Paul Baron^{1}, Dirk H.J. Poot^{1,2,3}, Piotr A. Wielopolski^{1}, Edwin H.G. Oei^{1}, and Juan A. Hernandez-Tamames^{1}

Diffusion weighted STEAM with UTE readout is a promising acquisition method to measure diffusion in tissues with short T2 and T2*, such as tendons, ligaments and menisci. However, the accuracy of the ADC obtained with this method has not been studied before. Using experiments and Bloch simulations we show that the ADC can be biased, especially when a short TR is used, and that this bias depends on T1 and T2. Randomization of the diffusion gradient direction reduces the bias, providing clear suggestions to improve acquisition and/or post processing that reduces the ADC bias.

**Introduction**

Diffusion weighting imaging of
tissues such as tendons, ligaments and menisci is challenging due their short T2
and T2* relaxation times. A diffusion weighted (DW) STEAM-based 3D-Cones UTE (DW-UTE)
sequence has been proposed^{1} to combine the benefits of a stimulated echo
acquisition (i.e. to minimize T2 decay for high b-values) with ultrashort UTE
echo times (to minimize T2* decay during readout). However, the accuracy of the Apparent
Diffusion Coefficient (ADC) with this sequence has not been investigated and may
be affected by undesired coherence pathways. In this study, we investigate the accuracy
of the calculated ADC in phantoms with different T1 and T2 relaxation times. Additionally,
we propose a method that improves the accuracy of the ADC by randomly changing
the direction of the diffusion weighting gradient in each TR.

Experimental setup: To achieve a range of T1 and T2, six agar gels
(2% agar) were prepared and doped with MnCl_{2} concentrations from 10
mg/L to 80 mg/L. The gels were contained in 50 ml falcon tubes and placed with the long axis parallel to the B0
field in a 8-ch T/R knee coil in a 3T whole body human MR scanner (Discovery
MR750, GE Healthcare). The addition of agar and MnCl_{2} is not expected to influence the diffusion coefficient of water^{2}, which was verified by
measuring the ADC with conventional DW-SE-EPI.

Acquisition and ADC
calculation: The DW-UTE images were acquired axially. UTE-parameters (also see Fig. 1): TE=32
µs, FA=10°, 8 spokes per diffusion preparation, FOV = 15×15×19.2 cm, Matrix =128×128×64.
Diffusion parameters: T_{mix} = 120 ms, gradient duration (δ)= 4ms in right-left-direction
with amplitudes: 10, 20, 30, and 40 mT/m resulting in b-values 14, 57, 127, and
227 s/mm^{2}. To correct T1 relaxation effects^{1}, one acquisition was
acquired with the last STEAM RF pulse as +90° (‘flip up’), and -90°
(‘flip down’). Separate acquisitions were acquired with TR’s of 200, 300, and
500 ms. ADC was calculated by subtracting the two (flip up/down) complex
images^{1}, and subsequently fitting the magnitude as function of b-value with a
mono-exponential function^{3}. Additionally, acquisitions were conducted to
measure the T1 and T2 times of the gels. (T1:
IR-SE,TR=4s,TE=9ms,ST=10mm, 19×TI’s in range 50-3500 ms, T2: SE, TR=500ms, ST=5mm, 14×TE’s in
range 5-120 ms).

Evaluation and correction: To evaluate the influence of T1, T2, and TR on
the calculated ADC, Bloch simulations of the DW-UTE acquisitions were
performed. It included diffusion of an ensemble of spins with input parameters
T1 and T2 as measured in the gels and ADC=2.20×10^{-3} mm^{2}/s.
The ADC was computed with the same procedure as for the experiments. To reduce signal
formation from pathways longer than TR we simulated randomly alternating the orientation
of diffusion gradients for each TR, averaging over 50 TRs.

**Results**

1. M. Carl et al. Proc. ISMRM 2016, abstract number 3021

2. I. Lavdas et al. A phantom for diffusion-weighted MRI (DW-MRI). J. Magn Reson Imag. 2013;38(1):173-9

3. D.H.J. Poot et al. Detecting statistically significant differences in quantitative MRI experiments, applied to diffusion tensor imaging. IEEE Transactions on Medical Imaging, 2015;34(5):1164-1176

Fig.1: Schematic of the DW-STEAM 3D Cones UTE sequence. The
last RF pulse in the diffusion preparation part (marked in pink) was acquired
once with +90° (flip up) and once with -90° (flip down).

Fig.2: (a) T1 map and (b) T2 map. Image (a) has
numbers indicating gel number. (c-e) The calculated ADC for different TR times. (f)
Illustration of the corrected ADC map using the offsets given by the Bloch simulation

Table 1: Column I and II show the
mean T1 and T2 values measured for the gels, respectively. Column III are the
ADC values measured with DW-SE-EPI. Additionally the experimentally (Exp) and simulated
(Sim) ADC (assumed ADC= 2.2×10^{-3} mm^{2}/s) with DW-UTE for TR = 500, 300, and 200 ms is shown.

Fig.3: Influence of using random diffusion gradient orientation
per TR on the calculated ADC. rand =
random diffusion orientations used.