It was previously shown that the contraction force scales positively with the BOLD signal change¹, but due to the ambiguous nature of the BOLD contrast alone, it is impossible to say to which extent this proportionality comes from an increased oxidative metabolism rather than the vasculature. A useful tool for investigating oxygen consumption in healthy human brain is Davis model for fMRI signal calibration². However, applicability at high field is limited by low CNR of ASL sequences. Vascular-Space Occupancy (VASO) measures CBV, providing a valid alternative to CBF to estimate CMRO₂ changes. Here, we aimed to disentangle the ambiguity between metabolic and vascular effects by combining a gas manipulation challenge (mild hypercapnia) with BOLD/VASO measurements during a controlled pinch-force task.

Experiments comprising a gas-manipulation and a pinch-force session were performed at 7T (Siemens, Erlangen, Germany) in 6 healthy participants (23±2 years). Two were excluded due to intolerance to gas manipulation or excessive motion (>3 mm).

Gas manipulation: Hypercapnia was achieved by administering a gas containing 5% CO₂, 21% O₂ and 74% N₂ through a non-rebreathing mouth piece, twice for 3 min (15min total session duration).

Pinching task: The force task involved pinching with 4 fingers simultaneously (thumb, index, middle and ring finger) between 10% and 66% of the participant’s MVC in a controlled sinusoidal fashion (wavelength = 2s). Activation periods of 1 minute were interleaved to rest periods of the same duration (Fig. 1), for a total time of 15 min. Force data were recorded simultaneously, with a temporal resolution of 0.1 ms and undersampled to match the temporal resolution of the MR acquisitions.

Acquisition: Interleaved BOLD and VASO acquisitions were recorded with simultaneous multi-slice (SMS) SS-SI VASO³ with TI1/TI2/TR/TE = 1100/2600/3000/26 ms; 16 slices of 1×1×1.5 mm³ nominal resolution (no slice gaps); SMS factor = 2; CAIPI-shift = 2, GRAPPA factor = 4, segmented reference line acquisition, no partial Fourier.

Analysis: An ROI encompassing the primary motor area (M1) was defined based on the FEAT-detected VASO activation (more specific to grey matter than BOLD⁴). For the calculation of CMRO2 changes, the Davis model was solved at every timepoint combining BOLD and VASO data from the same session with the M value calculated from the gas manipulation acquisition using the following parameters⁵: α_tot = 0.38, α_vein = 0.2, β = 1. VASO data were first interpolated (mean between neighboring timepoints), and the Davis model was calculated at each timestep.

The average MVC across participants was (8.2±2.1) kg. Statistically significant activation was found and used to define the ROI in the contralateral M1 (Fig. 2). Corresponding time courses are shown in Fig. 1.

Average signal changes for the pinch-force task were found to be (5.1±0.3)% for BOLD and (-3.0±0.4)% for VASO. For hypercapnia, a BOLD signal change of (5.3±0.7)% and a VASO signal change of (-1.3±1.1)% were obtained. The parameter M was found to be (16±3)%, and the average CMRO₂ change during activation was estimated as (59±27)%.

A timecourse of CMRO₂ changes was constructed at each timestep of the pinch-force task session as shown in Fig. 3. Correlation between BOLD signal changes, VASO signal changes and estimated CMRO₂ changes versus normalized force values are visualized in scatter plots (Fig. 4).

All contrasts (BOLD signal, inverted VASO signal, and estimated CMRO₂ changes) show a positive correlation with the force level (note that VASO is a negative contrast, i.e. a positive CBV change leads to a negative VASO change). This positive trend is weaker for CMRO₂ changes than for BOLD and VASO changes. This may suggest that the stronger correlation of BOLD with the force level comes from CBF and CBV changes primarily, and only to a lesser extent from an increased oxygen consumption. However, it may also be a result of the higher noise level in CMRO₂-time courses compared to BOLD and VASO.

As expected, the CMRO₂ timecourse (Fig. 3), resemble the functional paradigm, however, it is much more affected by noise than the BOLD or VASO timecourses.

The application of the Davis model itself was shown to be straightforward at 7T when BOLD and VASO contrasts are acquired. Constructing a CMRO₂ timecourse is feasible but challenging even at high resolution, due to the noise amplification inherent to the model. It was shown that the applied contraction scales positively with the BOLD signal, as previously observed, but has only a weak correlation with CMRO₂ changes.