Arterial Spin Labeling (ASL) is a non-invasive technique to obtain quantitative maps of perfusion. At higher magnetic fields, it benefits from both higher signal-to-noise ratio and longer T₁ but could suffer from higher RF power deposition and thus temperature increase. The latter issue has however not been characterized in animals. In this study, the specific absorption rate (SAR) delivered to a rat was measured in vivo at 9.4T using continuous ASL (CASL) and pseudo-continuous ASL (pCASL) with and without a dedicated labeling coil.

Experiments:

Nine rats were anesthetized with 1-2% isoflurane. Their respiration rate was maintained at around 50brpm. The animal’s temperature was measured continuously at three different locations using fiber optic probes placed as follows: inside the brain (Fig.1), subcutaneously near the labeling plane (between carotids) and in the rectum.

Experiments were performed on a 9.4T scanner (Bruker Biospec, AVIII-HD) with a volume transmit/surface (phased-array) receive coil configuration. A separate RF-transmit coil (diameter 23mm) was placed under the animal’s throat for labeling. After an anatomical TurboRARE-T2 scan, three sets of ASL experiments were performed: a pCASL and a CASL (called ‘volCASL’) sequence using the volume transmit RF coil for labeling, and a CASL sequence (called ‘labCASL’) using the separate labeling coil. Within each set of experiments, a range of mean label B₁ amplitudes (B_1ave) were applied: 3, 5 and 7µT for volCASL and labCASL, and 3 and 5µT for pCASL. To evaluate the reproducibility, the 3µT pCASL scan was repeated at the end of the experiment. All other scans were performed once, in random order.

For each ASL scan, labeling pulses were applied in the neck (2cm from the isocenter) during 3s followed by a 300ms post-labeling delay and a single-shot EPI acquisition (in-plane resolution=234µm², slice thickness=1mm, 1 slice, TE=22ms, TR=4s). 30 pairs of label/control images were acquired within 4 minutes. Sequence specific parameter settings were:

- The pCASL labeling pulse train consisted of Hanning window shaped RF pulses¹ lasting 400µs, repeated every 800µs. For a given B_1ave, note that the B_1RMS (root mean square B₁) is √3 times higher for pCASL than for volCASL.

- During labCASL, the RF power during the control experiment was turned off, since the labeling coil induced no magnetization transfer effects in the imaged slice. Inversion efficiency (IE, n=3) was measured 5mm downstream of the labeling plane with a flow compensated, ASL encoded FLASH sequence for each B_1ave value and each of the three ASL sequences. Both label and control inter-pulse phases were optimized prior to the pCASL IE measurement².

Data Processing:
The temperature time-course was corrected for baseline drift. SAR was calculated with: $$$SAR=S_{scan}*C_{tissue}$$$, where S_scan is the fitted slope of the temperature increase and C_tissue the specific heat capacity of tissue³ (3664 J/kg/K). IE was calculated with: $$$IE=\frac{M_C-M_L}{2 M_C}*100$$$, where M_C and M_L are respectively the complex magnetizations from the control and label experiments. The relative IE difference between carotids was computed to evaluate the asymmetry of labeling.

Fig.2 shows temperature timecourses for the different techniques at identical B_1ave=5µT and for the anatomical scan as a reference. For this animal, temperature increased up to 1.3°C (brain) and 1.4°C (carotids) within 4 minutes. Since SAR is proportional to B_1RMS² (Fig.4) and B_1RMS,pCASL is higher than B_1RMS,volCASL, this explains the SAR difference between pCASL and volCASL at equal mean labeling power B_1ave (Fig.2-3-4). However the measured ratio was lower than the factor 3 expected. SAR is lower for labCASL than for pCASL by at least 45%(brain)-75%(carotids), and lower than volCASL by 32%(brain)-62%(carotids) (Fig.3).

The SAR derived from the two 3µT-pCASL scans performed on each animal were comparable (Fig.3), suggesting experimental stability. The SAR standard deviation across rats is mainly due to the positioning of the probes: we observed lower SAR deeper in the brain.

Considering these SAR values, a compromise must be found between SAR, inversion efficiency, acquisition time and/or coil configuration. As shown previously⁴, IE increases with B_1ave (Fig.5a). Reducing B_1ave to less than 3µT led to lower IE (<70%) and asymmetrical labeling between carotids (Fig5b). For B_1ave≥5µT, acceptable IE and low asymmetry were obtained. Note that pCASL reached higher IE for lower B_1ave values than both CASL methods but still needed B_1ave>3µT to produce symmetrical labeling.

This study shows that ASL experiments with standard parameters (B_1ave=3.5-5µT) at 9.4T lead to non-negligible increases in brain temperature. This should be taken into account when using them for long scanning periods or with diseased animals such as stroke or trauma. SAR effect can be dramatically reduced by using a dedicated labeling coil.