In lung cancer, diffusion weighted MRI (DW-MRI) is used to derive an apparent diffusion coefficient (ADC) from malignant lesions, as ADC is recognised as a potential biomarker for both lesion characterisation and response evaluation (1, 2). Standardization of DW protocols for use in multicentre clinical trials is challenging due to inter-vendor variations. Initially, it is informative to explore the effect of common sequence variations related to inter-vendor deviations (e.g fat suppression technique, diffusion gradient mode and echo time) on the ADC reliability (reproducibility/ repeatability) individually at each centre before protocol development in multi-centre clinical trials.

To assess the reliability of ADC estimates of lung lesions using two optimised DW protocols together and separately by mimicking common vendor-related sequence variations.8 patients with malignant lung masses, having at least one lesion >2 cm, were recruited for a single centre study with their informed consent. They were scanned twice with a time separation ranging from 1 hour to 1 week. Anatomical and DW imaging were performed using a Siemens Avanto 1.5T MR scanner with 2 anterior phased-array body coils. Axial T₁-weighted and T₂-weighted images were acquired. Two DW protocols (DW-STIR and DW-SPAIR) were employed with the following parameters: free-breathing, single shot echo planar imaging, FoV=380x273mm, acquired matrix=128x92, pixel=3x3mm, parallel imaging factor=2, PE direction=AP, TR=8500ms (DW-STIR) or 9000ms (DW-SPAIR), TE=72ms (DW-STIR) or 83ms (DW-SPAIR), short time inversion recovery (STIR) fat suppression (DW-STIR) or spectral adiabatic inversion recovery (SPAIR) fat suppression (DW-SPAIR), three scan trace (DW-STIR) or three orthogonal directions (DW-SPAIR), single spin echo, b=100, 500, 800 s/mm², 25 to 40 slices, slice thickness=5mm, NSA=1 repeated 4 times. A mono-exponential decay model for all b-values from all the acquisitions was applied for the production of the calculated ADC maps and computed DW images using in house-built software using the geometric mean image of the three acquired directions of both DW protocols assuming isotropic diffusion. Regions of interest (ROIs) were drawn on the computed DW-MRI (b=800 s/mm²) (3) encompassing the larger area of the lesion on three central contiguous slices. The calculated ADC values of all pixels within these slices were combined in order to give a volume of interest (VOI) per examination and the mean ADC value of the whole volume was recorded. The ADC comparisons between the DW protocols and the repeatability of each protocol were assessed by performing a t-test and by using coefficient of variation (CoV) and Bland Altman plots.T-tests showed no difference comparing the ADC estimates of the DW-STIR and DW-SPAIR protocols (p=0.93) and by comparing the two baselines of each protocol separately (p=0.52 for DW-STIR and p=0.58 for DW-SPAIR). This result was in agreement with the CoV (Table 1) and the Bland Altman plot (Figure 1a) of the ADC estimates comparing the two DW protocols exhibiting only one outlier. By removing this outlier, the CoV was 3.5%. The Bland Altman plots (Figure 1b, 1c) presented a good repeatability of the ADC estimates of each protocol exhibiting similar CoVs.In multi-centre clinical trials, there are vendor specific variations of the DW sequence parameters in measurement protocols across the centres. These technical factors are most commonly related to the fat suppression technique, the diffusion gradient mode and the gradient performance. We explored these key variations comparing STIR, which is less sensitive to B₀ inhomogeneity, but results in loss of signal and a T₁-weighted contrast contribution, to SPAIR, which is sensitive to B₀ inhomogeneity and provides higher SNR due to its spectrally selective fat suppression character. Different diffusion gradient modes were also used for further optimisation of the fat suppression techniques resulting in different echo times (TE). These differences impact on ADC estimates with the CoVs being increased between the 2 protocols (CoV=7.7%) compared to each protocol individually (CoV_DW-STIR=3.7% and CoV_DW-SPAIR=4,6%). However, no significant influence was observed on the absolute ADC estimates using different fat suppression techniques, diffusion gradient modes and TEs in this study.In a single centre study, several of the most common protocol variations in DW-MRI were employed. The selection of the fat suppression technique, the diffusion gradient mode and the different TE increased the CoV of the ADC estimates of lung lesions using together the 2 DW protocols to using each one separately, although absolute values did not differ significantly.