Calibrated BOLD fMRI with an Optimized ASL-BOLD Dual-Acquisition Sequence
Maria A. Fernandez-Seara1, Zachary B. Rodgers2, Erin K. Englund2, Hee-Kwon Song2, John A. Detre3, Michael C. Langham2, and Felix W. Wehrli2

1Radiology, University of Navarra, Pamplona, Spain, 2Radiology, University of Pennsylvania, Philadelphia, PA, United States, 3Neurology, University of Pennsylvania, Philadelphia, PA, United States

Synopsis

Calibrated fMRI techniques estimate task-induced changes in the cerebral metabolic rate of oxygen (CMRO2) based on simultaneous measurements of cerebral blood flow (CBF) and blood-oxygen-level-dependent (BOLD) signal changes evoked by stimulation. To determine the calibration factor M (corresponding to the maximum possible BOLD signal increase), BOLD signal and CBF are measured in response to a gas breathing challenge (CO2, O2). Here we describe an ASL dual-acquisition sequence that combines a background-suppressed 3D readout with 2D multi-slice EPI. In five subjects we found an average gray matter M-value of 8.71±1.03 and fractional changes of CMRO2 of 12.5±5% in response to a bilateral motor task.

Introduction

Calibrated fMRI techniques estimate task-induced changes in the cerebral metabolic rate of oxygen consumption (CMRO2) based on simultaneous measurements of cerebral blood flow (CBF) and blood-oxygen-level-dependent (BOLD) signal changes evoked by stimulation (1). To calibrate the BOLD signal, a calibration factor M (corresponding to the maximum possible signal increase) is measured using blood gas manipulation techniques, involving CO2 or O2 gas mixture breathing. This measurement is quite sensitive to arterial spin labeling (ASL) noise, especially when using the hypercapnia (HC) method. Nevertheless, data acquisition is commonly done using double-echo EPI, which is suboptimal for ASL, due to the difficulty of implementing background suppression (BS) in 2D-multislice readouts (2). A background-suppressed double acquisition sequence has previously been proposed (3) using two separate EPI readouts. However, 3D readouts are desirable for ASL since they can be optimally combined with BS (4).

The goal of this study was to implement a dual-acquisition sequence for simultaneous measurement of BOLD and CBF and evaluate its potential for M-mapping and quantification of CMRO2 changes in response to a motor task.

Methods

Pulse sequence: The sequence (Fig. 1) consists of a pseudo-continuous ASL (PCASL) module with a background-suppressed single-shot 3D-GRASE readout for acquisition of ASL images and multi-slice 2D-EPI for acquisition of BOLD images. The two readouts are separated by a delay of 1.3s to allow for recovery of the longitudinal magnetization.

Subjects: 5 healthy subjects (34±6 years) participated in the study, after signing written informed consent.

Scanning protocol: The study was performed on a 3T Siemens Trio using a 32-channel head-array. The scanning session included an anatomic T1-weighted image. Functional data were acquired with the following parameters: PCASL labeling time=1.6s, post-labeling delay=1.5s, TE (GRASE/EPI)=29.3/30 ms, TR=6s, in-plane resolution=4x4 mm2, 16 slices, slice thickness=6mm. First, data were acquired throughout a gas mixture breathing protocol (5 min room air (baseline), 5 min breathing 5% CO2 in room air (HC) and 5 min room air (recovery)) (5), followed by one run with a motor activation paradigm (3 blocks of bilateral finger tapping alternated with 3 rest blocks of 1 min duration).

Data pre-processing (MATLAB scripts and SPM8): GRASE and EPI images were separately pre-processed. EPI images were realigned and smoothed (6-mm kernel). GRASE images were realigned, followed by subtraction of label and control to obtain perfusion-weighted images and smoothing.

Gas manipulation data analysis: EPI images obtained during gas manipulation were entered into a general linear model (GLM) with two conditions (room air and hypercapnia), excluding the first minute after each transition. The model involved a regressor to account for the image being acquired during control or label conditions and a linear drift term. The regression coefficients were used to compute signal baseline level (BOLD0) and signal change induced by HC (ΔBOLD). Perfusion-weighted images were entered into a similar GLM without regressors to compute perfusion difference signal at baseline and during HC (ΔS0 and ΔS1) and HC-induced signal change. M-maps were calculated via Eq. 1 (α=0.18, β=1.5) (6).

$$M=\frac{\frac{\triangle BOLD}{BOLD_{0}}}{1-\left(\frac{\triangle S_{1}}{\triangle S_{0}}\right)^{\left(\alpha-\beta\right)}} [1]$$

