-specific large vessel contribution, poor sensitivity and reproducibility, and lack of independent evaluation. Furthermore, regions with the highest conventional BOLD signals may indicate neurally inactive domains rather than active columns if, for instance, the results from optical intrinsic signal (OIS) are directly applicable. To examine these issues, here we performed BOLD fMRI at an ultrahigh magnetic field to map orientation-selective columns of isoflurane-anesthetized cats. [Methods] Gradient-echo (GE) BOLD fMRI (TE = 18 ms, TR = 500 ms) and spin-echo (SE) BOLD fMRI (TE = 40 ms, TR = 2 s) with 0.3x0.3x1.0 mm3 resolution were obtained in six anesthetized cats (10 hemispheres) at 9.4 Tesla. SE BOLD technique at high magnetic fields is mostly sensitive to microvessels including capillaries, and has been shown to improve spatial localization to the parenchyma. As a reference for the known neural interpretation, cerebral blood volume (CBV)-weighted fMRI (TE = 10 ms, TR = 1 s) was also obtained after the injection of monocrystalline iron oxide nanoparticles (MION). Visual responses were induced by 800-seclong full-field moving gratings with eight orientations (10 s for each orientation, 10 repetitions) without inter-stimulation gap. Fourier analysis was applied to obtain maps of orientation-specific signal (1/80 Hz). For neural interpretation, BOLD iso-orientation maps were directly compared to MION-based CBV fMRI maps and correlation coefficients (R) were calculated. For further evaluation (n = 3 of 6 cats, 4 hemispheres), the BOLD maps were also compared to the iso-orientation maps in the same cats from CBV-weighted OIS. [Results] Temporally-encoded continuous cyclic orientation stimulation with Fourier analysis reduces orientation-non-selective signals such as draining artifacts and allows us to obtain conventional GE BOLD iso-orientation maps with high reproducibility (R = 0.74 ± 0.12, n = 10 hemispheres). Further reduction of large vessel contribution was achieved using, SE BOLD, but with overall decreased sensitivity; the ratio of SE BOLD contrast-to-noise ratio (CNR) to GE BOLD CNR was approximately 1 to 3. The both GE and SE BOLD iso-orientation maps were significantly correlated with maps determined by MION-based CBV fMRI (R = 0.78 ± 0.06 for GE and 0.76 ± 0.08 for SE). Furthermore, the average least-distance for local maxima of iso-orientation domains between GE BOLD and OIS maps was 0.22 ± 0.02 mm (n = 4 hemispheres). Since these distances are shorter than the column width of 0.70 mm (i.e., a half of the measured inter-columnar distance), we consider that the BOLD iso-orientation maps match with OIS isoorientation maps. These results suggest that the highest BOLD signals are localized to the sites of increased neural activity when column-non-selective signals are suppressed. [Conclusions] Thus, sub-millimeter-scale functional columns can be reliably mapped using conventional BOLD fMRI.