Measurement of T1 and T2 relaxation times of the pancreas at 7 T using a multi-transmit system

Objective To determine T1 and T2 relaxation times of healthy pancreas parenchyma at 7 T using a multi-transmit system. Materials and methods Twenty-six healthy subjects were scanned with a 7 T MR system using eight parallel transceiver antennas, each with two additional receive loops. A Look-Locker sequence was used to obtain images for T1 determination, while T2 was obtained from spin-echo images and magnetic resonance spectroscopy measurements with different echo times. T1 and T2 times were calculated using a mono-exponential fit of the average magnitude signal from a region of interest in the pancreas and were tested for correlation with age. Results The age range of the included subjects was 21–72 years. Average T1 and T2 relaxation times in healthy pancreas were 896 ± 149 ms, and 26.7 ± 5.3 ms, respectively. No correlation with age was found. Conclusion T1 and T2 relaxation times of the healthy pancreas were reported for 7 T, which can be used for image acquisition optimization. No significant correlations were found between age and T1 or T2 relaxation times of the pancreas. Considering their low standard deviation and no observable age dependence, these values may be used as a baseline to study potentially pancreatic tissue affected by disease.


Introduction
Magnetic resonance (MR) imaging is an imaging modality with good soft tissue contrast and high sensitivity for detection of pancreatic cancer, which facilitates early tumor detection [1,2]. Modern MRI techniques can visualize biliary and pancreatic ductal systems noninvasively, have a high sensitivity for tumor detection, and, unlike endoultrasound imaging (EUS) with high sensitivity (up to 89%) and specificity (up to 99%) for detection of small-tumor pancreatic cancer [3][4][5], reveal the pancreatic three-dimensional anatomy, possible invasion into surrounding tissue or vascular involvement [6]. However, it is still difficult to distinguish chronic pancreatitis and pancreatic carcinoma using T 1 -weighted and/or T 2 -weighted MR images on 1.5 and 3 Tesla (T) [7], even when gadolinium contrast agents are used [8].
Higher magnetic field strengths, such as 7 T, offer higher signal-to-noise (SNR) and contrast-to-noise ratios (CNR), as well as higher spatial and spectral resolutions [9]. However, artifacts are more profound at higher field strengths, particularly in the abdominal region, mainly due to the reduced wavelength in the body that leads to an inhomogeneous B 1 + field distribution and signal voids [10,11]. In addition, the local specific absorption rate (SAR) increases with increasing field strength [12], leading to longer repetition times needed and thus longer acquisitions. Using a multi-transmit system, the B 1 + efficiency can be optimized within a region of interest (ROI). This can be achieved by optimizing the transmit phases for each channel in the body coil. As a result, improved local B 1 + homogeneity and magnitude can be achieved [10].
In addition to the technical challenges that MRI at higher magnetic field strengths brings, high-quality diagnostic imaging of pancreatic cancer requires development and optimization of imaging protocols. Moreover, a new baseline assessment of the images obtained at ultra-high field is required. Therefore, knowledge of relaxation times (T 1 and T 2 ) is essential. These characteristics are known to change substantially with field strength and this change cannot be predicted accurately with theoretical calculations due to the complex tissue behavior [13]. Consequently, although T 1 and T 2 values have been well established for many tissues, including pancreatic tissue at 1.5 T and 3 T [14][15][16], T 1 and T 2 relaxation times of pancreatic tissue at 7 T remain unknown.
The purpose of this study was to determine the mean T 1 and T 2 relaxation times of healthy pancreas parenchyma over a wide age range at 7 T to assess baseline levels to develop optimized imaging protocols at this magnetic field strength.

Study population and hardware
Twenty-six healthy subjects were scanned with a 7 T MR system (Philips, Best, The Netherlands) after providing a written informed consent. A multi-transmit system with eight parallel transmit channels was used, where each channel was connected to a transmit-receive fractionated dipole antenna (MR Coils BV, Drunen, The Netherlands) [17]. Each antenna had 2 additional receive loops integrated in its housing (16 in total) and were interfaced to a 16-channel receiver box (Philips, Best, The Netherlands).

Image acquisition
All scans were acquired in a 'feet first' supine position, with the eight channels positioned symmetrically around the abdomen, approximately centered at the height of the pancreas. First, a gradient echo image was obtained for anatomy localization and optimization steps (2D T1 fast field echo (FFE), FA = 15°, TR/TE = 10/5 ms, FOV = 704 × 704 mm 2 , 0.7 × 0.7 × 10 mm 3 voxels). Part of the optimization was RF phase shimming, in which the RF phase of each antenna is steered to maximize and homogenize the B 1 + field in the pancreas region (ROI). An in-house developed MATLAB ( © MATLAB 2015b) script was used for loading a dynamic imaging series (fast field echo (FFE), FA = 4°, TR/TE = 30/1.68 ms, 0:25 min total acquisition time, 1.1 × 1.1 × 10 mm 3 voxels), drawing an ROI including the whole pancreas, and optimizing the phase of each element, while keeping a fixed power using a numerical optimization algorithm. After this, image-based B 0 shimming was performed. Subsequently, a B 1 + map was acquired (fast field echo (FFE), FA = 50°, TR/TE = 50-250/2.2 ms, 1:29 min total acquisition time, 1.2 × 1.2 × 10 mm 3 voxels) for power optimization in the pancreas region. Finally, to guarantee optimal performance in the T 2 -weighted MRI protocol, the power was adjusted to reach a B 1 + of 10 µT within the pancreas. A Look-Locker sequence was used for T 1 relaxation determination in one slice containing the pancreas in 22 subjects (inversion recovery turbo field echo (TFE), FA = 3°, TR/TE = 40/1.68 ms, 20 measurement times separated by 100 ms, during a 6 s cycle, 8:33 min total acquisition time, 4 × 4 × 10 mm 3 voxels).