Task activation data analysis: EPI images obtained during the activation paradigm were entered into a GLM with two conditions (rest and task) and a regressor to model the ASL effect. Perfusion-weighted images were analogously, excluding the regressor. Contrast images comparing task and rest were obtained. Activated regions were identified using uncorrected p-value<0.005. Regions of interest (ROI) were defined in the left and right motor cortex (MC), including voxels that appeared active in both analyses, and used to extract signal time-courses. Relative CMRO2 maps were computed via Eq. 2.

$$rCMRO_{2}=\left(\frac{\triangle S_{1}}{\triangle S_{0}}\right)^{1-\frac{\alpha}{\beta}}\left(1-\frac{\frac{\triangle BOLD}{BOLD_{0}}}{M}\right)^{\frac{1}{\beta}} [2]$$

Results and Discussion

Fig. 2 shows EPI and background-suppressed GRASE images, signal difference images and M-map, computed from the gas manipulation data. Gray matter M values averaged across the 5 subjects (Table 1) are in very good agreement with reported results (6). Fig. 3 shows maps of signal change induced by bilateral finger-tapping, revealing activation in bilateral MC. Relative CMRO2 time-courses in the activated regions are shown in Fig. 4. Group-averaged calibrated fMRI parameters, including mean M values for the left and right MC ROIs, as well as task-evoked signal changes (%BOLD, %ΔS and rCMRO2) are shown in Table 1. Mean evoked CMRO2 estimates are 12 and 13% for left and right MC, respectively. Assuming a resting state value for gray matter CMRO2 of 146 mmol/100g/min (7), these relative values would correspond to absolute changes of 17 and 19 mmol/100g/min.

Conclusion

This work demonstrates the feasibility of measuring task-evoked CMRO2 changes using a dual-acquisition PCASL sequence that combines a background-suppressed 3D readout optimized for ASL with 2D-EPI.

Acknowledgements

Grants RYC-2010-07161, NIH RO1 HLI22754.

References

1. Davis TL, Kwong KK, Weisskoff, et al. Calibrated functional MRI: Mapping the dynamics of oxidative metabolism. Proc Natl Acad Sci USA. 1998;95:1834-39.

2. Ghariq E, Chappell MA, Schmid S, et al. Effect of background suppression on the sensitivity of dual-echo arterial spin labeling MRI for BOLD and CBF signal changes. Neuroimage. 2014;103:316-22.

3. Wesolowski R, Gowland PA, Francis ST. Double acquisition background suppressed (DABS) FAIR at 3T and 7T: advantages for simultaneous BOLD and CBF acquisition. Proc Intl Soc Mag Reson Med 17. 2009; p.1526.

4. Vidorreta M, Wang Z, Rodriguez I, et al. Comparison of 2D and 3D single-shot ASL perfusion fMRI sequences. Neuroimage. 2013; 66:662-71.

5. Rodgers ZB, Englund EK, Langham MC, et al. Rapid T2-and susceptometry-based CMRO2 quantification with interleaved TRUST (iTRUST). Neuroimage. 2015;106:441-50.

6. Gauthier CJ, Hoge RD. A generalized procedure for calibrated MRI incorporating hyperoxia and hypercapnia. Hum Brain Mapp. 2013;34:1053-69.

7. Bulte DP, Kelly M, Germuska M, et al. Quantitative measurement of cerebral physiology using respiratory-calibrated MRI. Neuroimage. 2012; 60:582-91.

Figures

Fig.1: Pulse sequence diagram showing presaturation (PreSat), background suppression (BS) and PCASL pulses, followed by the single shot 3D-GRASE and 2D-EPI readouts, separated by a delay time of 1.3s. The timing of BS pulses has been optimized to null the static tissue signal at the beginning of the 3D-GRASE readout.

Fig. 2: Data obtained during the gas manipulation run from a representative subject, showing: (a) Baseline background suppressed 3D-GRASE images; (b) Baseline EPI images; (c) Baseline perfusion-weighted images; (d) Perfusion signal increase induced by HC; (e) BOLD signal increase induced by HC; (f) Computed M-map.

Fig.3: Maps of signal change induced by the bilateral finger tapping task in a single subject, showing activation in bilateral motor cortex: (a) BOLD (% of baseline); (b) Perfusion (% of baseline); (c) Relative CMRO2 change during activation.

Fig. 4: Relative CMRO2 time courses during the fMRI paradigm computed from the BOLD and perfusion signals extracted from the activated regions in the left and right motor cortex, in the subject shown in Fig. 3.

Table 1: Calibrated fMRI results, shown as group means (standard deviation).



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3729