Diffusion‐weighted MRI of the lung at 3T evaluated using echo‐planar‐based and single‐shot turbo spin‐echo‐based acquisition techniques for radiotherapy applications

Abstract Purpose To compare single‐shot echo‐planar (SS‐EPI)‐based and turbo spin‐echo (SS‐TSE)‐based diffusion‐weighted imaging (DWI) in Non‐Small Cell Lung Cancer (NSCLC) patients and to characterize the distributions of apparent diffusion coefficient (ADC) values generated by the two techniques. Methods Ten NSCLC patients were enrolled in a prospective IRB‐approved study to compare and optimize DWI using EPI and TSE‐based techniques for radiotherapy planning. The imaging protocol included axial T2w, EPI‐based DWI and TSE‐based DWI on a 3 T Philips scanner. Both EPI‐based and TSE‐based DWI sequences used three b values (0, 400, and 800 s/mm2). The acquisition times for EPI‐based and TSE‐based DWI were 5 and 8 min, respectively. DW‐MR images were manually coregistered with axial T2w images, and tumor volume contoured on T2w images were mapped onto the DWI scans. A pixel‐by‐pixel fit of tumor ADC was calculated based on monoexponential signal behavior. Tumor ADC mean, standard deviation, kurtosis, and skewness were calculated and compared between EPI and TSE‐based DWI. Image distortion and ADC values between the two techniques were also quantified using fieldmap analysis and a NIST traceable ice‐water diffusion phantom, respectively. Results The mean ADC for EPI and TSE‐based DWI were 1.282 ± 0.42 × 10−3 and 1.211 ± 0.31 × 10−3 mm2/s. The average skewness and kurtosis were 0.14 ± 0.4 and 2.43 ± 0.40 for DWI‐EPI and −0.06 ± 0.69 and 2.89 ± 0.62 for DWI‐TSE. Fieldmap analysis showed a mean distortion of 13.72 ± 8.12 mm for GTV for DWI‐EPI and 0.61 ± 0.4 mm for DWI‐TSE. ADC values obtained using the diffusion phantom for the two techniques were within 0.03 × 10−3 mm2/s with respect to each other as well as the established values. Conclusions Diffusion‐weighted turbo spin‐echo shows better geometrical accuracy compared to DWI‐EPI. Mean ADC values were similar with both acquisitions but the shape of the histograms was different based on the skewness and kurtosis values. The impact of differences in respiratory technique on ADC values requires further investigation.


| INTRODUCTION
Magnetic resonance imaging (MRI) in the lung is challenging due to breathing and cardiac motion. At the same time, MRI in the lung is also appealing because any pathology in the lung will have a higher proton density than surrounding normal tissue and therefore a higher MR signal with a strong inherent contrast against the dark background. Recently, there has been tremendous interest in the use of diffusion-weighted MRI in lung cancer for diagnosis, staging and response assessment. 1-7 Diffusionweighted imaging (DWI) is typically acquired using an echo-planar imaging (EPI)-based acquisition that provides high signal-to-noise ratio (SNR) and is very fast to minimize the effects of physiological motions arising from respiration, cardiac or any bulk motion. 8 In an EPI acquisition, the echo trains are formed by gradient pulses, which do not rephase spins that have become dephased due to intravoxel field inhomogeneity. Therefore, the EPI signal can be greatly reduced in the presence of large differences in magnetic susceptibility at air/tissue interfaces due to rapid intravoxel dephasing and the extremely short resultant T2*. 9 In addition to signal loss, field inhomogeneity results in image distortion when spins encoded by frequency are mapped to the incorrect location. The spatial shift is proportional to the ratio of the field inhomogeneity over the voxel (in Hz) to the voxel acquisition bandwidth (BW) and can be several mm or more in EPI where voxel bandwidths are low. Furthermore, the effect of susceptibility differences scales with field strength and is therefore more severe on 3 T MR scanners. On the other hand, if the echo train is formed by radiofrequency pulses, such as turbo spin-echo (TSE)/ fast spin-echo (FSE) based acquisition, 10 the effect of static field inhomogeneities will be refocused, increasing the signal at a given echo time and permitting longer sampling windows, higher voxel bandwidths and less spatial distortion than EPI-based DWI.

2.A | DWI-TSE sequence implementation
In DWI-TSE, diffusion gradients are applied before and after the 180-degree refocusing pulse to allow for diffusion acquisition using a TSE sequence. The single-shot TSE-diffusion pulse sequence provided by Philips healthcare incorporates the following features in order to shorten echo train length, and minimize blurring: (a) averaging of modulus data instead of complex data to minimize the effect of phase differences between echoes, (b) a short refocusing pulse that has a less sharp profile than the standard refocusing pulse but reduces echo spacing, and (c) sensitivity encoding-based parallel imaging. 12

2.B | Phantom study
A National Institute of Standards and Technology traceable, temperature-controlled ice-water diffusion phantom (High Precision Devices, Inc, Boulder, CO, USA) was scanned using both single-shot echo-planar (SS-EPI)-based and turbo spin-echo (SS-TSE)-based DWIbased acquisitions at 0°C. 13 The phantom was scanned using a 16element head coil on the 3 T Philips Ingenia scanner with four different b values: 0, 500, 900, and 2000, with the established scan parameter values. 13 The phantom consisted of 13 vials containing 30 ml of polymer polyvinylpyrrolidone in aqueous solution at various concentrations. The phantom scan was repeated twice 2 month apart, and two sets of ADC measurements were performed for both EPI and TSE acquisition on each day. A 1 cm diameter region of interest (ROI) was defined in the center of each vial to calculate the mean ADC and standard deviation for each technique.

