Compared to the dominant gradient-echo EPI (GE-EPI) technique, spin-echo EPI is much less susceptible to signal drop out, and potentially less sensitive to large-vein contamination. Moreover, in our previous work¹, we had demonstrated the potential of using a spin-echo EPI (SE-EPI) scan for detecting fMRI activity in the white matter (WM). However, SE-EPI generally suffers from lower BOLD signal sensitivity at lower field strengths (e.g. 3 T). Conventionally, the optimal echo time (TE) for spin-echo EPI (SE-EPI) BOLD is assumed to be near the tissue T2 (65‑100ms at 3 T)². However, unlike for GE-EPI³, the optimal TE for SE-EPI was never experimentally verified in light of physiological noise and large-vein contamination. The purpose of this work is to experimentally identify the optimal TE by characterizing the sensitivity of BOLD signal at different TEs to a hypercapnia paradigm³.

Four healthy humans were scanned on a 3T Siemens Tim Trio scanner (Siemens Healthcare, Erlangen Germany). A total of five SE-EPI scans were performed on each subject at 5 different TEs: 35, 45, 55, 65 and 75 ms. The imaging parameters were TR = 2s, 192 time points, matrix size = 64 x 64 x 20 and voxel size = 3.4 X 3.4 X 5.8 mm³. A 3D T1-weighted MPRAGE scan was also acquired. During MR Imaging, we induced hypercapnia by administering mixtures of CO₂ and O₂ using the RespirAct™ breathing circuit (Thornhill Research, Toronto, Canada). Hypercapnic condition was achieved by increasing end-tidal CO₂ pressure (PETCO₂) by 5 mmHg from the baseline. The experimental paradigm was composed of 2 hypercapnic blocks as shown in Fig. 1.

The fMRI data were pre-processed using FSL (www.fmrib.ox.ac.uk/fsl), involving slice timing and motion correction, high pass filtering (<0.005 Hz) and spatial smoothing (5 mm FWHM). Lastly, fMRI images were registered to 3D anatomical images. WM and GM masks were created from the anatomical images using FMRIB's Automated Segmentation Tool (FAST). These masks were registered and down-sampled to fMRI native space, and used to compute average normalized percent change in BOLD signal (∆BOLD%), contrast-to-noise ratio (CNR) and temporal signal-to-noise ratio (tSNR) using the following equations:

$$\frac{\Delta\ BOLD \%}{mmHg}=\left(\frac{ mean\ BOLD_{hypercapnia}}{mean\ BOLD_{normocapnia}}-1\right)\diagup\Delta PETCO_{2}$$

$$\frac{CNR}{mmHg}=\left(\begin{array}a\frac{mean\ BOLD_{hypercapnia} \ \ -\ mean\ BOLD_{normocapnia}}{stdev. \ BOLD_{normocapnia}}\end{array}\right)\diagup\Delta PETCO_{2}$$

$$tSNR = \frac{mean \ BOLD}{stdev. \ BOLD}$$

∆BOLD% and CNR were then normalized to ΔPETCO₂ to account for the variable response of the subjects to hypercapnic manipulation. To assess the TE-dependence of venous contamination, we also calculated the ∆BOLD% and CNR in the sagittal sinus (SS).

As seen in Fig. 2, GM exhibits a higher normalized SE ∆BOLD%/mmHg and CNR/mmHg compared to WM, but the CNR difference between GM and WM is less than the ∆BOLD% difference. As expected, tSNR decreases with increasing TE in both tissue types. Specific to WM, the ∆BOLD% and CNR trends are summarized in Fig. 3, and show that a TE of ~50ms enables a trade-off between CNR and tSNR. Lastly, hypercapnia induced a higher ∆BOLD%/mmHg in the sinus compared to both WM and GM, as shown in Fig. 4. Interestingly, however, while the tissue BOLD CNR does not change significantly with lengthening TE, the BOLD CNR associated with the sinus increases at longer TEs.As expected, GM is associated with higher BOLD detectability than WM (based on all three detectability metrics), due to a higher cerebrovascular blood flow and volume. However, the CNR advantage of GM is somewhat diminished, likely since GM is also more sensitive to physiological noise. We note that a mid-range TE of ~50 ms is likely to offer a better trade off between tSNR and CNR. The conventionally assumed TE of ~75 ms at 3 T has been shown to result in rather high venous contamination (Fig. 4) for a modest gain in CNR, and may thus not be optimal for BOLD detectability at 3 T.We established that an approximate TE of 50ms is optimal for GM and WM BOLD signal detection when using SE-EPI. The high flow velocity in SS during hypercapnia may call for improved methods to determine macrovascular BOLD sensitivity as part of our future work. The results of this study might help improve the performance of SE-EPI BOLD in both grey and white matter in future fMRI studies.