Data processing
Mono-exponential decay with a non-linear least squares method was used to fit T 2 relaxation times [21]; for the T 2 -weighted images magnitude signals were used and for the STEAM data the area under the curve of the water peak was used. To fit the T 2 relaxation time from the lower refocusing angle series, the equivalent echo times (TE equiv ) were used. This TE equiv was the calculated apparent echo time (based on the average T 1 and T 2 relaxation times determined with the 180° refocusing pulse and STEAM) as if a 180° refocusing angle had been used [22]. An ROI was manually drawn within the area with the best B 1 + homogeneity and B 1 + closest to 100%, excluding big vessels (Fig. 1b). The position of the ROI in the pancreas was consistent within subjects, however, between subjects the location of the ROI varied between head, body and tail of the pancreas. The signal in the ROIs was subsequently averaged and fitted. Average T 1 and T 2 relaxation times were calculated. Finally, age dependency was examined using all three datasets.

Statistical analysis
An unpaired t test was performed to compare T 2 relaxation fitting methods (1) and (2). Since previously published research has shown a significant linear dependence of age with T 1 relaxation times [23] and it is known that the pancreas tissue composition is changing with age [16], a linear regression analysis for age with respect to T 1

Results
Out of 26 subjects, 22 T 1 and 23 T 2 datasets could be used for further analysis. The discarded datasets had insufficient image quality, caused by motion or not enough B 1 + amplitude in the pancreas region. Average age of the subjects was 37 years (range 21-72 years) for the T 1 datasets and 40 years (range 24-72 years) for the T 2 datasets. T 1 relaxation times (mean ± std) for all 22 subjects are summarized in Fig. 2, the average T 1 relaxation time was 896 ± 149 ms. An example of an B 1 + map is shown in Fig. 3, the mean ratio of the measured B 1 + over the expected B 1 + value in the ROI across the subjects was 0.92 ± 0.14. Linear regression analysis showed no correlation between age and T 1 relaxation time (slope = 1.1, offset = 856, R 2 = 0.01, p = 0.6).
Using method (1) (T 2 -weighted image series with a 180° refocusing pulse) and (2) (using STEAM spectroscopy) for T 2 determinations gave an average T 2 relaxation time of 28.1 ms (n = 3) and 25.6 (n = 4), respectively. An unpaired t test proved no significant difference (p = 0.6) between the results of the two methods. Combining these results as a reference value gave an average T 2 of 26.7 ± 5.3 ms.
Method (3) resulted in a lower average T 2 relaxation time of 19.5 ± 3.8 ms (n = 16) as expected due to the lower refocusing angle used. Fitted T 2 relaxation times for all 23 subjects are summarized in Fig. 4 Figure 5 shows examples of the monoexponential fits for all three methods, with the signal of the first acquisition normalized to 1, for three single subjects.

. No correlation between age
Multi 2D T 2 -weighted single-shot turbo spin-echo images FA 90°, refocusing flip angle 35°, TE 80 ms (TE equiv = 31 ms), TR 10 s, TSE factor of 121, SPAIR fat suppression, voxel size of 1.5 × 1.5 × 5 mm 3 (Fig. 6) were obtained. TE equiv close to the average T 2 relaxation time of pancreas and TR more than 5 times longer ( ~ 11 times longer) than the average T 1 relaxation time for improved contrast. The full field of view of the pancreas can be covered. However, signal voids can be still observed outside the region of interest (not B 1 + optimized region), which are more pronounced in spin-echo-based sequences.