2.C | Patient selection and Imaging protocol
Ten patients (eight men, two women; median age: 64 yr (range 51-74 yr) with locally advanced Non-Small Cell Lung Cancer undergoing chemoradiation were enrolled in a prospective IRB-approved study to undergo DWI using both the SS-EPI-and SS-TSE-based techniques. Table 1 shows the patient characteristics, such as age, diagnosis, TNM status, 14

2.D | DWI analysis
Diffusion signal decay was modeled exponentially as a function of b value where b is a factor representing diffusion weighting. Quantification of this signal loss is performed by calculating the ADC from: where S(b) is signal intensity measured for a given b value, and S(0) is the signal intensity for b = 0 s/mm 2 . Anatomical T2-w images and b = 0 DW images were imported into MIM VISTA TM for contouring EPI-DWI and TSE-DWI tumor volumes. Tumor contours drawn on the anatomic T2-w images were transferred to the DWI images after manual registration. Figure 1 shows EPI-and TSE-based DWI and T2w MRI of an example case. In the EPI images, susceptibilityrelated signal "pile up" can be seen at the tumor edge (arrows).

2.E | Distortion evaluation
To evaluate the extent of patient-specific distortions in lung DWI images, B 0 maps (in Hz), were derived from two gradient echo images with different echo times and obtained for three cases. The change in MR signal phase from one echo to the next is proportional to both the field inhomogeneity in that voxel and the echo time difference. 15 B 0 maps were converted to pixel shift maps based on the BW, as shown by the equations below.
where ΔB 0 is the B0 field distortion in Hz, φ 1 and φ 2 are the phase values of two images, TE 1 and TE 2 are the echo times of the two images, γ is the gyromagnetic ratio, Δυ x is the pixel size (mm) in the phase encoding direction and BWx is the pixel BW. The phase images are wrapped between −π and +π and were unwrapped using an two-dimensional phase unwrapping algorithm available in FSL. 16 For EPI-DWI, the pixel shifts predominantly occur in the phase encoding direction whereas for TSE-DWI, the shifts occur in the frequency encoding direction. The pixel BW along the phase encoding direction for EPI-DWI is calculated as where ES is the echo spacing and ETL is the echo train length.
Please note that the echo spacing and echo train length values obtained were calculated after applying for SENSE, partial Fourier or phase oversampling. The gross tumor volumes (GTVs) drawn on the T2-w image were overlaid on the field map after registration between T2 and the magnitude image. The mean, standard deviation, minimum, and maximum values of pixel shifts within the GTV ROI were then calculated. A pixel-by-pixel fit of the ADC, based on monoexponential behavior, was calculated using equation 1 and histograms were generated.

2.F | Image analysis
From each ADC histogram, the following descriptive parameters were calculated: mean, median, standard deviation, kurtosis, and skewness. These parameters were compared between EPI-and TSE-based DWI for the entire population. The statistical correlation between EPI-DWI and TSE-DWI was determined using the paired student's t test. A P value less than 0.05 was considered statistically significant.

3.A | Phantom measurements
Phantom images showed less susceptibility distortion with TSE-DWI as compared with EPI-DWI for all b values. This can be observed in Fig. 2 where the circular cross-sections of the embedded tubes appear distorted in the EPI images. ADC statistics within each ROI were calculated for two same day acquisitions and two different dates (four sets of ADC values for both EPI and TSE acquisitions). These are plotted as a function of vial concentration along with the published values for the phantom in Fig. 2(c). The average ADC difference for all vials between EPI-and TSE-DWI was −0.01 ± 0.008 × 10 −3 mm 2 /s with ADC TSE systematically higher than EPI, but not statistically significant. The average ADC differences for all vials for EPI-DWI and TSE-DWI, with respect to published reference ADC values, were 0.00 ± 0.011 and 0.02 ± 0.012 × 10 −3 mm 2 /s, respectively. On average, both EPI and TSE values were within 0 ± 3% and 3 ± 2%, respectively, with the pub-

3.B | Patient study
SNR comparison performed between the EPI-and TSE-DWI image varied between patients as shown in Table 3. SNR value for one

3.C | Distortion analysis
Distortion analysis showed that the mean shift in the GTVs for the three patients were 13.72 ± 8.12 mm for EPI-DWI, with a mean average minimum and maximum shift of −24.21 and 36.9 mm, respectively. For TSE-DWI, the mean, minimum, and maximum shift over both GTVs was 0.61 ± 0.4, −1.08, and 1.65 mm, respectively. Figure 5 shows the field maps in Hz for one example case. As shown in Table 1, the pixel BW of EPI was 34 Hz whereas that of TSE was 757 Hz. For the same pixel size, the extent of distortion in EPI would be 20× that of the TSE acquisitions. One of the patients (#10) had tumor next to heart. The presence of beating heart next to the tumor affected phase unwrapping of this patient resulting in inaccuracies in the phase maps. Without this patient, the mean, minimum, and

| DISCUSSION
In this study, the tumor ADC histograms derived from an SS-EPIbased acquisition and an SS-TSE-based acquisition were compared.
SS-TSE-DWI was superior to EPI in minimizing susceptibility artifacts as shown by the field map analysis. With TSE-DWI, the geometric accuracy is of the order of a standard anatomic T2w imaging.
Minimum distortion allowed easy registration and transfer of contours between T2w and TSE-DWI. With EPI, there was often a shift in the tumor position and lumping of pixels at the tumor-air interface due to susceptibility artifacts. This distortion required registration. The low geometric accuracy of EPI-DWI makes it challenging to incorporate the imaging modality into radiation therapy treatment planning.
T A B L E 3 SNR ratio between EPI and TSE-based acquisition (SNR EPI /SNR TSE ).