Discussion
In this study, we determined normal pancreas T 1 and T 2 relaxation times in 26 healthy subjects aged 21-72 years at 7 T. T 1 and T 2 relaxation times for pancreas at 7 T-to our knowledge-have not been reported before. However, they have been reported for pancreas in healthy subjects at 1.5 T and 3 T. In the literature, it can be found that T 1 relaxation The red arrow points at the pancreas. Field inhomogeneities with signal voids outside the pancreas (blue arrow) can be observed. The yellow arrows point at regions where overtipping occurs (~ 180°) due to B 1 + field inhomogeneity effects, leading to signal voids times increase and T 2 relaxation times decrease with increasing field strengths [13,23]. De Bazelaire et al. [13] used an inversion recovery method and different inversion times for T 1 measurements and a multiple spin-echo (SE) technique with different echo times for T 2 measurements in six healthy subjects, reported T 1 values of 584 ± 14 ms for 1.5 T and 725 ± 71 ms for 3 T and T 2 values of 46 ± 6 ms for 1.5 T and 43 ± 7 ms for 3 T, and concluded that pancreas T 1 relaxation times increased and T 2 relaxation times decreased with increasing field strength. Tirkes et al. [23] reported a mean T 1 relaxation time of 797 ms at 3 T for healthy pancreas (n = 53, sequence: dual flip angle 3D gradient echo). A slightly higher mean T 1 of 987 ± 52 ms and mean T 2 of 50 ± 3 ms at 3 T was reported by Chhor et al. [15] (n = 6, sequence: Look-Locker for T 1 and T 2 -prep for T 2 measurements). Our 7 T results confirm the expected trend of increasing T 1 and decreasing T 2 relaxation times with age.
In addition, T 1 and T 2 values are dependent on tissue composition [24]. In this study, we did not investigate tissue differences between subjects. Therefore, we could not confirm this at 7 T. Tirkes et al. [23] also found a weak correlation between age and T 1 relaxation time in the pancreas at 3 T, which we could not observe in our measurements at 7 T. The relatively young mean age of them could explain the lack of correlation (if any).
The use of standard TSE sequences attenuates the effect of molecular diffusion by the use of multiple refocusing pulses, the effect becoming stronger as the refocusing pulses get closer to each other [14,24]. In addition, differences in imaging methods and determining T 2 relaxation times, a more than doubled magnetic field strength, different age ranges, and the morphology of each subject, combined with the complex characteristics of tissues, may be responsible for the large decrease of T 2 relaxation time found at 7 T.
For T 1 measurements, a Look-Locker sequence is generally used with a TR (including the 6 s per cycle) of more than 5 times the T 1 value of pancreas tissue, which is required to measure the T 1 value accurately. Using T 2 -weighted imaging with full 180° refocusing angle and magnitude signals to fit the water peaks from the MR spectroscopy measurements, T 2 relaxation times for pancreas could be correctly determined. A spin-echo sequence such as the one used here in method (1) is not practical to use in such large groups of subjects. The high SAR deposition in the body due to the large refocusing angles requires long repetition times leading to long scanning times, to keep SAR under the guideline levels. Since the mentioned SAR levels was the worst-case possible SAR and it is known that the dipole array will never exceed a local SAR of 10 W/kg when driven at a time averaged power of 3 W per channel driving all eight antennas at any phase setting, the SAR could be overestimated but were always within (conservative) limits. In addition, there are hardware limitations associated to these high-power demanding sequences. T 2 imaging with a small refocusing pulse ensures a more practical sequence; with shorter scan times and lower SAR levels. However, T 2 imaging with a 35° refocusing pulse is affected by T 1 -weighted stimulated echoes during the signal formation [25], which leads to a lengthening of the signal-intensity decay. This resulted in longer T 2 relaxation times, which after correction with the equivalent echo times (to a 180° refocusing angle), resulted in slightly lower T 2 relaxation values and a comparable variability to the results found with the nominal 180° refocusing pulse and the spectroscopy data, as was seen in Fig. 4. No other explanation could be found for the underestimation of the results of method (3). Age-dependency testing using data from method (3) is, therefore, only suitable for assessing the relative differences of T 2 of the pancreas with respect to age.
Inter-subject variability in the measured relaxation times can be caused by different factors. Variations within characteristics of (pancreatic) tissue, and their change with age were different for individuals, even between men and women (which we did not take into account). As Sato et al. [16] describes in their paper, the pancreas changes with increasing age; characteristic changes with age are pancreatic atrophy, lobulation and fatty degeneration. These changes, such as the fraction of fat in the pancreas would lead to altered relaxation times.
Results were fitted over scans with different echo times since a single sequence with multiple echo times given by the turbo factor resulted to be too sensitive to motion and breathing artifacts that would result in blurring of the image. This leads to slight differences in anatomy position between echo times. Therefore, the delineated ROIs did not correspond to exactly the same anatomical position, leading to outliers and a higher variation in the raw data, which can influence the fitting. In our study, data were obtained from all regions in the pancreas. Since the ROIs were drawn in the area with the best B 1 + homogeneity, the location was different between subjects. Tirkes et al. [23] showed a slight but not significant difference between the relaxation times measured in head, body or tail. As we have not accounted for these differences, this might have been another factor affecting the standard deviation in our calculations.
Lastly, imaging the abdomen at higher field, especially with T 2 -weighted sequences, is challenging. The scans are more sensitive to artifacts given the increased B 0 and B 1 + inhomogeneities (even with good B 0 and B 1 + shimming) found at higher field strengths, leading to greater standard deviations sources.
The measured T 1 and T 2 relaxation times for pancreas at 7 T reported here is a starting point for pancreatic research at higher field strengths to improve abdominal MR imaging, and may be a base for further use of MR techniques to noninvasive pancreatic diagnostics particularly in early stages of disease.

Conclusion
T 1 and T 2 relaxation times of the healthy pancreas were reported for 7 T, which can be used for image acquisition optimization. No significant correlation was found between age and T 1 or T 2 relaxation times of the pancreas.